A Study On Application Of Elman Neural Network

Some foundation knowledge based on artificial neural network is introduced in this paper. Moreover the dynamic feedback artificial neural networks, namely, the Elman neural network (Elman NN) and output-input feedback Elman neural network (OIF Elman NN), whose mathematics models and learning algorithms, are also introduced respectively.Considering that the main purpose of investors is to obtain more profit, in order to get more accurate forecasting precision and obtain more accurate results, then obtain more profit. Based on the Elman NN, an improved model is proposed, namelyTPFENN (Time Profit Factor Elman Neural Network) and it is applied to forecast the composite index of stock. Comparisons and calculations of stock profit are also made when the stock exchange is performed using prediction results from different models. To evaluate the effectiveness of the prediction models, we use two types of error functions to represent the forecasting precision, i.e., the least square error (LSE) and absolute average error (AAE). Experimental results show that the proposed Elman NN model is effective and feasible in the field of stock investing, which can not only improve forecasting precision for Elman NN obviously but also more profits can be obtained. Therefore it can realize the main objective to obtain more profit as possible as for investors.In addition, in order to control the atmospheric pollution and forecast the atmospheric change timely in future, which can prevent the serious pollution occurring. Based on the OIF Elman NN, this paper proposed an improved OIF Elman NN model by introducing direction profit factor to OIF Elman NN, which is called DPFOIFENN (Direction Profit Factor OIF Elman Neural Network). Moreover DPFOIFENN model is used to forecast and assess the atmospheric pollution of Changchun City in the Northeast of China. Experimental results show that DPFOIFENN model developed in this paper can obviously improve the prediction precision of OIF Elman NN and it has the character of the better approach performance, which is basically accordant to the reality.So TPFENN and DPFOIFENN models proposed in this paper can provide novel technology and effective approaches for finance investing and the atmospheric pollution forecasting and assessment, which has greatly potential and good application...