Optimization of Stock Price Prediction Using Long Short-Term Memory (LSTM) Algorithm and Cross-Industry Standard Process Approach for Data Mining (CRISP-DM)

Predicting stock prices accurately is an integral part of investment analysis as it permits forecasting movements in the financial markets and tailoring strategies accordingly. In this study, the LSTM (Long Short-Term Memory) algorithm is used with the aim of improving predictive accuracy, particularly the forecasting of stock price movements. This research follows the CRISP-DM framework or Cross-Industry Standard Process for Data Mining, which incorporates six defined steps including: understanding the business context, data understanding, data preparation, model building, evaluation, and implementation. Stock price data for the ticker symbol “ANTM.JK” was sourced from Yahoo Finance for the date range of October 29, 2005 to July 11, 2024. Along with the consistency, several model accuracy enhancing preprocessing steps such as data cleaning, feature selection, and normalization with Python were performed before modeling. Hyperparameter tuning to reduce the error margins on predictions was conducted after training the LSTM model. Testing the hypotheses showed that the LSTM model demonstrated a low Root Mean Square Error (RMSE) on the test dataset indicating outstanding forecasting accuracy. The ability of the model to outperform conventional time series forecasting techniques is attributed to its ability to effectively retain nonlinear time-series relationships and long-term dependencies. These findings suggest that the LSTM algorithm can serve as a reliable tool for stock price forecasting in emerging markets. This study provides practical insights for investors and lays the groundwork for future research on hybrid or ensemble models to further improve prediction robustness and accuracy