IMPLEMENTASI METODE RANDOM FOREST UNTUK MEMPREDIKSI PENJUALAN PRODUK

Abstract

The fast and rapid development of the world of information technology has also resulted in competition in the business world becoming complicated and complex. One important role for companies in determining sales strategies is sales prediction. Forecast is forecasting what will happen, for example, predicting the number of products that will be sold in the next period. Various time series forecasting models have been used by researchers to predict future events. The time series modeling used in forecasting includes the Random Forest method, Single Exponential Smoothing and Double Exponential Smoothing. The random forest method is a method that is able to work more efficiently and effectively for discrete and continuous data. This method is able to produce smaller errors and has a high level of accuracy compared to other methods. The aim of this research is to implement the Random Forest method in carrying out the product sales prediction process. The variable used in the prediction process is the total nominal sales of each product. From the results of the tests carried out, information was obtained that the greater the number of decision trees, the lower the MSE value. This means that to obtain more accurate prediction results, the n_estimator value must be large.