Analysis of Closing Stock Price Predictions Energy Sector in Indonesia Using Convolutional Neural Network

In industry, accurate stock price prediction is crucial as it allows investors to better plan and allocate resources. This study uses CNN with window sizes of 20 and 100 to predict the closing prices of Indonesia's five largest market capitalization energy sector stocks. While existing methods like LSTM have proven effective, they are not always practical for all scenarios. The research investigates the effects of long and short-term window sizes on model accuracy. Additionally, different filter configurations for three CNN layers (64, 128, 64; 32, 64, 32; and 8, 6, 4) are applied. The model can predict the next day's stock price, enhancing decision-making and reducing financial risks for investors. The model's accuracy is analyzed using each stock's median, mean, and variance of the mean squared error. Results indicate that the 100-window size model outperforms the 20-window size model in predicting stock prices.