Using Back Propagation Neural Network In Selecting Securities

In this paper, portfolio selection based on data mining technology is discussed for Chinese stock market. Two parts from the view of technology analysis and fundamental analysis are investigated respectively. To do technology analysis, 100 stocks from Shanghai Stock Exchange are chosen as samples. Using their transaction data and ten technology indexes, three principle factors are achieved by means of principal component analysis.Then, a three layer neural network is constructed and trained with real data. Its input-s are from the principle factors above and the outputs are the returns of stock. After training, a portfolio selection approach based on the above neural network is delivered,Meanwhile, the efficiency of the method is examined by new data. Similar process is also done for fundamental analysis, that is, another principal component analysis and neural network model are taken out. Lastly, the feasibility of medium-term investment strategy is also discussed with empirical data. It is satisfactory that the test shows the model is efficient.