investorguide.co.za
Home | News | Services | Downloads | Forum | About us | Contact us
Business Analytics
Explanation of business analytics and the services provided by Investorguide
Financial time-series prediction
Example project and results of stock market prediction with machine learning techniques
Resources
Investment Strategies
Discussion of various
investment strategies
Publications
Fundamental Financial
Concepts
Investment Planning
Investing
Financial Statment Analysis
Fundamental Investing
Strategies
Advanced Investing
Strategies
Terminology
Investment Terminology
Glossary of investment terms
Financial time series prediction
We have successfully employed support vector machines (SVMs) on financial time series data to predict the index values of the Johannesburg Stock Exchange on a monthly basis for the period of September 2006 to June 2008. A relative gain of 37.3 % over the market was obtained with an absolute percentage gain of 8.91 %. We made use of various macro-economic indicators to perform these predictions.The following figures illustrate the results that were obtained.
Figure 1. SVM predictions vs. target index values
Figure 2. Cumulative returns of SVM predictions vs. the market return
Figure 3. SVM monthly percentage gain/loss
Figure 4. Market monthly percentage gain/loss
The final results are summarized in Table 1.
|
Performance measure |
Percentage (%) |
|
SVM
return on investment |
32.8 |
|
Market
return on investment |
23.89 |
|
Absolute
percentage gain over market |
8.91 |
|
Relative
gain over market |
37.3 |
|
Classification
accuracy |
72.72 |
Table 1. Performance measures of SVM predictions
Details regarding the construction of the data set and implementation of the SVMs are confidential at this stage.
Financial 24
Latest news and stock market figures for JSE
Google Finances
Excellent source of financial news and figures