OPTIMIZATION OF TIME DEPOSIT CLASSIFICATION USING KNN ALGORITHM WITH INFORMATION GAIN FEATURE SELECTION
Abstract
A broad marketing strategy in the banking and financial services industry to serve business, investment, and banking companies has the opportunity to carry out promotions and marketing strategies efficiently with the help of technology. Customer identity that provides a promotional response to a product is an important part of marketing. Data mining can provide solutions to these problems. With the method used in completing direct marketing, there are still data features that have little effect on performance in direct marketing so it provides a low level of accuracy to increase the level of performance is done by reducing several features that do not affect direct marketing by using add information gain on the KNN algorithm to get good performance. The first step is to model with the KNN algorithm in a limited way using 5-fold cross-validation to produce the best K value, namely K = 9 with an accuracy of 89.72%, 98% recall, and 91% precision. The addition of information gain to the KNN algorithm at K=9 using 5-fold-cross-validation produces an accuracy value of 90.49%, a recall of 97 %, and 92% precision so that the addition of information gain to the KNN algorithm can increase the accuracy value and provide an increase in the proportion of precision.