Implementation of Diffusion Variational Autoencoder for Stock Price Prediction with the Integration of Historical and Market Sentiment Data

Abstract

This study aims to predict stock prices using a Diffusion Variational Autoencoder (D-VAE) model that integrates technical data and market sentiment. Technical data is obtained from historical stock prices and trading volume, while sentiment data is derived from financial news analyzed using the IndoBERT model for sentiment classification. The research findings indicate that the integration of sentiment data in the D-VAE model enhances the accuracy of stock price predictions compared to a model that uses only technical data. Model evaluation is conducted using metrics such as Mean Squared Error (MSE), Mean Absolute Error (MAE), and R-squared (R²). The model with sentiment data integration has an MSE of 2753.204, MAE of 42.751, and R² of 0.94489, which are better than the model without sentiment data integration. This study demonstrates that the use of sentiment analysis can significantly contribute to improving stock price prediction performance using machine learning technology.