IMPLEMENTATITON OF RANDOM FOREST ALGORTIHM ON SALES DATA TO PREDICT CHURN POTENTIAL IN SUZUYA SUPERMARKET PRODUCTS

Abstract

Concentration of sales that are focused on products that are in great demand and are popular is one of the supermarket sales techniques. Seasonal sales techniques like this sometimes have an impact that can be seen obviously by the imbalance in sales of existing products in supermarkets. Sales imbalance can be the initial cause for a product to lose interest and become a product that is eventually removed from store. With a classification model made to predict which products will be eliminated or churn, it can assist staff in distributing the sales of each product. The more products are churn due to lack of enthusiasts which can affect the overall sales of the supermarket. The purpose of this study is to assist staff in classifying potentially churn products. The classification model consists of 3 models with different algorithms and the results show that the application of the Random Forest algorithm is more effective for predicting data with 96% accuracy compared to 81% for the Logistic Regression algorithm and 46% for the Support Vector Machine algorithm.