Application of Data Mining Techniques to Predict Children Datasets: The Case of Love for Children Organization

Gebreyohans, Gebremedhin (2012) Application of Data Mining Techniques to Predict Children Datasets: The Case of Love for Children Organization. Masters thesis, Addis Ababa University.

PDF (Application of Data Mining Techniques to Predict Children Datasets: The Case of Love for Children Organization) Gebremedhin, Gebreyohans final.pdf - Accepted Version Restricted to Repository staff only Download (808kB) | Request a copy

Abstract

This research describes and evaluates classification of children to four classes as orphan, single orphan, vulnerable and safe that will help for the organizations to donors from outside of the organization and to get full information about each child for internal purpose of the organization. To classify this three algorithms of classification techniques are used which are Decision tree, Bayesian learning and Neuron Network algorithm have been explored within the framework of KDD data mining model has been used and The data have been analyzed and interpreted using the WEKA 3.7.4 version software. Data are collected, cleaned, transformed and integrated for experimenting with the classification model. The final dataset consists of 17044 records have been experimented and evaluated against their performances. The collection of data set are experimented with the 10-fold cross-validation and splitting the datasets in to 70% for training and 30% for testing, 66% for training and 44% for testing as well as 50% for training and 50% for testing is used. Finally a comparison of decision tree, Bayesian network and neural network model in terms of the overall classification accuracy and their advantage is made. As a result, Decision Tree is selected because it gives better results than Bayesian learning and better advantage over Neuron Network so due to more advantages of decision tree over the others: The accuracy of these algorithms are by Decision Tree C4.5, and Naïve Bayes, Multilayer Perception, 98.83%, 98.32%, 98.86% respectively. This shows that the Decision Tree classifier performed better than the other on a specific children’s dataset. Finally, in overall decision tree is selected as a model for the organizations children classification.

Item Type:	Thesis (Masters)
Subjects:	H Social Sciences > H Social Sciences (General) Z Bibliography. Library Science. Information Resources > Z665 Library Science. Information Science
Divisions:	Africana
Depositing User:	Selom Ghislain
Date Deposited:	16 Jul 2018 08:42
Last Modified:	16 Jul 2018 08:42
URI:	http://thesisbank.jhia.ac.ke/id/eprint/7359

Actions (login required)

View Item