New approach to stockforecasting using HMM methods combined with model averaging. The HMM method is based on
NGUYEN, N. (2016): “Stock Price Prediction using Hidden Markov Model,” (https://editorialexpress.com/cgi-bin/conference/download.cgi?db_name=SILC2016&paper_id=38)
Several Models with a different number of hidden states are trained and their results are averaged by three different methods:
- Naive Model Averaging
- Total Model Averaging
- Selective Model Averaging
The results of testing this new approach on 4 different real data series shows better results compared to the standard method.
For the whole model description and the explanation of the research approach see the attached pdf file (German only)