The growth of Fast Moving Consumer Goods (FMCG) industry is still showing double-digit and Indonesia becomes a potential market for the products FMCG, so that the competition between companies will be intense. The company have to attempted to survive, one of the way is to maintain customer loyalty. Data mining techniques can be used to predict customer loyalty. In data mining pra-processing, feature selection is one of the important thing to reduces the number of features, removes irrelevant, redundant, or data noise, and brings the immediate effects for applications: speeding up a data mining algorithm, improving mining performance such as the accuracy of the prediction and the comprehensive result. This paper aims to identify the relevant factors that affect the performance of the classification of customer loyalty with several feature selection method and to compare the classification performance in customers loyalty prediction of FMCG products. Data was obtained from the results of fast moving consumer goods customers questionnaires towards several brands of instant noodles in Lampung that was ranked TOP Brand Award Phase 1 2016, using nonprobability sampling method and convenience sampling technique. The result in this paper, chi square feature selection methods with threshold > 0.01 showed the best results, it is indicated by the highest accuracy of random forest classification algorithm, that is 83.2% for thirteenth features
Copyrights © 2017