Predicering av aktiekursutveckling för svenska aktier utifrån konjunkturdata

Sammanfattning: This study aims to investigate whether Swedish economic indicators can be used to predict stock market performance on the Stockholm Stock Exchange. The study is expected to contribute to new research in the field and also explore the potential utility of these predictions for individual investors. Using Gaussian Naive Bayes, models have been developed to classify a stock’s price development one month in advance. The objective of these models is to maximize their classification ability. This is achieved by evaluating various machine learning performance metrics for the models and assessing a simulated stock portfolio invested by the models against the OMXS30 stock market index. The results reveal rather weak machine learning performance metrics for the models. However, the simulations demonstrate significant returns, with the models’ predictions outperforming the index by as much as 439 percentage points over the period from 2015-01-01 to 2023-04-01. Thus, this study concludes that it is possible to predict stock market development with some success using economic indicators. However, the use of the economic indicators has several limitations, that might be misleading when used with historical data and may explain the high returns of the simulation. Furthermore, the reason for the discrepancy between the performance metrics and portfolio simulations cannot be determined based on this study alone, but several hypotheses are presented. Further investigation is therefore warranted.