Prediction of Stock Trend Using Random Forest Optimization

Stocks are financial investment instruments used by investors to generate a profit. Depending on the prevailing trend, investors typically profit from fluctuating stock price movements. Researchers in the field of artificial intelligence and investors are interested in this because the development of technology can assist investors in making decisions. One of the artificial intelligences utilized in this study is machine learning employing the Random Forest method to forecast stock market trends. Random Forest was selected because the ensemble method algorithm is particularly well-suited for use with data containing numerous features. It's just that this machine learning approach is not optimal because it requires adjustments based on the prediction target and has poor accuracy. Consequently, this study optimizes Random Forest by utilizing Random Search to obtain the optimal parameters for the classification model. In order to reduce noise in the data, Exponential Smoothing is applied to the features that will be utilized. Using the Random Forest method, this Research predicts stock price trends from transactional data. The dataset consists of historical stock transactions obtained from tradingview for 10 IDX stock issuers from several stock indices that are deemed to represent several areas of the company's industrial sector. The transaction data period is from November 2016 to November 2021, with a total of 1200 days of transaction data per issuer. This research includes the steps of data collection, data preprocessing, feature extraction, training mode, and model testing. Multiple methods, including F1-score, accuracy, precision, and recall, will be employed to evaluate the constructed model. The constructed model is compared to a random forest model in which the default parameters are used and the data is not smoothed. Based on the results of the comparison test, the optimized model has an average evaluation value that is 14.89% higher than the default random forest model.