This document summarizes a research paper that proposes a new model called Quantum-Inspired Neuro-Evolutionary Computation (QINEA-BR) for optimizing the configuration of neural networks. QINEA-BR extends an existing quantum-inspired evolutionary algorithm (QIEA-BR) by adding a binary representation to allow optimization of categorical neural network parameters like input variables and hidden neuron count. The paper shows that QINEA-BR can successfully perform binary classification on a credit risk evaluation problem, outperforming other models. It divides the neural network parameters into binary and numerical parts to optimize using a hybrid binary-real chromosome representation in QIEA-BR.
IJERA (International journal of Engineering Research and Applications) is International online, ... peer reviewed journal. For more detail or submit your article, please visit www.ijera.com
Exploring Quantum Supremacy in Access Structures of Secret Sharing by Coding ...Ryutaroh Matsumoto
Quantum supremacy or quantum advantage is the potential ability of quantum computing devices to solve problems that classical computers practically cannot (Wikipedia). The speaker recently found that quantum computation can realize secret sharing schemes that cannot be realized by any classical computation. That finding was enabled by combinatorial studies of quantum error-correcting codes and classical secret sharing. This talk introduces those studies to non-specialists with mathematical backgrounds.
Reference: arXiv:1803.10392
IJERA (International journal of Engineering Research and Applications) is International online, ... peer reviewed journal. For more detail or submit your article, please visit www.ijera.com
Exploring Quantum Supremacy in Access Structures of Secret Sharing by Coding ...Ryutaroh Matsumoto
Quantum supremacy or quantum advantage is the potential ability of quantum computing devices to solve problems that classical computers practically cannot (Wikipedia). The speaker recently found that quantum computation can realize secret sharing schemes that cannot be realized by any classical computation. That finding was enabled by combinatorial studies of quantum error-correcting codes and classical secret sharing. This talk introduces those studies to non-specialists with mathematical backgrounds.
Reference: arXiv:1803.10392
Blind Audio Source Separation (Bass): An Unsuperwised Approach IJEEE
Audio processing is an area where signal separation is considered as a fascinating works, potentially offering a vivid range of new scope and experience in professional and personal context. The objective of Blind Audio Source Separation is to separate audio signals from multiple independent sources in an unknown mixing environment. This paper addresses the key challenges in BASS and unsupervised approaches to counter these challenges. Comparative performance analysis of Fast-ICA algorithm and Convex Divergence ICA for Blind Source Separation is presented with the help of experimental result. Result reflects Convex Divergence ICA with α=-1 gives more accurate estimate in comparison of Fast ICA..
On Fixed Point error analysis of FFT algorithmIDES Editor
In this correspondence the analysis of overall
quantization loss for the Fast Fourier Transform (FFT)
algorithms is extended to the case where the twiddle factor
word length is different from the register word length. First,
a statistical noise model to predict the Quantization error
after the multiplication of two quantized signals, of different
precision, is presented. This model is then applied to FFT
algorithms. Simulation results, that corroborate the
theoretical analysis, are then presented.
Dear Students
Ingenious techno Solution offers an expertise guidance on you Final Year IEEE & Non- IEEE Projects on the following domain
JAVA
.NET
EMBEDDED SYSTEMS
ROBOTICS
MECHANICAL
MATLAB etc
For further details contact us:
enquiry@ingenioustech.in
044-42046028 or 8428302179.
Ingenious Techno Solution
#241/85, 4th floor
Rangarajapuram main road,
Kodambakkam (Power House)
http://www.ingenioustech.in/
Generating images from a text description is as challenging as it is interesting. The Adversarial network
performs in a competitive fashion where the networks are the rivalry of each other. With the introduction of
Generative Adversarial Network, lots of development is happening in the field of Computer Vision. With
generative adversarial networks as the baseline model, studied Stack GAN consisting of two-stage GANS
step-by-step in this paper that could be easily understood. This paper presents visual comparative study of
other models attempting to generate image conditioned on the text description. One sentence can be related
to many images. And to achieve this multi-modal characteristic, conditioning augmentation is also
performed. The performance of Stack-GAN is better in generating images from captions due to its unique
architecture. As it consists of two GANS instead of one, it first draws a rough sketch and then corrects the
defects yielding a high-resolution image.
Naver learning to rank question answer pairs using hrde-ltcNAVER Engineering
The automatic question answering (QA) task has long been considered a primary objective of artificial intelligence.
Among the QA sub-systems, we focused on answer-ranking part. In particular, we investigated a novel neural network architecture with additional data clustering module to improve the performance in ranking answer candidates which are longer than a single sentence. This work can be used not only for the QA ranking task, but also to evaluate the relevance of next utterance with given dialogue generated from the dialogue model.
In this talk, I'll present our research results (NAACL 2018), and also its potential use cases (i.e. fake news detection). Finally, I'll conclude by introducing some issues on previous research, and by introducing recent approach in academic.
Pairwise and Problem-Specific Distance Metrics in the Linkage Tree Genetic Al...Martin Pelikan
The linkage tree genetic algorithm (LTGA) identifies linkages between problem variables using an agglomerative hierarchical clustering algorithm and linkage trees. This enables LTGA to solve many decomposable problems that are difficult with more conventional genetic algorithms. The goal of this paper is two-fold: (1) Present a thorough empirical evaluation of LTGA on a large set of problem instances of additively decomposable problems and (2) speed up the clustering algorithm used to build the linkage trees in LTGA by using a pairwise and a problem-specific metric.
http://medal.cs.umsl.edu/files/2011001.pdf
IJERA (International journal of Engineering Research and Applications) is International online, ... peer reviewed journal. For more detail or submit your article, please visit www.ijera.com
MULTI-OBJECTIVE ENERGY EFFICIENT OPTIMIZATION ALGORITHM FOR COVERAGE CONTROL ...ijcseit
Many studies have been done in the area of Wireless Sensor Networks (WSNs) in recent years. In this kind of networks, some of the key objectives that need to be satisfied are area coverage, number of active sensors and energy consumed by nodes. In this paper, we propose a NSGA-II based multi-objective algorithm for optimizing all of these objectives simultaneously. The efficiency of our algorithm is demonstrated in the simulation results. This efficiency can be shown as finding the optimal balance point among the maximum coverage rate, the least energy consumption, and the minimum number of active nodes while maintaining the connectivity of the network
Robust Watermarking through Dual Band IWT and Chinese Remainder TheoremjournalBEEI
CRT was a widely used algorithm in the development of watermarking methods. The algorithm produced good image quality but it had low robustness against compression and filtering. This paper proposed a new watermarking scheme through dual band IWT to improve the robustness and preserving the image quality. The high frequency sub band was used to index the embedding location on the low frequency sub band. In robustness test, the CRT method resulted average NC value of 0.7129, 0.4846, and 0.6768 while the proposed method had higher NC value of 0.7902, 0.7473, and 0.8163 in corresponding Gaussian filter, JPEG, and JPEG2000 compression test. Meanwhile the both CRT and proposed method had similar average SSIM value of 0.9979 and 0.9960 respectively in term of image quality. The result showed that the proposed method was able to improve the robustness and maintaining the image quality.
IJERA (International journal of Engineering Research and Applications) is International online, ... peer reviewed journal. For more detail or submit your article, please visit www.ijera.com
COMPARISON OF WAVELET NETWORK AND LOGISTIC REGRESSION IN PREDICTING ENTERPRIS...ijcsit
Enterprise financial distress or failure includes bankruptcy prediction, financial distress, corporate performance prediction and credit risk estimation. The aim of this paper is that using wavelet networks innon-linear combination prediction to solve ARMA (Auto-Regressive and Moving Average) model problem.ARMA model need estimate the value of all parameters in the model, it has a large amount of computation.Under this aim, the paper provides an extensive review of Wavelet networks and Logistic regression. Itdiscussed the Wavelet neural network structure, Wavelet network model training algorithm, Accuracy rateand error rate (accuracy of classification, Type I error, and Type II error). The main research opportunity exist a proposed of business failure prediction model (wavelet network model and logistic regression
model). The empirical research which is comparison of Wavelet Network and Logistic Regression on training and forecasting sample, the result shows that this wavelet network model is high accurate and the overall prediction accuracy, Type Ⅰerror and Type Ⅱ error, wavelet networks model is better thanlogistic regression model.
