Customer Segmentation Analysis Using Random Forest & Naïve Bayes Method In The Case of Multi-Class Classification at PT. XYZ

Abstract

Cases of the COVID-19 pandemic are gradually decreasing every day in Indonesia, but the impact of the COVID-19 pandemic has greatly affected various sectors, especially the economy and business. Sales transactions have not yet reached the company's target due to weak public purchasing power. The accuracy of customer segmentation analysis and attractive promo voucher offers are needed to increase the opportunity for people's purchasing power for a product. This study aimed to predict the level of customer purchasing power using the random forest and naïve Bayes methods in the case of multi-class data classification at PT. XYZ. The classification is carried out to determine the type of promo voucher suitable to be offered to customers according to the level of customer purchasing power. The data used is historical daily transaction data from January 1, 2022, to December 31, 2022, which is the transition period for the COVID-19 pandemic. Evaluation using the random forest method produces an accuracy of 99.99%, while the naïve Bayes method produces an accuracy of 92.99%. The random forest and naïve Bayes methods can work very well on large data volumes. However, from the comparison results, it can be concluded that the performance of the random forest method is better than the naïve Bayes method in the multi-class classification case in predicting the level of customer purchasing power at PT. XYZ.