Can AI Predict the Stock Market? A CNN-BiLSTM Based Analysis of Macroeconomic Indicators' Effect on OMXS30

Abstract

Forecasting time series within the financial markets is one of the most fundamental subjects in economics. These markets are highly complex due to their nonlinear behaviour, where uncertainty and unpredictability play a central role. Fluctuations occur without any clear patterns, resulting in more challenging investment decisions for all the market participants. Based on four domestic macroeconomic indicators, this thesis predicts the return of the OMXS30 index by implementing deep learning approaches, specifically the CNN-BiLSTM model. Analysis reveals that while increasing the model complexity by adding additional variables, the out-of-sample test performance R_OS^2 improved by 12.1 %. The observed results indicate the model’s ability to explain 12.1 % of the variation in the OMXS30 index returns by including all the selected variables. Despite improvements in predictions for both the training and test data, the model still struggles with limited generalization, exhibiting a mild overfitting.