Comparing Machine Learning Techniques and Classical Approach for child’s Education and Alternative Activities in case of Pakistan
Author: Namal Kinza

This study investigates the factors which are more affective in the decision of child activities through classification techniques. Further, it compares the classical approach and machine learning techniques of classification on the basis of overall accuracy of confusion matrix and area under the curve (AUC). The data is taken from Pakistan social and living standards measurement (PSLM) survey for year 2014-2015 and is based on urban and rural areas of four provinces of Pakistan. Two separate models are made based on age groups 4-9 and 10-14. Results showed that accuracy from confusion matrix and area under the curve ROC analysis of classification tree model is greater than the MLR and LDA for the age group 4-9. While accuracy from confusion matrix of classification tree is greater than MLR and LDA for age group 10-14. However, accuracy checked in the context of area under the ROC analysis showed no significant difference between the accuracy of three model. Our finding show that classification tree is best technique among others as it also identifies the most significant variables. Such as ,child gender, kaccha house, fuel for cooking mother education ,mother employment,region ,child’sage, infants, toilet facility, aggland, cattle, 16-64 female, source of drinking water and father employment. Therefore, it is recommended to reducing the gender discrimination towards child activities. Gender disparity should be minimized through public awareness about girls’ education. Woman education in both model have significant effect on the decision of child activities. We have to focus on girl’s education because in future girls can play important role as woman. It has an increasing effect on human capital through the education. Supervisor:- Dr. Hafsa Hina

Meta Data

Keywords : Alternative Activities in Pakistan, Child’s Education, Classical Approach, Comparing Machine Learning Techniques
Supervisor: Hafsa Hina

Related Thesis​