TataKelola dan KamSiber Kecerdasan Buatan v022.pdf
A Fuzzy Multi-Stage ML Method for Cyber-Hate Detection
1. Base paper Title: A Multi-Stage Machine Learning and Fuzzy Approach to Cyber-Hate
Detection
Modified Title: A Fuzzy and Multi-Stage Machine Learning Method for Cyber-Hate Detection
Abstract
Social media has revolutionized the way individuals connect and share information
globally. However, the rise of these platforms has led to the proliferation of cyber-hate, which
is a significant concern that has garnered attention from researchers. To combat this issue,
various solutions have been proposed, utilizing Machine learning and Deep learning techniques
such as Naive Bayes, Logistic Regression, Convolutional Neural Networks, and Recurrent
Neural Networks. These methods rely on a mathematical approach to distinguish one class
from another. However, when dealing with sentiment-oriented data, a more ‘‘critical thinking’’
perspective is needed for accurate classification, as it provides a more realistic representation
of how people interpret online messages. Based on a literature review conducted to explore
efficient classification techniques, this study applied two machine learning classifiers,
Multinomial Naive Bayes and Logistic Regression, to four online hate datasets. The results of
the classifiers were optimized using bio-inspired optimization techniques such as Particle
Swarm Optimization and Genetic Algorithms, in conjunction with Fuzzy Logic, to gain a
deeper understanding of the text in the datasets.
Existing System
It was the advancement of technology and the impulse of human communication that
led to the evolution of social media, which altered how individuals interact online. Prior to the
introduction of Information Communication Technology (ICT), human interactions were
largely confined to geographical locations; however, Online Social Networks (OSNs) have
eliminated geographical barriers [1]. It has become increasingly apparent that cyber-hate is a
widespread issue due to the pervasiveness of easy-to-use technologies. Social media platforms
have emerged as a medium for the perpetration of aggressiveness and bullying, making it a
dangerous and elusive phenomenon. The ease with which perpetrators can commit harmful acts
through the utilization of a laptop or mobile device connected to the internet renders young
individuals highly vulnerable to online harassment. A conventional approach to detecting
cybercrime involvesmanual flagging of data [2]; however, this method has been demonstrated
2. to be neither ‘‘effective nor scalable’’ [2]. This has prompted researchers to investigate the
potential of utilizing Machine Learning and Deep Learning techniques to design automated
systems capable of detecting and preventing cyber-hate.
Drawback in Existing System
Complexity and Computational Overhead: The multi-stage approach involves
multiple models, leading to increased computational complexity, longer processing
times, and higher resource requirements.
Model Interpretability: Combining different models and fuzzy logic systems might
reduce the interpretability of the overall system, making it challenging to understand
the decision-making process and trace the source of errors or biases.
Optimization Challenges: Tuning parameters and optimizing multiple stages in the
pipeline, each with its own set of hyperparameters, might be complex and require
meticulous fine-tuning for optimal performance.
Integration Complexity: Integrating multiple models and fuzzy logic systems into a
coherent and scalable hate speech detection pipeline can be challenging, especially in
real-world deployment scenarios.
Proposed System
Furthermore, [34] proposed a modified fuzzy approach with two-stage training for the
classification of four types of hate speech, specifically those based on religion, race,
disability, and sexual orientation.
The proposed approach is designed to address the challenge of text ambiguity. The
features are extracted using a combination of bag-of-words and word embedding
methods, and the correlation-based feature subset selection method is used to select the
relevant features.
The performance of the proposed fuzzy approach is compared to popular methods and
existing fuzzy approaches.
The results of the experiments indicate that the proposed fuzzy approach outperforms
the other methods (DT, NB, SVM, GBT, and DNN) in most cases
Algorithm
3. Genetic Algorithm is described as an Evolutionary Algorithm which finds the optimal
solution in the process of natural selection and crossover.
Algorithm process is terminated due to the given constraints, the optimum of the
solutions is obtained to solve the problem.
To optimize the accuracy of machine learning algorithms, libraries such as PySwarms
for Particle Swarm Optimization and TPOT for Genetic Algorithm were utilized. The
entire program required approximately a day to run all models across various platforms.
Advantages
Improved Detection Accuracy: Leveraging multiple stages of machine learning
models and fuzzy logic systems can enhance the precision and recall rates in identifying
cyber-hate speech, covering a wider range of patterns and contexts.
Robustness to Noise: A multi-stage approach can potentially mitigate the impact of
noise or ambiguity in textual data by refining the detection process through successive
stages, enhancing the model's robustness.
Customization and Adaptability: Fuzzy logic systems allow for the creation of
flexible rule-based models that can adapt to diverse linguistic patterns and nuances,
providing a degree of customization in hate speech identification.
Reduced False Positives: By employing multiple stages for classification and
validation, the approach may reduce false positive rates, enhancing the accuracy of hate
speech detection.
Software Specification
Processor : I3 core processor
Ram : 4 GB
Hard disk : 500 GB
Software Specification
Operating System : Windows 10 /11
Frond End : Python
Back End : Mysql Server
IDE Tools : Pycharm