Predicting the French Stock Market Using Social Media Analysis

In this article, we try to predict the next-day CAC40 index. We apply the idea of Johan Bollen et al. from [1] on the French stock market and we conduct our experiment using French tweets. Two analysis are applied on tweets: sentiment analysis and subjectivity analysis. Results of these analysis are then used to train a simple neural network. The input features are the sentiment, the subjectivity and the CAC40 closing values at day-1 and day-0. The single output value is the predicted CAC40 closing value at day+1. We propose an architecture using the JEE framework resulting in a better scalability and an easier industrialization. Experiments are conducted over 5 months of data. We train our neural network on 3/4 of the data and we test predictions on the remaining quarter. Our best run gives a direction accuracy of 80% and a mean absolute percentage error (MAPE) of 2.97%. Results are not as good as those reported in [1] but many modifications can be applied in order to improve the performance: classically by adding more features but we mainly aspire to a system where the user can integrate its own expertise.