|
COMPARATIVE ANALYSIS OF TRADITIONAL AND SOFT COMPUTING MODELS FOR TRADING SIGNALS PREDICTION |
|---|---|
| รหัสดีโอไอ | |
| Creator | Beenish Bashir, Faheem Aslam |
| Title | COMPARATIVE ANALYSIS OF TRADITIONAL AND SOFT COMPUTING MODELS FOR TRADING SIGNALS PREDICTION |
| Contributor | - |
| Publisher | TuEngr Group |
| Publication Year | 2563 |
| Journal Title | International Transaction Journal of Engineering, Management, & Applied Sciences & Technologies |
| Journal Vol. | 11 |
| Journal No. | 4 |
| Page no. | 11A04J: 1-16 |
| Keyword | Moving averages, Autoregressive models, Artificial neural network, Support vector machine, Financial marketing, Extreme learning machine. |
| URL Website | http://TuEngr.com/Vol11_4.html |
| Website title | ITJEMAST V11(4) 2020 @ TuEngr.com |
| ISSN | 2228-9860 |
| Abstract | Financial market forecasting always remains a challenging issue due to the nature of the time series data as well the information stock prices reflect. Economist takes this data as a linear process whereas the soft computing models assume that time-series data is nonlinear, complex and dynamic in nature and can be better analyzed through nonlinear models. This research paper settles the debate by conducting a comparative analysis of traditional forecasting models and soft computing models for the trading signals prediction of Pakistan stock exchange covering the daily data from 1997 to 2018. Our study is unique regarding the target variable it predicts which reflects the overall dynamics of the market not just the next period future price. In traditional models, we include moving averages as well as autoregressive models whereas in soft computing models we take artificial neural networks, support vector machines and extreme learning machines to have a comprehensive analysis. This comparison shows that soft computing models perform better than traditional models in trading signals prediction showing non-linear behavior of financial time series data. However, amongst all soft computing models, the artificial neural network becomes the best predictive model. Our results are more convincing as compare to existing literature regarding the careful selection of market features, which can help investors to better understand market behavior and improve the predictive ability of the model. |
