Introduction PROS Maria FVR, Brodjol SS U Classifying the poor fulltext

Proceedings of the IConSSE FSM SWCU 2015, pp. MA.66–70 ISBN: 978-602-1047-21-7 SWUP MA.66 Classifying the poor household using neural network Maria F.V. Ruslau and Brodjol Sutijo S. Ulama Department of Statistics, Sepuluh Nopember Institute of Technology, Surabaya, Indonesia Abstract The issue of poverty has recently been brought to the public’s attention. The picture of Indonesian development reveals many families are not benefiting from national economic growth. Many families were still poor and hovering below the poverty line. The classification of the individual or of poor households in a class or poverty status can be a good instrument to focus on the living conditions of the poor. In this study, back propagation algorithm was used to build models of neural networks that can classify each poor household appropriate their poverty status. Network is built using the weights of the selection of the best network. The best networks have been training on the sub-sub smaller dataset. Classification is done by replication 10-fold cross validation. Average accuracy of classification in the training data is 58.89 percent while the testing data of 56.42 percent. Keywords back propagation, classification, neural network, poverty

1. Introduction

Classification is one of the most frequently encountered decision making tasks of human activity and application areas of neural networks. A classification problem occurs when an object needs to be assigned into a predefined group or class based on a number of observed attributes related to that object. Many problems in business, science, industry, and medicine can be treated as classification problems. Examples include bankruptcy prediction, credit scoring, medical diagnosis, quality control, handwritten character recognition, and speech recognition TNP2K, 2013a. Many social studies analyze attitude responses using regression models. The classification and recognition of individual characteristics and behaviours constitute a preliminary step and is an important objective in the behavioural sciences TNP2K, 2013b. Current statistical methods do not always give satisfactory results. Neural network are an alternative method for classification can work with large numbers of qualitative variables such as behaviours, provided that they can be coded, and they are able to use non-linear linked variables. A methodology based on one of the principles of artificial neural networks, the backpropagation, can improve performance in this area Boonkiatpong Sinthupinyo, 2011. This research focus on the analysis with large data of poor household by applying backpropagation neural networks. Measurement and analysis of poverty is necessary to identify individuals and households in need of government assistance and aid. Cases in this study was a multi-class classification for ordinal response or target. To analyze these data we Corresponding author. Tel.: +682335111988; E-mail address: mariafvruslauyahoo.com M.F.V. Ruslau, B.S.S. Ulama SWUP MA.67 adopt the techniques introduced by Boonkiatpong Sinthupinyo 2011. Classification accuracy is not increased significantly. However, the network integration method, simply helping to reduce errors and improve accuracy, especially for large and heterogeneous data sets.

2. Materials and methods