The use of optimized random forest in analyzing the prediction of stock trends in Indonesia

Stocks are ownership symbols of an individual or an institution, which are long-term financial instruments and can be traded. This research employs the optimized random forest as the primary method to assist in predicting stock trends in the energy sector. The stocks used are ADRO.JK, INDY.JK, PTBA.JK, TOBA.JK, and UNTR.JK. The first step involves identifying the technical indicators that will be used as variables to aid in predicting stock trends, using adjusted closing price of the stocks. The model will be constructed using two sets of data: stock-based and year-based data. There are four stages in the model development: random forest model, optimized random forest model, random forest model with feature importance, and optimized random forest with feature importance. The model will be evaluated based on accuracy, F1 score, and AUC value. The results obtained from the optimized random forest method will be compared with those from the random forest method. This research demonstrates that the optimized random forest is overall a more accurate method for predicting stock trends either in stock-based and year-based data based on the F1-score. Furthermore, the predictions from the best-performing model will be used in a stock trading simulation.