Study On Hybrid Neural Network Application In Petroleum Price Prediction

Petroleum is one of the indispensable energy for development of world economy and politics, which brings an important influence on the development of the national economy. The petroleum price is not only influenced by international military affairs and politics, but also affected by the situation of economy and diplomacy at the same time, with some characters such as uncertainty and complex nonlinear. To grasp the trend of the petroleum price is the key means to reduce the negative impact of petroleum price changes, how to accurately predict the price petroleum become the hot issues in many countries around the world. But due to the influence factor of the petroleum price is too much and the petroleum price data contains noise, it has much more deficiency in the traditional petroleum price prediction methods, such as qualitative analysis only, unreliable prediction results or insufficient prediction precision. Therefore, research on prediction model with higher precision, better performance, efficiency, correlation and wider adaptability are still the first-line goal in academia and industrial community.By means of the review of the petroleum price situation both at home and abroad, the advantages and disadvantages of the petroleum price prediction model have been analyzed and compared firstly. In this thesis, a hybrid neural network model to predict the petroleum price is proposed based on the combination of particle swarm optimization algorithm, chaos theory and artificial neural network technology. The main research content are as follows:1. A novel improved particle swarm algorithm (CSAPSO) is proposed. The standard particle swarm optimization has been improved by combining the self-adaptive weight adjustment strategy and chaos theory, the adaptive weight adjustment strategy is used to improve the convergence speed of algorithm, and the chaotic sequence produced by chaos theory is used to tune the learning factor with the purpose of balance the exploitation and exploration, and improving the premature convergence problem of the algorithm. 2. A hybrid neural network model (HANN) is proposed. The hybrid training algorithm is combined with the CSAPSO and BP algorithm, the CSPSO-BP ANN model is developed trained by the hybrid algorithm based on chaos self-adaptive PSO and BP algorithm. The hybrid training algorithm combines the strong global search ability of particle swarm optimization algorithm and strong local search ability of BP algorithm, which improve the prediction performance of the model.3. According to problems of petroleum price prediction and the feasibility of petroleum price prediction model, the HANN model for petroleum price prediction is developed. Through the petroleum price test example, it is shown that the HANN model is feasible and reliable to predict the petroleum price. Compared with conventional BP ANN and PSO-BP ANN, the HANN shows better performance with better accuracy and correlation.In this thesis, a high performance petroleum price prediction model is proposed based on PSO algorithm, chaos theory and ANN technology combined with the petroleum industry development strategy. It may be used to provide a viable approach for he petroleum industry prediction. At the same time, the proposed CSAPSO algorithm may be used for reference for design in many industrial and research fields, it has a good application prospect.