Quantum Evolutionary Algorithm for Solving Bin Packing Probleminventionjournals
International Journal of Engineering and Science Invention (IJESI) is an international journal intended for professionals and researchers in all fields of computer science and electronics. IJESI publishes research articles and reviews within the whole field Engineering Science and Technology, new teaching methods, assessment, validation and the impact of new technologies and it will continue to provide information on the latest trends and developments in this ever-expanding subject. The publications of papers are selected through double peer reviewed to ensure originality, relevance, and readability. The articles published in our journal can be accessed online.
Blind Audio Source Separation (Bass): An Unsuperwised Approach IJEEE
Audio processing is an area where signal separation is considered as a fascinating works, potentially offering a vivid range of new scope and experience in professional and personal context. The objective of Blind Audio Source Separation is to separate audio signals from multiple independent sources in an unknown mixing environment. This paper addresses the key challenges in BASS and unsupervised approaches to counter these challenges. Comparative performance analysis of Fast-ICA algorithm and Convex Divergence ICA for Blind Source Separation is presented with the help of experimental result. Result reflects Convex Divergence ICA with α=-1 gives more accurate estimate in comparison of Fast ICA..
On Fixed Point error analysis of FFT algorithmIDES Editor
In this correspondence the analysis of overall
quantization loss for the Fast Fourier Transform (FFT)
algorithms is extended to the case where the twiddle factor
word length is different from the register word length. First,
a statistical noise model to predict the Quantization error
after the multiplication of two quantized signals, of different
precision, is presented. This model is then applied to FFT
algorithms. Simulation results, that corroborate the
theoretical analysis, are then presented.
Dear Students
Ingenious techno Solution offers an expertise guidance on you Final Year IEEE & Non- IEEE Projects on the following domain
JAVA
.NET
EMBEDDED SYSTEMS
ROBOTICS
MECHANICAL
MATLAB etc
For further details contact us:
enquiry@ingenioustech.in
044-42046028 or 8428302179.
Ingenious Techno Solution
#241/85, 4th floor
Rangarajapuram main road,
Kodambakkam (Power House)
http://www.ingenioustech.in/
Generating images from a text description is as challenging as it is interesting. The Adversarial network
performs in a competitive fashion where the networks are the rivalry of each other. With the introduction of
Generative Adversarial Network, lots of development is happening in the field of Computer Vision. With
generative adversarial networks as the baseline model, studied Stack GAN consisting of two-stage GANS
step-by-step in this paper that could be easily understood. This paper presents visual comparative study of
other models attempting to generate image conditioned on the text description. One sentence can be related
to many images. And to achieve this multi-modal characteristic, conditioning augmentation is also
performed. The performance of Stack-GAN is better in generating images from captions due to its unique
architecture. As it consists of two GANS instead of one, it first draws a rough sketch and then corrects the
defects yielding a high-resolution image.
Naver learning to rank question answer pairs using hrde-ltcNAVER Engineering
The automatic question answering (QA) task has long been considered a primary objective of artificial intelligence.
Among the QA sub-systems, we focused on answer-ranking part. In particular, we investigated a novel neural network architecture with additional data clustering module to improve the performance in ranking answer candidates which are longer than a single sentence. This work can be used not only for the QA ranking task, but also to evaluate the relevance of next utterance with given dialogue generated from the dialogue model.
In this talk, I'll present our research results (NAACL 2018), and also its potential use cases (i.e. fake news detection). Finally, I'll conclude by introducing some issues on previous research, and by introducing recent approach in academic.
Pairwise and Problem-Specific Distance Metrics in the Linkage Tree Genetic Al...Martin Pelikan
The linkage tree genetic algorithm (LTGA) identifies linkages between problem variables using an agglomerative hierarchical clustering algorithm and linkage trees. This enables LTGA to solve many decomposable problems that are difficult with more conventional genetic algorithms. The goal of this paper is two-fold: (1) Present a thorough empirical evaluation of LTGA on a large set of problem instances of additively decomposable problems and (2) speed up the clustering algorithm used to build the linkage trees in LTGA by using a pairwise and a problem-specific metric.
http://medal.cs.umsl.edu/files/2011001.pdf
IJERA (International journal of Engineering Research and Applications) is International online, ... peer reviewed journal. For more detail or submit your article, please visit www.ijera.com
MULTI-OBJECTIVE ENERGY EFFICIENT OPTIMIZATION ALGORITHM FOR COVERAGE CONTROL ...ijcseit
Many studies have been done in the area of Wireless Sensor Networks (WSNs) in recent years. In this kind of networks, some of the key objectives that need to be satisfied are area coverage, number of active sensors and energy consumed by nodes. In this paper, we propose a NSGA-II based multi-objective algorithm for optimizing all of these objectives simultaneously. The efficiency of our algorithm is demonstrated in the simulation results. This efficiency can be shown as finding the optimal balance point among the maximum coverage rate, the least energy consumption, and the minimum number of active nodes while maintaining the connectivity of the network
Robust Watermarking through Dual Band IWT and Chinese Remainder TheoremjournalBEEI
CRT was a widely used algorithm in the development of watermarking methods. The algorithm produced good image quality but it had low robustness against compression and filtering. This paper proposed a new watermarking scheme through dual band IWT to improve the robustness and preserving the image quality. The high frequency sub band was used to index the embedding location on the low frequency sub band. In robustness test, the CRT method resulted average NC value of 0.7129, 0.4846, and 0.6768 while the proposed method had higher NC value of 0.7902, 0.7473, and 0.8163 in corresponding Gaussian filter, JPEG, and JPEG2000 compression test. Meanwhile the both CRT and proposed method had similar average SSIM value of 0.9979 and 0.9960 respectively in term of image quality. The result showed that the proposed method was able to improve the robustness and maintaining the image quality.
IJERA (International journal of Engineering Research and Applications) is International online, ... peer reviewed journal. For more detail or submit your article, please visit www.ijera.com
COMPARISON OF WAVELET NETWORK AND LOGISTIC REGRESSION IN PREDICTING ENTERPRIS...ijcsit
Enterprise financial distress or failure includes bankruptcy prediction, financial distress, corporate performance prediction and credit risk estimation. The aim of this paper is that using wavelet networks innon-linear combination prediction to solve ARMA (Auto-Regressive and Moving Average) model problem.ARMA model need estimate the value of all parameters in the model, it has a large amount of computation.Under this aim, the paper provides an extensive review of Wavelet networks and Logistic regression. Itdiscussed the Wavelet neural network structure, Wavelet network model training algorithm, Accuracy rateand error rate (accuracy of classification, Type I error, and Type II error). The main research opportunity exist a proposed of business failure prediction model (wavelet network model and logistic regression
model). The empirical research which is comparison of Wavelet Network and Logistic Regression on training and forecasting sample, the result shows that this wavelet network model is high accurate and the overall prediction accuracy, Type Ⅰerror and Type Ⅱ error, wavelet networks model is better thanlogistic regression model.
