Percorrer por autor "Beeger, Dominik Lukas"
A mostrar 1 - 1 de 1
Resultados por página
Opções de ordenação
- Stock market prediction via machine learning and investor sentiment data : a quantitative investment strategyPublication . Beeger, Dominik Lukas; Tran, DanThis study shows sentiment data’s ability to predict stock prices by applying modern machine learning models and investment strategies on data from different industry sectors. Sentiment-based investment strategies outperform benchmarks in return and risk measurements. The findings are consistent with previous academic research that already proved sentiment’s forecasting ability on major stock markets with other statistical methods. The author finds that sentiment-based strategies work best for negative performing industry sectors and high volatile sectors. In general, high volatile time periods enable sentiment-based models to predict stock returns more accurately; in low volatile periods, sentiment strategies underperform. Additionally, data based on single companies supposedly provide clearer signals than index-based features; thus, it can be concluded that input data lose significance through indexing. Based on these results, the author created an investment strategy that can be used for further research and professional investment strategies The author also finds out that high accuracy scores of applied machine learning models must not be followed by high financial performances. This finding can be explained by the complex distribution of returns. In our case, the majority is concentrated on returns with small magnitudes, which increases the importance of days with extremely positive and negative returns. These returns determine the overall performance more than returns with small magnitudes. It can be concluded that in finance data relations are more complex, which is why it is more important to adapt models to the data structure instead of maximizing machine learning performance.