Sales Profit Forecasting In Indonesian State Owned Enterprise: A Comparative Study Of Machine Learning Algorithms

Abstract

Forecasting is a crucial step in management planning to predict future condition of the business. Company needs to find forecasting method with best performance to create company’s strategy and avoid inaccuracy. Recent studies have attempted to find the best prediction method using machine learning to predict sales demand and sales forecast in various industries. By using machine learning algorithm, accuracy rate can be measured to evaluate prediction method. This study aims to find best sales profit forecast method by comparing three machine learning model: Linear Regression, Neural Network, and Gradient Boost Regression. Data source for this study was from 32 branch offices of Indonesian State Owned Enterprise PPI (Perusahaan Perdagangan Indonesia / Indonesian Trade Center), with data range of year 2017 to 2022. Study finds that Neural Network has the highest performance with the smallest error rate, compared to Linear Regression and Gradient Boost Regression, with 96.97% accuracy rate.