Quantum Evolutionary Algorithm for Solving Bin Packing Probleminventionjournals
International Journal of Engineering and Science Invention (IJESI) is an international journal intended for professionals and researchers in all fields of computer science and electronics. IJESI publishes research articles and reviews within the whole field Engineering Science and Technology, new teaching methods, assessment, validation and the impact of new technologies and it will continue to provide information on the latest trends and developments in this ever-expanding subject. The publications of papers are selected through double peer reviewed to ensure originality, relevance, and readability. The articles published in our journal can be accessed online.
https://quspeakerseries9.splashthat.com/
Lecture 1: Dr.Jorg Kientz
In this talk we outline the use of Machine Learning algorithms and their potential application. We focus on Deep Neural Networks. The aim is to outline different network architectures. Then, we wish to find a way of choosing the architecture that best fits from a model validation perspective. This approach is illustrated with examples.
A New Model for Credit Approval Problems: A Neuro-Genetic System with Quantum...Anderson Pinho
This paper presents a new model for neuro-evolutionary systems. It is a new quantum-inspired evolutionary algorithm with binary-real representation (QIEA-BR) for evolution of a neural network. The proposed model is an extension of the QIEA-R developed for numerical optimization. The Quantum-Inspired Neuro-Evolutionary Computation model (QINEA-BR) is able to completely configure a feed-forward neural network in terms of selecting the relevant input variables, number of neurons in the hidden layer and all existent synaptic weights. QINEA-BR is evaluated in a benchmark problem of financial credit evaluation. The results obtained demonstrate the effectiveness of this new model in comparison with other machine learning and statistical models, providing good accuracy in separating good from bad customers.
ABSTRACT: Once introduced the fundamental ideas of quantum computing, we will discuss the possibilities offered by quantum computers in machine learning.
BIO: Davide Pastorello obtained an M.Sc. in Physics (2011) and a Ph.D. in Mathematics (2014) from Trento University. After serving at the Dept. of Mathematics and DISI in Trento, he is currently an assistant professor at the Dept. of Mathematics, University of Bologna. His main research interests concern the mathematical aspects of quantum information theory, quantum computing, and quantum machine learning.
Control chart pattern recognition using k mica clustering and neural networksISA Interchange
Automatic recognition of abnormal patterns in control charts has seen increasing demands nowadays in manufacturing processes. This paper presents a novel hybrid intelligent method (HIM) for recognition of the common types of control chart pattern (CCP). The proposed method includes two main modules: a clustering module and a classifier module. In the clustering module, the input data is first clustered by a new technique. This technique is a suitable combination of the modified imperialist competitive algorithm (MICA) and the K-means algorithm. Then the Euclidean distance of each pattern is computed from the determined clusters. The classifier module determines the membership of the patterns using the computed distance. In this module, several neural networks, such as the multilayer perceptron, probabilistic neural networks, and the radial basis function neural networks, are investigated. Using the experimental study, we choose the best classifier in order to recognize the CCPs. Simulation results show that a high recognition accuracy, about 99.65%, is achieved.
Min-based qualitative possibilistic networks are one of the effective tools for a compact representation of decision problems under uncertainty. The exact approaches for computing decision based on possibilistic networks are limited by the size of the possibility distributions.
Generally, these approaches are based on possibilistic propagation algorithms. An important step in the computation of the decision is the transformation of the DAG into a secondary structure, known as the junction trees. This transformation is known to be costly and represents a difficult problem. We propose in this paper a new approximate approach for the computation
of decision under uncertainty within possibilistic networks. The computing of the optimal optimistic decision no longer goes through the junction tree construction step. Instead, it is performed by calculating the degree of normalization in the moral graph resulting from the merging of the possibilistic network codifying knowledge of the agent and that codifying its preferences.
A Novel Approach to Mathematical Concepts in Data Miningijdmtaiir
-This paper describes three different fundamental
mathematical programming approaches that are relevant to
data mining. They are: Feature Selection, Clustering and
Robust Representation. This paper comprises of two clustering
algorithms such as K-mean algorithm and K-median
algorithms. Clustering is illustrated by the unsupervised
learning of patterns and clusters that may exist in a given
databases and useful tool for Knowledge Discovery in
Database (KDD). The results of k-median algorithm are used
to collecting the blood cancer patient from a medical database.
K-mean clustering is a data mining/machine learning algorithm
used to cluster observations into groups of related observations
without any prior knowledge of those relationships. The kmean algorithm is one of the simplest clustering techniques
and it is commonly used in medical imaging, biometrics and
related fields.
Brief introduction of neural network including-
1. Fitting Tool
2. Clustering data with a self-organising map
3. Pattern Recognition Tool
4. Time Series Toolbox
Algoritmos Genéticos com Inspiração Quântica e Representação Mista Aplicados ...Anderson Pinho
Apresentação final de dissertação para obtenção do título de mestre em métodos de apoio a decisão com ênfase em métodos estatísticos e inteligência computacional.
PLANO DE NEGÓCIOS: RESTAURANTE EMPÓRIO APRAZÍVELAnderson Pinho
Este plano de negócios objetivou estudar e analisar o mercado de restaurantes no Rio de Janeiro para, posteriormente, estruturar um novo conceito de prestação de serviço de alimentação customizado. Tendo em vista os hábitos e cultura, preocupada com bem-estar, saúde e forma física da população do Rio de Janeiro, identificou-se a oportunidade de um restaurante cuja oferta alimentar se enquadre ao perfil e à necessidade individual dos consumidores. O restaurante propõe acompanhar o cliente em seu programa de dieta. Este programa terá
objetivos distintos conforme a necessidade de cada indivíduo, como por exemplo, perda, manutenção ou aumento de peso, dieta específica para diabéticos e hipertensos e etc. Haverá o acompanhamento nutricional dos clientes através de um relacionamento estreito com forte investimento em ações de CRM, data mining e pós-venda. O restaurante montará programas customizados para cada cliente dentro
de uma oferta pré-elaborada por nutricionistas. Além de criar a maximização de satisfação e bem-estar, a oferta customizada gera alta percepção de valor. Isso mostra que o restaurante passará a fazer parte do cotidiano do cliente, assim, será mais que um restaurante será uma solução em alimentação.
Estratégias de Retenção de Clientes no Marketing de RelacionamentoAnderson Pinho
As atuais práticas do mercado provam que é mais lucrativos para a empresa o investimento no relacionamento com o cliente com foco na retenção, do que outros resultados obtidos por incrementos na carteira. Conhecendo-se os riscos associados ao desligamento de um cliente, as empresas podem direcionar estratégias de marketing na busca da retenção para os clientes com maiores chances de evasão. Isto otimiza os resultados de marketing, pois clientes com baixo risco de perda deixam de ser atingidos.
Empresas no ramo de seguros de plano de saúde devem se preocupar em criar valor para seus assegurados neste mercado que pouco se assemelha ao transacional, como ocorre nos grandes varejos. Clientes que optam por adquirir um plano de saúde em sua grande maioria buscam um relacionamento de médio para longo prazo. Seguradoras que souberem utilizar uma gestão inteligente do negócio agregando valor contínuo, identificando e antecipando-se a possível perda de um cliente, estarão contribuindo para o sucesso da empresa. Por estes motivos, nosso objetivo aqui será de desenvolver um modelo estatístico que identifique uma possível futura perda do cliente, para que as empresas foquem seus recursos na retenção deste, maximizando os resultados da corporação e satisfação do cliente.
Previsão de Inadimplência por Redes Neurais Artificiais na Presença de Baixo ...Anderson Pinho
A análise de inadimplência creditícia tem sido objeto de estudo para diversos pesquisadores e profissionais. Mesmo em casos onde há ocorrência da inadimplência é um evento quase raro, empresas concessionárias de empréstimos precisam se apoiar em boas práticas de concessão de empréstimos por diversos motivos. O presente artigo visa estudar o fenômeno da inadimplência através de redes neurais artificiais neste ambiente de baixa inadimplência. Para isto, utilizaremos uma base de dados real de uma cooperativa de crédito brasileira para estudo e aplicação das técnicas aqui apresentadas.
