Improving Financial Time Series Prediction Through Output Classification by a Neural Network Ensemble
One topic of great interest in the literature is time series prediction. This kind of prediction, however, does not have to provide the exact future values every time: in some cases, knowing only time series future tendency is enough. In this paper, we propose a neural network ensemble that receives as input the last values from a time series and returns not its future values, but a prediction that indicates whether the next value will raise or fall down. We perform exhaustive experiments to analyze our method by using time series extracted from the North American stock market, and evaluate the hit rate and amount of profit that could be obtained by performing the operations recommended by the system. Evaluation results show capital increases up to 56 %.
KeywordsArtificial neural networks Classification Prediction Stock markets Time series
This work is partially supported by InWeb (Brazilian National Institute for Web Research), under the MCT/CNPq grant 45.7488/2014-0 and by the authors individual grants and scholarships from CNPq (National Counsel of Technological and Scientific Development) and CAPES (Coordination for the Improvement of Higher Education Personnel).
- 4.Hansen, L.K., Liisberg, C., Salamon, P.: Ensemble methods for handwritten digit recognition. In: Proceedings of the 1992 IEEE-SP Workshop on Neural Networks for Signal Processing  II, pp. 333–342. IEEE (1992)Google Scholar
- 5.Heaton, J., Reasearch, H.: Encog java and dotnet neural network framework. Heaton Research Inc, 20 July 2010Google Scholar
- 9.Silva, E., Castilho, D., Pereira, A., Brandao, H.: A neural network based approach to support the market making strategies in high-frequency trading. In: International Joint Conference on Neural Networks (IJCNN). pp. 845–852. IEEE (2014)Google Scholar
- 10.Taylor, B., Kim, M., Choi, A.: Automated stock trading algorithm using neural networks. In: Juang, J., Chen, C.Y., Yang, C.F (eds.) Proceedings of the 2nd International Conference on Intelligent Technologies and Engineering Systems (ICITES2013), Lecture Notes in Electrical Engineering. vol. 293, pp. 849–857. Springer, Switzerland (2014)Google Scholar