Research On Time Series Prediction And Program Trading Based On QPSO-RPNN

According to the Chaos Theory,if a financial market has chaos feature,then the law hidden in a time series can be found by an appropriate model and then being used to make prediction.Based on the theory,the paper studies Recurrent Predictor Neural Network and Quantum-Behaved Particle Swarm Optimization on the purpose of constructing a time series prediction model.RPNN is a dynamic time delay neural network invented for time series.Training RPNN by QPSO can overcome the fault of run into local optimum and solve imprecision problems.Firstly,the paper used appropriate methods to construct the QPSO-RPNN model,then simulation and prediction were applied on Shanghai composite index with different periods,and prediction results were analyzed.The results show that the model makes effective short term prediction.Finanly,this research combined theory and practice by using QPSO-RPNN to make simulation and prediction on selected stocks.Based on the prediction results,daily program trading and 15-minutes short-term trading were applied.The prediction results and the program trading results proved that QPSO-RPNN has high precision of prediction and thus it can be used in program trading.