Modelagem de Cross Selling no Marketing de Relacionamento por Máquinas de Vet...Anderson Pinho
Analistas de marketing estão freqüentemente em busca de novas maneiras de se vender mais produtos e serviços para os mesmos clientes. É de grande debate em processos KDD (Knowledge Discovery Database) a discussão sobre modelagem de cross-selling, ou venda cruzada, pela aplicação de tradicionais técnicas de inteligência computacional e estatísticas como Redes Neurais e Regressão Logística. O presente artigo abordará a técnica conhecida como Máquinas de Vetor de Suporte (MVS) para resolução de problemas desta natureza em contrapartida as demais já utilizadas. No presente artigo, utilizaremos uma base de dados real de uma empresa vendedora de motos a pessoas físicas como aplicação da técnica. MVS trata-se de um tipo especial de redes neurais alimentadas adiante, e representa real vantagem às outras técnicas, pois permite ao analista o domínio na determinação das superfícies de separação entre grupos, em nosso caso, de interessados e não interessados na compra de outra moto. Esta predição é interessante do ponto de vista de marketing, pois permite a antecipação pela empresa de um desejo do cliente, viabilizando uma ação de marketing pró-ativa para incremento de receita através da venda focada.
Mineração de Dados com Mapas de Kohonen: Uma Abordagem no Setor FinanceiroAnderson Pinho
O presente artigo objetivará a aplicação de Mapas de Kohonen na identificação de perfis de clientes, ou clusters, como estratégias de diferenciação no marketing empresarial. Clientes que decidem por investir em ações na Bolsa de Valores de São Paulo (Bovespa) podem optar por ações de empresas nos mais diversos ramos de atividade. Se gestores do marketing souberem explicitar o conhecimento necessário para diferenciar clientes de acordo com o setor em que operam na bolsa de valores, estes poderão incorporar este conhecimento nos processos da empresa, maximizando lucros e resultados. Mapas de Kohonen representam uma classe de modelos de redes neurais denominados mapas auto-organizáveis e seu potencial de aplicação em processos de descoberta de conhecimento representa alguma vantagem competitiva. Em nosso caso de aplicação isto significa separar clientes de acordo com suas concentrações de operações na Bovespa nos 10 possíveis mercados de ações. Os resultados mostrarão que clientes podem ser agrupados em 1 de 6 clusters diferentes, cada um com características bastante singulares.
Estratégias de Marketing através de Análise RFV do Cliente e Lógica Fuzzy, no...Anderson Pinho
O presente artigo objetivará a criação de um modelo de inferência fuzzy o qual classifique corretamente futuros clientes evasivos para uma empresa. Em estratégias de marketing, é de grande dúvida para a empresa quais clientes abordar numa campanha, ou quais clientes apresentam maiores chances de evasão. Para responderem a isto, muitos pesquisadores têm recorrido a informações de recência, freqüência e valor do cliente, na mineração de conhecimento valioso o qual possa ser utilizado. Lógica fuzzy será um diferencial competitivo na identificação destes clientes, pois permitirá a criação de um modelo preciso de classificação, alinhado ao conhecimento lingüístico explícito do especialista de marketing.
Análise RFV do Cliente na Otimização de Estratégias de Marketing: Uma Abordag...Anderson Pinho
O presente artigo objetivará a evolução de regras
de decisão por Algoritmos Genéticos as quais
classifiquem corretamente futuros clientes
evasivos para a empresa. Em estratégias de
marketing, é de grande dúvida para a empresa
quais clientes abordar numa campanha, ou quais
clientes apresentam maiores chances de evasão.
Para responderem a isto, muitos pesquisadores
têm recorrido a informações de recência,
freqüência e valor do cliente, na mineração de
conhecimento valioso o qual possa ser utilizado.
Algoritmos Genéticos demonstrará um diferencial
competitivo na explicitação deste conhecimento,
pois permitirá uma simples integração com
processos empresaria, de fácil entendimento para
o usuário.
Previsão de Inadimplência por Redes Neurais ArtificiaisAnderson Pinho
A análise de inadimplência creditícia tem sido objeto de estudo para diversos pesquisadores e profissionais. Mesmo em casos onde há ocorrência da inadimplência é um evento quase raro, empresas concessionárias de empréstimos precisam se apoiar em boas práticas de concessão de empréstimos por diversos motivos. O presente artigo visa estudar o fenômeno da inadimplência através de redes neurais artificiais neste ambiente de baixa inadimplência. Para isto, utilizaremos uma base de dados real de uma cooperativa de crédito brasileira para estudo e aplicação das técnicas aqui apresentadas.
Modelagem de Cross-Selling no Marketing de Relacionamento por Máquinas de Vet...Anderson Pinho
Analistas de marketing estão freqüentemente em busca de novas maneiras de se vender mais produtos e serviços para os mesmos clientes. É de grande debate em processos KDD (Knowledge Discovery Database) a discussão sobre modelagem de cross-selling, ou venda cruzada, pela aplicação de tradicionais técnicas de inteligência computacional e estatísticas como Redes Neurais e Regressão Logística. O presente artigo abordará a técnica conhecida como Máquinas de Vetor de Suporte (MVS) para resolução de problemas desta natureza em contrapartida as demais já utilizadas. No presente artigo, utilizaremos uma base de dados real de uma empresa vendedora de motos a pessoas físicas como aplicação da técnica. MVS trata-se de um tipo especial de redes neurais alimentadas adiante, e representa real vantagem às outras técnicas, pois permite ao analista o domínio na determinação das superfícies de separação entre grupos, em nosso caso, de interessados e não interessados na compra de outra moto. Esta predição é interessante do ponto de vista de marketing, pois permite a antecipação pela empresa de um desejo do cliente, viabilizando uma ação de marketing pró-ativa para incremento de receita através da venda focada.
Estratégias de Marketing através de Análise RFV do Cliente e Lógica Fuzzy, no...Anderson Pinho
O presente artigo objetivará a criação de um modelo de inferência fuzzy o qual classifique corretamente futuros clientes evasivos para uma empresa. Em estratégias de marketing, é de grande dúvida para a empresa quais clientes abordar numa campanha, ou quais clientes apresentam maiores chances de evasão. Para responderem a isto, muitos pesquisadores têm recorrido a informações de recência, freqüência e valor do cliente, na mineração de conhecimento valioso o qual possa ser utilizado. Lógica fuzzy será um diferencial competitivo na identificação destes clientes, pois permitirá a criação de um modelo preciso de classificação, alinhado ao conhecimento lingüístico explícito do especialista de marketing.
Análise RFV do Cliente na Otimização de Estratégias de Marketing: Uma Abordag...Anderson Pinho
O presente artigo objetivará a evolução de regras
de decisão por Algoritmos Genéticos as quais
classifiquem corretamente futuros clientes
evasivos para a empresa. Em estratégias de
marketing, é de grande dúvida para a empresa
quais clientes abordar numa campanha, ou quais
clientes apresentam maiores chances de evasão.
Para responderem a isto, muitos pesquisadores
têm recorrido a informações de recência,
freqüência e valor do cliente, na mineração de
conhecimento valioso o qual possa ser utilizado.
Algoritmos Genéticos demonstrará um diferencial
competitivo na explicitação deste conhecimento,
pois permitirá uma simples integração com
processos empresaria, de fácil entendimento para
o usuário.
Attending a job Interview for B1 and B2 Englsih learnersErika906060
It is a sample of an interview for a business english class for pre-intermediate and intermediate english students with emphasis on the speking ability.
Improving profitability for small businessBen Wann
In this comprehensive presentation, we will explore strategies and practical tips for enhancing profitability in small businesses. Tailored to meet the unique challenges faced by small enterprises, this session covers various aspects that directly impact the bottom line. Attendees will learn how to optimize operational efficiency, manage expenses, and increase revenue through innovative marketing and customer engagement techniques.
