The Study On Trade Index Of Futures Based On Optimized Neural Network

Neural network is an important research method in the field of machine learning,andneural network prediction ability is widely used in the field of finance.In the futures market,the research on the trading strategy for Programmed Trading is deepening,this paper aims to optimize the neural network,then train and predict,improve different trading strategies by studying the trading index that based on the prediction results.The prediction error of neural network prediction is very large due to insufficient information data storage,simple structure and other issues.Also,the prediction results vary greatly,and full of uncertainty with different initial weights.To solve this problem,this paper presents a multivariate prediction method.For the NARX neural network input,multiple external variables are allowed to join in the training,and they could be the last prediction index.In this paper,the forecast data is the price of soybean meal.Through detailed analysis,this paper sums up ten factors influencing the price of soybean meal,and takes them as the external input of neural network.At the same time,in order to avoid the repeated use of external factor information,this paper uses principal component analysis to reduce the dimension of the ten external variables,thus selects four uncorrelated principal components,which not only preserve the most information of original data,avoid the data redundancy,but also improve the training speed of neural network.In this paper,by comparing the multivariate prediction results with the univariate prediction results,we can find that multivariate prediction is more accurate.In this paper,the prediction results stored in the commodity data of Trade Blazer are used for the study of trading index,meanwhile write a trading strategy independently in the TradeBlazer platform and then optimize the parameter,to find the problems and add the prediction results to the trading strategy.Through several experimental analysis,the trading index that based on the prediction results could greatly improve the trade strategy.In order to further verify the rationality of the trading index that based on the prediction results,this paper adds the index to the different trading strategies to conduct the experimental analysis.The results show that the trading index that based on the prediction results can effectively improve the transaction success rate of the general trading strategy.