Supervised Machine learning is used to solve the binary classification problem on four datasets of European Survey of Schools: Information and Communication Technology (ICT) in Education (known as ESSIE) which is supported by Euro-pean Union (EU). To predict the gender of the principal based on their response for the ICT questionnaire, the authors applied four supervised machine learning algorithms (Sequential minimal optimization (SMO), Multilayer perception (ANN), Random Forest (RF) and Logistic Regression (LR) on ISCED-1, ISCED-2, ISCED-3A and ISCED-3B level of schools. The survey was conducted by the European Union in the academic year 2011-2012. The datasets have total 2933 instances & 164 attributes considered for the ISCED-1 level, 2914 in-stances & 164 attributes for the ISCED-2 level, 2203 instances & 164 attributes for the ISCED-3A level and 1820 instances & 164 attributes for the ISCED-3B level. One the one hand, SMO classifier outperformed others at ISCED-3A level and on the other hand, LR outperformed others at ISCED-1, ISCED-2 and ISCED-3B. Further, real time prediction and automatic process of the data sets are done by introducing the concepts of the web server. The server communicates with the European Union web server and displays the results in the form of web application. This smart approach saves the data process and interaction time of humans as well as represent the processed data of the Weka efficiently.
Basic Civil Engineering first year Notes- Chapter 4 Building.pptx
Smart Approach for Real Time Gender Prediction ofEuropean School's Principal Using Machine Learning
1. Smart Approach for Real Time Gender
Prediction of
European School's Principal Using
Machine Learning
Yatish Bathla, Chaman Verma, Neerendra Verma
Obuda University, ELTE University
Budapest, Hungary
Central University of Jammu,
Jammu, India
2. OUTLINE
• Introduction
• Methods and Techniques
• Knowledge Flow Environment
• Experiments Results, Analysis, and Evaluation
• Web Server for Real Time Prediction
3. Introduction
• Machine learning (ML) has been using in the various sectors
like Computer vision, text and speech recognition, spam
filter on the email
• European Commission has been conducted a survey over
190,000 filled questionnaires from students, teachers and
principals in 27 European Union (EU) countries to analysis
the Information and Communication Technology (ICT) in
ISCED level
• Four supervised machine learning algorithms i.e. sequential
minimal optimization (SMO), Multilayer perception,
Random Forest (RF) and binary Logistic Regression (LR)
are applied by using the Weka to predict the principals'
gender of based on the ICT questionnaire.
4. Methods and Techniques
• Dataset: Four secondary datasets of European School's
principal has been downloaded from the European Union
(EU) website. There is a various level of school's division
ISCED level-1, ISCED level-2 and ISCED level-3A and
level-3B
5. Methods and Techniques
• Preprocessing: Before use dataset, it is essential to improve
data quality. There are a few numbers of techniques used for
data pre-processing as aggregation, sampling, dimension
reduction, variable transformation, and dealing with missing
values.
6. Knowledge Flow Environment
• To predict the gender, we used Knowledge Flow
Environment (KFE) which is a substitute for the Weka
Explorer. The experimental layout of supervised machine
learning with filters, classifiers, evaluators, and visualizers
12. Conclusion
•During the experimental study, LR has been proven as best approach
that trained ISCED level-1, ISCED level-2 dataset and ISCED level-3B
dataset with k-fold cross-validation to predict the gender of the
European principals.
•The maximum accuracy is achieved with 164 attributes by LR
(61.6%) as compare to RF (61%) to predict principal gender at ISCED
level-1.
•Again, LR classifier obtained the highest accuracy (59.3%) as
compare to RF (59.1%) to predict principal gender at ISCED level-2.
•LR classifier also obtained the highest accuracy (58.4%) as compare
to RF (57.7%) to predict principal gender at ISCED level-3B.
•But, SMO classifier obtained the highest accuracy (63.5%) as
compare to RF (62.2%) to predict principal gender at ISCED level-3A.
•Finally, Evaluation web server saves the time and represents the data
smartly.