Discover the innovative and creative projects that highlight my journey throu...dylandmeas
Discover the innovative and creative projects that highlight my journey through Full Sail University. Below, you’ll find a collection of my work showcasing my skills and expertise in digital marketing, event planning, and media production.
Digital Transformation and IT Strategy Toolkit and TemplatesAurelien Domont, MBA
This Digital Transformation and IT Strategy Toolkit was created by ex-McKinsey, Deloitte and BCG Management Consultants, after more than 5,000 hours of work. It is considered the world's best & most comprehensive Digital Transformation and IT Strategy Toolkit. It includes all the Frameworks, Best Practices & Templates required to successfully undertake the Digital Transformation of your organization and define a robust IT Strategy.
Editable Toolkit to help you reuse our content: 700 Powerpoint slides | 35 Excel sheets | 84 minutes of Video training
This PowerPoint presentation is only a small preview of our Toolkits. For more details, visit www.domontconsulting.com
RMD24 | Retail media: hoe zet je dit in als je geen AH of Unilever bent? Heid...BBPMedia1
Grote partijen zijn al een tijdje onderweg met retail media. Ondertussen worden in dit domein ook de kansen zichtbaar voor andere spelers in de markt. Maar met die kansen ontstaan ook vragen: Zelf retail media worden of erop adverteren? In welke fase van de funnel past het en hoe integreer je het in een mediaplan? Wat is nu precies het verschil met marketplaces en Programmatic ads? In dit half uur beslechten we de dilemma's en krijg je antwoorden op wanneer het voor jou tijd is om de volgende stap te zetten.
Business Valuation Principles for EntrepreneursBen Wann
This insightful presentation is designed to equip entrepreneurs with the essential knowledge and tools needed to accurately value their businesses. Understanding business valuation is crucial for making informed decisions, whether you're seeking investment, planning to sell, or simply want to gauge your company's worth.
The world of search engine optimization (SEO) is buzzing with discussions after Google confirmed that around 2,500 leaked internal documents related to its Search feature are indeed authentic. The revelation has sparked significant concerns within the SEO community. The leaked documents were initially reported by SEO experts Rand Fishkin and Mike King, igniting widespread analysis and discourse. For More Info:- https://news.arihantwebtech.com/search-disrupted-googles-leaked-documents-rock-the-seo-world/
Premium MEAN Stack Development Solutions for Modern BusinessesSynapseIndia
Stay ahead of the curve with our premium MEAN Stack Development Solutions. Our expert developers utilize MongoDB, Express.js, AngularJS, and Node.js to create modern and responsive web applications. Trust us for cutting-edge solutions that drive your business growth and success.
Know more: https://www.synapseindia.com/technology/mean-stack-development-company.html
Buy Verified PayPal Account | Buy Google 5 Star Reviewsusawebmarket
Buy Verified PayPal Account
Looking to buy verified PayPal accounts? Discover 7 expert tips for safely purchasing a verified PayPal account in 2024. Ensure security and reliability for your transactions.
PayPal Services Features-
🟢 Email Access
🟢 Bank Added
🟢 Card Verified
🟢 Full SSN Provided
🟢 Phone Number Access
🟢 Driving License Copy
🟢 Fasted Delivery
Client Satisfaction is Our First priority. Our services is very appropriate to buy. We assume that the first-rate way to purchase our offerings is to order on the website. If you have any worry in our cooperation usually You can order us on Skype or Telegram.
24/7 Hours Reply/Please Contact
usawebmarketEmail: support@usawebmarket.com
Skype: usawebmarket
Telegram: @usawebmarket
WhatsApp: +1(218) 203-5951
USA WEB MARKET is the Best Verified PayPal, Payoneer, Cash App, Skrill, Neteller, Stripe Account and SEO, SMM Service provider.100%Satisfection granted.100% replacement Granted.
Cracking the Workplace Discipline Code Main.pptxWorkforce Group
Cultivating and maintaining discipline within teams is a critical differentiator for successful organisations.
Forward-thinking leaders and business managers understand the impact that discipline has on organisational success. A disciplined workforce operates with clarity, focus, and a shared understanding of expectations, ultimately driving better results, optimising productivity, and facilitating seamless collaboration.
Although discipline is not a one-size-fits-all approach, it can help create a work environment that encourages personal growth and accountability rather than solely relying on punitive measures.
In this deck, you will learn the significance of workplace discipline for organisational success. You’ll also learn
• Four (4) workplace discipline methods you should consider
• The best and most practical approach to implementing workplace discipline.
• Three (3) key tips to maintain a disciplined workplace.
Personal Brand Statement:
As an Army veteran dedicated to lifelong learning, I bring a disciplined, strategic mindset to my pursuits. I am constantly expanding my knowledge to innovate and lead effectively. My journey is driven by a commitment to excellence, and to make a meaningful impact in the world.
VAT Registration Outlined In UAE: Benefits and Requirementsuae taxgpt
Vat Registration is a legal obligation for businesses meeting the threshold requirement, helping companies avoid fines and ramifications. Contact now!
https://viralsocialtrends.com/vat-registration-outlined-in-uae/
Affordable Stationery Printing Services in Jaipur | Navpack n PrintNavpack & Print
Looking for professional printing services in Jaipur? Navpack n Print offers high-quality and affordable stationery printing for all your business needs. Stand out with custom stationery designs and fast turnaround times. Contact us today for a quote!
Affordable Stationery Printing Services in Jaipur | Navpack n Print
A New Model for Credit Approval Problems a Neuro Genetic System with Quantum Inspiration and Binary-Real Representation
1. Published in the World Congress on Nature and Biologically Inspired Computing (NaBIC'09), Coimbatore, India, Dec 09-11, 2009
A New Model for Credit Approval Problems: A Quantum-Inspired Neuro-
Evolutionary Algorithm with Binary-Real Representation
Anderson Guimarães de Pinho Marley Vellasco André Vargas Abs da Cruz
Department of Electrical Engineering, PUC-Rio
Rio de Janeiro, Brazil
agp@gmail.com marley@ele.puc-rio.br andrev@ele.puc-rio.br
Abstract—This paper presents a new model for neuro- topology and weights of a feed-forward neural network. This
evolutionary systems. It is a new quantum-inspired model is an extension of the QIEA- proposed by Cruz in
evolutionary algorithm with binary-real representation [6], which is a quantum-inspired evolutionary algorithm with
(QIEA-BR) for evolution of a neural network. The proposed real representation for numerical problems.
model is an extension of the QIEA- developed for numerical Cruz developed a quantum inspired evolutionary
optimization. The Quantum-Inspired Neuro-Evolutionary algorithm with a numerical representation to optimize the
Computation model (QINEA-BR) is able to completely weights of a neural network. On this paper, a binary
configure a feed-forward neural network in terms of selecting representation is added to the chromosome opening the
the relevant input variables, number of neurons in the hidden
possibility to optimize other types of variables that are
layer and all existent synaptic weights. QINEA-BR is evaluated
in a benchmark problem of financial credit evaluation. The
important for modeling a neural network, such as: which
results obtained demonstrate the effectiveness of this new attributes are relevant to be used on the input layer; how
model in comparison with other machine learning and many neurons to use on the hidden layer and which kind of
statistical models, providing good accuracy in separating good activation function to use on the hidden and output layers.
from bad customers. Such decisions are of categorical nature, and cannot be
efficiently represented as real numbers, thus leading to the
Keywords-quantum-inspired algorithms; genetic algorithms; use of a mix of representations into a single algorithm.
hybrid neuro-genetic systems;classification. This paper shows that the use of Quantum-Inspired
Evolutionary algorithms for training neural networks can be
I. INTRODUCTION used to successfully perform a binary classification. The
results presented here show a significant improvement when
A precise prediction of breach of contract has been the compared to other models.
objective of many companies in several segments. One of This paper is divided as follows: Section II presents
them, which has been extensively studied in the financial details of the new proposed quantum-inspired evolutionary
literature, is the credit default analysis. In this area, many algorithm, the QIEA-BR; Section III describes the
quantitative methods for creating models to separate bad and application of the proposed model to define the feed-forward
good customers have been explored [1], [2], [3], [14], [15], neural network configuration, the so-called QINEA-BR;
[16], [17], [18], [19], [20]. Section IV evaluates the proposed QINEA-BR model in a
Huang in [20] divides these methods in fields of the benchmark credit analysis application and compares its
science: such as discriminant analysis; logistic regression; results with other techniques; finally, Section V presents the
mathematical programming methods; recursive partitioning; conclusions and future work.
expert systems; neural networks; non-parametric methods of
smoothing; and models of time series.
