A novel model by evolving partially connected neural network for stock price trend forecasting
Expert Systems with Applications: An International Journal
| Hi-index | 0.00 |
In our economic society, future stock price trend is very hot focus that the investors concern about. Challenges still exist in stock price prediction model regarding significant time-effectiveness of prediction, the complexity of methods and selection of feature index variables. In this paper, we present a new approach based on Logistic Regression to predict stock price trend of next month according to current month. Characteristics of our method include: (1) Feature Index Variables are easy to both understand for the private investor and obtain from daily stock trading information. (2) the prediction procedure includes unique and crucial operation of selecting optimizing prediction parameters. (3) significant time-effectiveness and strong purposefulness enable users predict stock price trend of next month just through considering current monthly financial data instead of needing a long term procedure of analyzing and collecting financial data. Shenzhen Development stock A (SDSA) from RESSET Financial Research Database is chosen as a study case. The SDSA’s daily integrated data of three years from 2005 to 2007 is used to train and test our model. Our experiments show that prediction accuracies reach as high as at least 83%. In contrast to other methods, e.g. RBF-ANN prediction model, our model is lower in complexity and better accuracy in prediction.