Discovering Stock Market Trading Rules Using Multi-layer Perceptrons

Piotr Lipinski

Abstract

This paper presents an approach to extracting stock market trading rules from stock market data. Trading rules are based on two multi-layer perceptrons, one generating buy signals and one generating sell signals. Inputs of these perceptrons are fed with values of technical indicators computed on historical stock quotations. Results of a large number of experiments on real-life data from the Paris Stock Exchange confirm that the model of trading rules is reasonable and the trading rules are able to generate reasonable trading signals, not only over a training period, used in the training process, but also over a test period, unknown during constructing trading rules. Moreover, trading strategies defined by such trading rules are profitable and often outperform the simple Buy&Hold strategy.
BibTeX
@inproceedings{lipinski2007discovering,
  author = {Piotr Lipinski},
  title = {{Discovering Stock Market Trading Rules Using Multi-layer Perceptrons}},
  booktitle = {IWANN},
  year = {2007},
  doi = {10.1007/978-3-540-73007-1\_135},
}

Loading related content...

Figure