Bose and Chen in [1] detailed some machine learning II. THE QIEA-BR ALGORITHM
techniques, such as: artificial neural networks; support vector The quantum-inspired genetic algorithm with binary-real
machines; genetic algorithms; genetic programming; representation (QIEA-BR) is a model where numerical and
evolutionary programming; and hybrid models of these binary parameters must be optimized.
techniques, with or without other statistical models. The binary representation part of the problem is based on
The use of hybrid models (neuro-genetic algorithms, for the concept of q-bits [7]. A q-bit can be in the "1" state, the
example), can help to overcome minor issues like neural "0" state, or in a superposition of both [4]. A state of a q-bit
networks overfitting problems and, thus, be more attractive can be represented as:
to solve complex problems with large volumes of data, such
as credit approval problems. 0 1 (1)
In this paper, a new model to evaluate credit approval in
financial problems is proposed. It is a Quantum-Inspired Where and
are complex numbers that determine
Neuro-Evolutionary Algorithm with binary-real the probability of observing the corresponding state, such
representation (QINEA-BR), which determines the final that:
2. Published in the World Congress on Nature and Biologically Inspired Computing (NaBIC'09), Coimbatore, India, Dec 09-11, 2009
2 2 ii) while t <= T
1 (2) t = t+1
iii) generate classic population P(t) with mix
Thus, a q-bit, the smallest unit of information, can be representation, observing Q(t)
defined as a pair of numbers ( , ) as:
iv) evaluate P(t)
v) if t=1 then
B(t) = P(t)
otherwise
(3)
vi) P(t) = Classic recombination of P(t) and B(t-1)
vii) evaluate P(t)
viii) B(t) = Best individuals of [P(t) U B(t-1)]
where (2) applies. ix) updates the binary part of Q(t) with the best
individuals of B(t), using q-gate
The representation of the numerical (continuous) part of x) updates the real part of Q(t) with the best
the chromosome is performed as in [6], using a probability individuals of B(t), using quantum-crossover
density functions (PDF). For a simple PDF, a uniform end
distribution inside a defined interval, this numerical gene can end
be represented as a pair of parameters called center and q0
width of the gene. Thus, the center () and width () are all i) Each j in Q(t) is initialized with equal probabilities
the parameters that are needed to represent this function, as
follows: for all states. In
q j b , all q-bits are equal to 1/ 2 .
0
In
q j r , considering that all the weights of a network
0
(4)
could be optimized assuming values in the range of (-
2,2), and would be equal to 0 and 2, respectively.
For a specific x ( , ) , the number to be q0
optimized, where: Note that initially, a j is the linear superposition of all
possible states, with equal probability of occurrence.
p( x) dx 1 , (5) ii) While t, the current generation, is less than the total
number of generations, QIEA-BR continues looping.
Thus, a quantum individual j, in an instant of time t, iii) The classical population P(t) is generated in compliance
representing mixed numeric and binary features, can be to the quantum states of individuals in Q(t). For each q-
defined as: bit, the algorithm generates a random number between 0
2
qtj qtj qt
b j r
and 1. If this number is between 0 and , then the
classic bit is generated with a value of 0; otherwise the
t t classical bit is 1. For the real classic gene, a number in
tjk tj1 tj 2 tjm
j1 j 2
... ... the interval ( ) to ( ) is randomly chosen.
t t t t t t
j1 j 2 jk b j1 j 2 jm r
(6) iv) Since the QIEA-BR model was developed for
classification problems, the evaluation of each
Thus, a population Q(t) in generation t, with n possible qt
individual j in P(t) considers the number of correctly
qt classified patterns, that is, correctly indicated states “1”
solutions j can be given as follows:
and “0”. Therefore, the evaluation function is calculated
t t t
by (8):
Q(t ) q1 , q2 ,..., qn (7)
( A j D j rc1/ c0 )
A detailed description of all steps of the QIEA-BR fj (8)
(C j rc1/ c0 B j )
algorithm is provided below:
QIEA-BR Algorithm Aj D
Where, is the number of true-positive samples, j
start
B C
t=0 true-negative, j false-negative and j false-positive.
i) initializes quantum population Q(t) with mixed rc1/ c0
representation is the ratio between total of “1” and “0”, in the
3. Published in the World Congress on Nature and Biologically Inspired Computing (NaBIC'09), Coimbatore, India, Dec 09-11, 2009
training sample. This ratio is used to avoid problems of
the quantum individual jn . Then, the update of jn
specialization in unbalanced databases. Finally, is a occurs as follows:
small parameter to avoid division by zero when the
problem is fully separable by the optimized model (i.e. jn' jn ( g jn jn )* random (10)
Cj B
and j are equal to zero). where random is a random number between 0 and 1,
generated for each quantum center, that determines the
v) In the first generation, the population of best individuals
founded in B(1) is the population observed P(1). speed of the update of jn in the direction of g jn . The
second parameter of the numerical gene, the pulse
vi) If it is not the first generation, the recombination occurs
- as in traditional genetic algorithms - between B(t-1), width jn , is updated in a similar way, by calculating
the population of the best individuals in the previous the total height of B(t) among all individuals, given
generation, and P(t). In all experiments carried out in by max( g jn ) min(g jn ) . So, jn is updated by:
this work the uniform crossover operator has been
applied, acting differently if gene is real or binary and jn' jn ((max( g jn ) min( g jn )) jn )* random (11)
applied in a pair of genes of an individual of B(t-1) and
another individual of P(t). Probability of crossover was This type of update for numerical representation on
specified by the user. No mutation operators have been quantum algorithms is inspired on the work of Cruz [6].
employed. Since quantum algorithms have good
potential of exploration and exploitation simultaneously,
classic mutation operation is no longer motivated on this III. QIEA-BR MODEL APPLIED TO NEURO-EVOLUTION
study, because we want to obtain potential results The main objective of the QIEA-BR is to apply it to
proved essentially by the quantum algorithm and neuro-evolution, that is, to completely configure a feed-
operators. forward neural network, with one hidden layer, to binary
classification models. With this objective in mind, the
vii) Evaluate the new population P(t)', resultant of the following parameters must be defined by the QIEA-BR
recombination of P(t) and B(t-1), applying (8). model:
a. Which variables, among the available ones, are
viii) The new population in B(t) is given by the best relevant to be used as inputs for the neural network?
individuals from the union of P(t) with B(t-1), b. How many neurons must be used in the single
respecting the size of B(t) population defined by the hidden layer?
user. c. What kind of activation function must be applied
qt throughout the network? Sigmoid logistic or
ix) The q-bit of a quantum individual j is updated by the hyperbolic?
q-gate rotation operators [7], [11]. Initially, one classic The above parameters will be represented as binary genes
individual from B(t) and another from Q(t) are selected in the hybrid chromosome representation of the QIEA-BR
randomly. Each q-bit is updated in the direction of the model. The main reason that we choose between logistic and
individual from B(t), increasing or decreasing the hyperbolic activation function is that according to Haykin
probability of a state "0" or "1". A new q-bit can be [23], page 40, these function are the most common used
obtained as follows: when constructing artificial neural networks.
