Applications of data mining in empirical finance

OData support
Supervisor:
Zlatniczki Ádám
Department of Computer Science and Information Theory

In my thesis I forecast stock price movement wherefore to create a profitable investing strategy. For the forecasting I use data mining techniques, especially classification. My goal is to build such a binary classifier that tells us whether the given stock price will increase or decrease tomorrow. I use the previous days’ closing prices and the previous days’ movement directions during the model construction. I try out various kinds of classifiers for the strategy building like decision tree, random forest, Bayes-classifier and Support Vector Machine (SVM). I compare the results of the different type of classifiers with various performance indices. Then use the best model to build an investing strategy and compare my strategy against a benchmark.

To achieve this results, I read some articles about similar researches as mine to collect ideas and analyse theirs results. I processed the theoretical aspects of the applied techniques and I wrote a summary of them. Then I downloaded the AT&T stock prices from the Yahoo! Finance and made some descriptive statistics to be familiar with the data. After this I used the data from AT&T stock through several iterations to find and improve the results of the four classifiers. In the end I defined the Support Vector Machine as the best classifier, therefore I used that to predict movements and built an investment strategy based on that. I compared my strategy against the Dow Jones Industrial Average. As curiosity I tried to find association rules among the stocks movements. My goal with it was to explore connections between the stock price changes.

Downloads

Please sign in to download the files of this thesis.