Stock Price Prediction During the Pandemic Period with the SVM, BPNN, and LSTM Algorithm

Abstract

The stock market volatility during the pandemic was a challenge that affected investors' decisions in making their investments. Machine learning was one of the options to cope with the issue, for it helped develop a predicted algorithm that analyzes time series data as part of the investor's investment consideration. Thus, the algorithm in machine learning can be the answer to the issue. The three comparable algorithms included SVM, BPNN, and LSTM within the BBRI stock report case study from November 14, 2019, to November 13, 2020. The study compared those three algorithms to figure out which is the best one. This research emphasizes CRISP-DM methodology, business understanding, data comprehension, data preparation, algorithm development, evaluation, and deployment. This research concluded that SVM has the best prediction accuracy with 0,003 MSE and 0,058 RMSE, followed by LSTM with 0,008 MSE and 0,087 RMSE, and lastly BPNN with 0,017 MSE and 0,132 RMSE. Reviewing this trend, SVM had the closest forecast to the exact result. BPPN had the highest RMSE, nevertheless, it showed a closer forecast to the exact result, compared to LSTM. This research benefits investors in delivering more accurate predictions to execute accurate decisions regarding stock forecast and investment.