Some Improved Methods Based On Computational Intelligence And Their Applications In Financial And Environmental Fields

Computation intelligent is based on the idea of bionics and the knowledge of biology intelligence to simulate and realize human being intelligence by using numeric calculation. It is a new crossover subject based on mathematics, physics, biology, psychics, neurology, computer science and intelligence technology, which is also studying focus in the world. Its theory and application study has made greatly progress and produce enormous economic profits and society benefit, which has been applied widely in many fields such as industry, agriculture, national defence, engineering, traffic, finance, communication etc. Especially in the fields of finance and environment, many economists, mathematicians, meteorologists and some reachers of computer are focusing on the study of computational intelligent methods. They use computational intelligent methods to solve some problems in the fields of finance and environment and some satisfactory results are also obtained.Some theoretical study and applications study based on the approaches of computation intelligence are made in the paper. The major contents could be summarized as follows:(1) In order to enhance the dynamic competition and clustering capability of self-organizing map (SOM) neural network and improve the precision of solutions, multi-winners SOM models are proposed by extending numbers of winners based on an unsupervised SOM neural network and improving the neighbor function and weight function in the network. In addition, a tabu-mapping method is proposed to avoid that the same output node is mapped by more than one input. The clustering analysis for the stock is used to examine the effectiveness of the proposed models. The financial indexes reflecting the whole performance level of companies are used in the simulated experiments. Simulation results show that the clustering effect of the SOM with 2 winners (SOM2W) is the best compared with those from the standard SOM and other proposed multi-winner SOMs. The proposed model could provide a feasible approach for analyzing and selecting stock, which has potential applications in the financial field.(2) Considering that the main purpose of investors is to obtain more profits, it seems that the profit should be paid more attention rather than forecasting precision when the stock indexes'prediction is performed. By experimenting, it could be found that the profits of stock exchange dealt according to the prediction results are not always accordant with the forecasting precision but accordant with the correctness of the fluctuation trend forecasting. While by further observing, it was found that if the fluctuation trend of stock indexes is predicted correctly by the calculating model, the profits are always high; and vice versa. Therefore considering this point, we propose improved Elman neural networks by introducing direction profit factor and time profit factor into Elman model, which are called ENNDPF, ENNTPF, ENNDTPF network. Experimental simulation results show that the proposed models are feasible and effective in the field of stock investment. They could improve evidently the forecasting precision and achieve the aim for obtaining more income compared with Elman model. Therefore it could be a novel effective approach applied to the financial field.(3) At present, investors'investment ideas get to science and reason during they are investing. So reference gist has important value and reality significance. Considering the factors, in order to obtain a new feasible, scientific, reasonable stock price approach, we applied SOM2W model to simulate clustering through a large number data in this paper. So the property and category of new stock companies can be confirmed. In order to make the price of new stock can express the value of stock companies in deed, the financial indexes reflecting the compositive performance of companies are used as data sample in the simulated experiments. Radial basis function (RBF) neural network is used to confirm the new stock price by simulating the black box of stock market reasonable. Experimental simulation results and numeric results show that the proposed SOM2W-RBF network can provide a new reference tool for making new stock price.(4) The air environment is closely linked with human health and life, whereas the atmospheric quality has been deteriorating with the quickening rhythm of economic growth and industrialized progress. The problems regarding the atmospheric pollution have attracted more and more attention. In order to control and evaluate the grade of the atmospheric pollution, in this paper, particle swarm optimization (PSO) algorithm is used to optimize the parameters in the universal formula for calculating harm rate of atmospheric pollution, therefore the calculating harm rate formula and index formula of atmospheric pollution suited for the cases of multi-pollutants can both be obtained. In addition, the comprehensive models based on the particle swarm optimization algorithm that have optimization advantage are also proposed in this paper, which include the model of calculating the harm rate of pollution and the harm index of pollution for the assessment of atmospheric quality with multi-pollutants. The models are applied to assess the atmospheric pollution of a city in the Northeast of China. Experimental results show the advantages of the proposed models, such as pellucid principle and physical explication, predigested formula and low computation complexity.(5) In order to forecast timely and correctly atmospheric change, avoid the serious pollution events and improve people's living quality, the atmospheric quality forecasting has become an important research subject. Considering the factors, in this paper, in order to obtain more accurate forecasting precision, three improved OIF Elman neural networks are proposed by introducing the direction and/or time profit factors into the output-input feedback (OIF) Elman neural network. Simulations show that the proposed models are pretty well, which could improve the forecasting precision evidently. The accurate assessment results could be obtained by using the improved models. Experimental results show that the proposed OIF Elman neural networks are feasible and effective in the fields of forecasting and assessment of the atmospheric quality, which has great potential in the field of atmospheric environment.