My task was to develop a stock market decision making system which downloads actual equity prices from Budapest Stock Exchange, and by evaluating the market situation suggests further trading to the portfolio holder.
First I reviewed the related literature to define trading rules which I can implement. I found some frequently followed strategies in the real business, and I implemented three of them in my framework.
The Long Strategy buys when the prices going up and sells when the prices fall, while the Short Strategy does the opposite. Beyond these primitive strategies I used WEKA API to develop a more intelligent, machine learning aided rule, which forecast the price in the next period.
The framework was written in Java, running on a MySQL database. I implemented two data providing methods: the required retrieving algorithm, which frequently parses and archives the exchange rates from the website of the Budapest Stock Exchange, and another one, which continuously ‘feeds’ the decision maker with historical data.
The tests ran on this historical dataset including the BUX index and four equities, which means the input data was genuine. The results were various, depending on the analyzed stock, but the WEKA Time Series evaluator provided more than satisfying results on the entire historical dataset.
In the nest semester I would like to continue the development of the system by developing all of the planned decision rules. All of the methods came up with a very good result at least on one stock. My goal is to find something criterion which allows to assign decision rules to equities to provide better results.
In the future I would like to improve the graphical user interface, and implement one from the potential client-side interfaces.