Other important neural network configuration
cos( ) sin( ) j1
t parameters, which will be represented as numerical genes in
(9) the QIEA-BR chromosome, are the following:
sin( ) cos( ) tj1
d. Synaptic weights of the single hidden layer;
e. Synaptic weights of the output layer;
where is the angle of rotation and should be f. Threshold value, at the output neuron, that separates
assigned considering the type of problem by the user. the two output classes.
Depending on the intention of increase or decrease the Therefore, to optimize all these parameters, the proposed
probability, the sine terms (positive and negative) are QIEA-BR model, described in the previous section, is
exchanged. applied, with the following parameters:
nh: Maximum number of neurons in the hidden
x) To update the numeric genes of the quantum individual layer;
Q(t), the same classic individual B(t), used in the numQuantum: number of quantum individuals;
previous step, is employed. Consider the n-th numeric numClassic: number of classic individuals;
g numGeneration: Number of generations;
gene of a classic individual as jn , and the n-th gene of C- Crossover: classical crossover rate;
4. Published in the World Congress on Nature and Biologically Inspired Computing (NaBIC'09), Coimbatore, India, Dec 09-11, 2009
: parameter to update the binary part of the chromosome, only if its weights are active by the binary
quantum individuals; genes. If the binary genes are inactive, real genes are kept
Q- Crossover: parameter to update the real part of unchangeable when applying q-crossover. On the other side,
the quantum individual. if binary genes are active, real genes could suffer updates by
Therefore, the final chromosome representation depends the crossover at least a number of “minGeneration”
on the maximum number of neurons defined by the user and controlled by the user.
the maximum number of possible variables in the input It is important to stress that all parameters presented
layer, so the length of the chromosome is fixed during the above should be adjusted according to the specific
evolutionary process. For example, in the case of a application problem.
maximum of 20 neurons in hidden layer and 30 available
input variables, the representation of an individual j, in an IV. EXPERIMENTAL RESULTS AND DISCUSSION
instant of time t, will contain 51 pairs of genes in the
quantum binary part: 30 to determine whether a variable in The proposed QIEA-BR model was evaluated in a
the input layer is active or not; 20 for activating the neurons benchmark application related to credit approval: the
in the hidden layer; and 1 gene to define the activation "Australian credit approval problem", which is available in
function that should be used in all layers (sigmoid or the UCI Machine Learning Repository [21]. As in any credit
hyperbolic). The numeric part of the chromosome analysis problem, this database provides a set of customers
representation contains, in this case, 621 pairs: 600 to define that are divided in good and bad payers. For confidential
the weight values for the synapses between input and hidden reasons, the meaning of the attributes is not provided by their
layers; 20 for the weights between hidden and output layers; administrators.
and 1 for the threshold value. This chromosome The database consists of 690 samples, with 307 (44.5%)
representation is provided below: composed of bad payers and 383 (55.5%) of good payers.
There are a total of 15 continuous and categorical attributes
qtj qtj qt
b j r (14 explanatory variables, and 1 that informs the class of the
costumer: good or bad). The database contained about 5% of
t t tj 51 tj 52 tj 53 tj 672 (12) customers with missing values in at least one attribute, which
j1 j 2 were treated by including the average and median (only one
t t ... ...
t t t t
j1 j 2 j 51 b j 52 j 53 j 672 r
for each attribute treated). Both categorical and numerical
attributes were pre-processed. Categorical attributes
(variable 5, 6 and 12) were transformed into the 1-of-N
It must be pointed out that the binary and numeric genes encoding. Variables 2, 3, 10, 13, and 14 were normalized by
are dependents. When a neuron is inactive in the classic mean and standard deviation.
individual, the crossover operator must no be applied to the The database was divided into 70% for training and 30%
weights associated to this neuron in the numeric part. for testing, in a 3-fold cross-validation process, thereby
Similarly, these weights can not be used to update the obtaining three sets of data for training and testing. We
parameters of the quantum individual. choose 3-fold for the cross-validation process, for
There are two additional parameters included in the comparisons with authors (see Lacerda page 178).
algorithm:
minGeneration: minimum number of generations
that a neuron remains active before it can be A. Results
disabled by a crossover. This parameter is important
to avoid that a neuron is turned off before the The QINEA-BR model was implemented and tested in
evolutionary process had enough time to optimize Matlab. Varying the parameters presented in section III, we
its weights. could observe the results given by the evaluation function in
updateGeneration: number of rounds that must be (8) from the best individual. After many tests, the parameters
executed before thee q-gate and q-crossover are presented in Table I were defined. Here we are going to
applied on the binary and numeric genes of the present just the final parameters adjusted. Sensitive analysis
quantum individual. This parameter controls the of how to control these parameters will be considered in
exploitation and exploration aspects of the futures works.
evolutionary process. That is, if a slower and After setting the parameters, each of the three training
gradual optimization process is desired, for a and test samples was subjected to evaluation by 3 neural
greater exploration of the search space, the higher networks, each developed independently. At the end, 9
his parameter should be. neural networks were obtained by the neuron-quantum
evolution. The results were evaluated by the percentage of
See that the binary part of a classic chromosome which wrong classified patterns (PWCP) and can be seen in Table
is responsible for enable or disable a neuron on the input and II.
hidden layers, is conditioned to the real part which As can be observed from Table II, among all the
determines the weights between neurons. And so, the experiments and samples, the model QINEA-BR showed an
algorithm must consider updating the real genes of a average PWCP of 15.0%, with a standard deviation of 2.9%.
5. Published in the World Congress on Nature and Biologically Inspired Computing (NaBIC'09), Coimbatore, India, Dec 09-11, 2009
The results obtained by the QINEA-BR were also M odel PWCP Average Standart D.
compared with other models provided in Lacerda & NEIQ-BR 15,0% 2,9%
Carvalho in [14], for the same issue of Australian credit Average Other M odels 16,5% 3,4%
approval. This comparison is provided in Table III below.
M LP-Backprop 17,1% 1,8%
It can be observed that, on average, the QINEA-BR
model presents a lower PWCP average than the models Cascade correlation 18,0% 3,0%
estimated by Lacerda and Carvalho. However, the statistical Tower 14,7% 3,2%
t-test [22] for the difference between means leads to the Pyramid 16,9% 2,1%
acceptance of the equal means hypothesis by the level of 5% SVM 16,7% 2,6%
of confidence, i.e., the difference is not significant. RBF - Batch 16,7% 3,9%
Considering the confidence of 10%, the t-test is significant
RBF - DF 16,3% 2,5%
with p-value of 0,096. We should say that before we applied
the t-test to compare means, each sample was proved to be RBF - IO 17,8% 4,0%
normally distributed by the Kolmogorov-Smirnov test, where RBF - DFIO 16,7% 4,3%
null hypothesis was accepted [22]. RBF - IODF 17,3% 4,4%
Other algorithms from Carvalho and Lacerda apud Jones RBF - On line 16,9% 4,4%
and Quinlan, and their comparison with the QINEA-BR RBF - Optimal 15,9% 4,7%
model can be found in Table IV.
