The application of neural network system for multi-dimensional stock market data analysis is presented in the paper. Developed system predicts stock price movements based on daily quotation data like: volume, minimum and maximum session price, opening and closing price. Several studies were carried out, to compare systems investment decisions, with decisions that were made on the basis of some commonly used methods of stock market analysis. These methods are: MACD, Bootstrap, Markowitz Portfolio. For valuation purpose, the real stock market data of the four largest Polish companies were used. All companies are quoted on the Warsaw Stock Exchange and belong to the WIG 20 index. For the benchmarking, only stock data from the year 2009 were used. In order to enrich the benchmarking tests, three investment scenarios were added. First known as the skeptical assume that only incorrect investment decisions were made. Second known as the optimistic assume that only correct investment decisions were made. Last one known as passive assume that no investment decision were made - it is so called "buy and hold" conception. The benchmarking results confirmed, that the neural network system is able to make investment decisions, that significantly increase the profitability of the investment portfolio. Neural network system provide investment suggestions, that can be considered as an alternative to other commonly used methods of stock market analysis. However statistical tests proved a high correlation between quality of systems investment decisions and market trend and lack of correlation to the "optimistic" scenario. Neural network systems may help in investment process, but cannot be considered as fully reliable way of investment process automation.
JavaScript is turned off in your web browser. Turn it on to take full advantage of this site, then refresh the page.