In this project, we checked the possibility of predicting stock prices using current affair events, which is expressed by words’ commonness in internet blogs and forums.
In this project, we checked the possibility of predicting stock prices using
current affair events, which is expressed by words’ commonness in internet
blogs and forums. It is known that there is a strong bond between current
affair events and stock prices. Moreover, there is also a connection between
public mind-set and stock prices. In this project, we will try to find an
empiric correlation between the words commonness to stock prices using statistical
methods. This project includes computer vision, machine learning and optimization.
Although there have been many approaches to predict stock prices, there
have been little success due to the large uncertainty in this area. Hence,
in this project we used statistical techniques based on current affairs
which we believe improves the prediction results.
The algorithm was implemented using Matlab-7.3.0 environment.
Results and Conclusions
On a 142 days period, we earned 5%.
The optimal estimator gained 35% profits, the worst estimator, lost -30%,
and the average investor (buy at the beginning – sell at the end), lost
-2%. Focusing the last three months, our estimator earned 9%, while the
ideal was 25.5%, the worst -18%, the average 1.5%.
Therefore, we conclude that:
- We didn’t lose any money, even though it was a period of many uncertainties
(-30% loss was the largest downfall). The results were good, although
we learned on a different period (mostly rising).
- Indication to the above is the fact that the last three months were
- When we tested some other estimators, which were chosen by other
predefined characteristics, we gained only 1-2% profits.
For conclusion, we can see that the results are good, and might be improved
by several suggestions you can find in our project book.
 R.O.Duda, P.E. Hart,D.G Stork – Pattern Classification and Scene
Analysis (Second edition)
We would like to thank our supervisor Dani Pinkovitch for his support
and guidance throughout this project.
We are also grateful to the Ollendorf
Minerva Center Fund for supporting this project.