RBF - GA 14,0% 3,5%
TABLE I. PARAMETERS SET
nh 20 TABLE IV. COMPARISON WITH OTHER MODELS IN CARVALHO AND
numQuantum 2 LACERDA APUD JONES AND QUINLAN
numClassic 400 M odel PWCP Average
numGeneration 200 NEIQ-BR 15,0%
C-Crossover 0,95
Average of Other M odels 16,5%
0,020*pi
Q-Crossover 0,95 C4.5 Rules 15,5%
minGeneration 3 C4.5 Trees 15,1%
updatesGeneration 10 Foil trad.1 17,8%
Foil trad.2 17,4%
Foil trad.3 17,0%
TABLE II. RESULTS OBTAINED FOR THE "AUSTRALIAN CREDIT
APPROVAL PROBLEM” Foil exd.1 18.,0%
Foil exd.2 16,4%
Samples PWCP
Experiment Foil exd.3 16,4%
1 2 3 Aver. Stand. D.
1 16,9% 13,5% 13,0% 14,5% 2,1% Again, the QINEA-BR model presented the lowest
2 18,4% 18,8% 11,1% 16,1% 4,3% PWCP average, but this difference is not proved by statistical
3 17,4% 13,5% 12,1% 14,3% 2,7% evidences.
Average 17,6% 15,3% 12,1% 15,0% -
Standart D. 0,7% 3,1% 1,0% - 2,9% V. CONCLUSIONS
This paper presented a new quantum-inspired
evolutionary computation model based on a hybrid binary
TABLE III. COMPARISON WITH OTHER MODELS PROVIDED IN and numeric representation, named QIEA-BR. The proposed
CARVALHO AND LACERDA [14] model was developed for a neuro-evolution application, and
tested in a benchmark application of credit approval.
The resultant neuro-evolution model provides the user
with a high degree of flexibility, avoiding the necessity to
perform variable selection and the specification of all neural
networks parameters, such as number of neurons in the
hidden layer and the threshold value used in the output layer
to define the final classification of the input pattern.
Although the difference in the percentage of wrong
classified patterns (PWCP) obtained by the QINEA-BR and
others models used in the literature has not been significant,
the results were quite surprising. It was shown that, on
average, the QINEA-BR model could replace but it is not
significant better than the others.
6. Published in the World Congress on Nature and Biologically Inspired Computing (NaBIC'09), Coimbatore, India, Dec 09-11, 2009
Many parameters, however, must be defined in the 2005, Springer Science and Business Media, Inc. Manufactured in
QINEA-BR model, which is directly related to the potential The Netherlands.
for classification and generalization of the neural network [15] YU, Lean; WANG, Shouyang; LAI, Kin Keung. Credit risk
assessment with a multistage neural network ensemble learning
modeling. A more detailed analysis of the dependency and approach. Expert Systems with Applications 34 (2008) 1434-1444,
impact on the results of these parameters must be carried out Elsevier.
in the future. [16] CHEN, Mu-Chen; HUANG, Shih-Hsien. Credit scoring and rejected
The model has been tested for a binary classification instances reassigning through evolutionary computation techniques.
model. However, the model can be easily extended to Expert Systems with Applications 24 (2003) 433-441, Elsevier.
multiple class problems, with a more general multi-layer [17] HUANG, Jih-Jeng; TZENG, Gwo-Hshiung; ONG, Chorng-Shyong.
perceptron topology. Two-stage genetic programming (2SGP) for credit scoring model.
Applied Mathematics and Computation (2005), Elsevier.
REFERENCES [18] PIRAMUTHU, Selwyn. Financial credit-risk evaluation with neural
and neurofuzzy systems. European Journal of Operational Research
112 (1999) 310-321.
[1] BOSE, Indranil; CHEN, Xi. Quantitative models for direct marketing. [19] HUANG, Zan et al. Credit rating analysis with support vector
European Journal of Operational Research 195 (2009) 1-16. machines and neural networks: a market comparative study. Decision
[2] YOBAS, M. B.; CROOK, J. N.; ROSS; P.. Credit scoring using Support Systems 37 (2004) 543-558, Elsevier.
neural and evolutionary techniques. IMA Journal of Mathematics [20] HUANG, Zan et al. Statistical Classification Methods in Consumer
Applied in Business and Industry 11 (2000) 111-125. Credit Scoring: a Review. Journal of Royal Statistical Society 160
[3] SAKPRASAT, Sum; SINCLAIR, Mark C.. Classification Rule (1997) 523-541.
Mining for Automatic Credit Approval using Genetic Programming. [21] MURPHY, C. A.; AHA, D. W. “UCI repository of machine learning
2007 IEEE Congress on Evolutionary Computation. databases”. Irvine, CA, University of California, 1994.
[4] KIM, Yehoon; KIM, Jong-Hwan; HAN, Kuk-Hyun. Quantum- [22] JOHNSON, Richard Arnold.; WICHERN, Dean W. “Applied
inspired Multiobjective Evolutionary Algorithm for Multiobjective Multivariate Statistical Analysis”. 6th ed., Prentice Hall, 2007.
0/1 Knapsack Problems. 2006 IEEE Congress on Evolutionary
Computation. Vancouver, BC, Canada. [23] HAYKIN, Simon. Neural Networks: A Comprehensive Foundation.
2nd edition, Prentice Hall, 1999.
[5] JANG, Jun-Su; HAN, Kuk-Hyun; KIN, Jong-Hwan. Face Detection
using Quantum-inspired Evolutionary Algorithm. Congress on
Evolutionary Computation, 2004. Pag. 2100-2106, Vol.2.
[6] CRUZ, André Vargas Abs. VELLASCO, Marley Maria Bernardes
Rebuzzi, PACHECO, Marco Aurélio Cavalcanti, Quantum-Inspired
Evolutionary Algorithm for Numerical Optimization, Book Series
Studies in Computational Intelligence, Book Quantum Inspired
Intelligent Systems Vol. 121, Nedjah, Nadia; Coelho, Leandro dos
Santos; Mourelle, Luiza de Macedo (Eds.), Publisher Springer Berlin
/ Heidelberg, pp.115-132, 2008.
[7] HAN, Kuk-Hyun; KIN, Jong-Hwan. Quantum-Inspired Evolutionary
Algorithm for a Class of Combinatorial Optimization. IEE
Transactions on Evolutionary Computation, Vol. 6, No. 6, December
2002.
[8] LI, Zhiyong; RUDOLPH, Günter; LI, Kenli. Convergence
performance comparison of quantum-inspired multi-objective
evolutionary algorithms. Journal of Computers and Mathematics with
Applications, Elsevier, 2008.
[9] TALBI, Hichem; MOHAMED, Batouche; DRAA, Amer. A
Quantum-Inspired Evolutionary Algorithm for Multiobjective Image
Segmentation. International Journal of Mathematical, Physical and
Engineering Sciences, Volume 1, Number2.
[10] MAHDABI, Parvaz; JALILI, Saeed; ABADI, Mahdi. A Multi-Start
Quantum-Inspired Evolutionary Algorithm for Solving Combinatorial
Optimization Problems. GECCO, 2008. Atlanta, Georgia, USA.
[11] HAN, Kuk-Hyun; KIN, Jong-Hwan. Genetic Quantum Algorithm and
its Application to Combinatorial Optimization Problem. Congress of
Evolutionary Computation, Piscataway, NJ, 2000.
[12] ARPAIA, Pasquale; MECCARIELLO, Giovanni; RAPONE, Mario;
ZANESCO, Antonio. Quantum-Inspired Evolutionary Classification
of Driving Sequences in Vehicle Emission Factor Measurement. 15th
IMEKO TC4, Symposium on Novelties in Electrical Measurements
and Instrumentation, Iasi, Romania.
[13] MOORE, Mark; NARAYANAN, Ajit. Quantum-inspired computing.
Department of Computer Science Old Library, University of Exeter,
UK, 1995.
[14] LACERDA, Estefane; CARVALHO, André C. P. L. F.; BRAGA,
Antônio P.; LUDERMIR, Teresa B.. Evolutionary Radial Basis
Functions for Credit Assessment. Applied Intelligence 22, 167-181,