Predicting Stock Market Using Reddit Comments
Year: 2024
Type: preprint
Abstract: This study investigates the relationship between Reddit comment sentiments and stock market performance, focusing on predicting the S&P 100 index using data from 2007 to 2015. The researchers employed a large-scale distributed machine learning sys- tem, combining sentiment analysis using SentiWordNet 3.0, dimensionality reduction through Principal Component Analysis (PCA), and logistic regression modeling. Key findings reveal observable correlations between Reddit sentiments and stock prices, with past sentiments more predictive of future market performance. The study achieved a best 5-fold cross-validation accuracy of 65.47% and a test accuracy of 64.87%. Re- sults indicate that aggregating sentiment data over longer periods and focusing on long- term cumulative trends improve prediction accuracy. This research suggests that an- alyzing social media sentiments can provide valuable insights for stock market pre- diction, though further investigation is needed to refine the approach and explore its application to individual stocks.
Source: HAL
Authors: Chen, Yilin, Ma, Heting
Institution: Xiamen University
Language: English
Field: Decision Sciences
Domain: Social Sciences
Open Access status: green
OpenAlex ID: W4401407151