摘要
为了提高锌矿价格预测精度,采用改进的QPSO算法优化BP网络的权值与阈值,将通过优化搜索得到的粒子位置向量解码作为网络的权值与阈值,优化BP神经网络,对锌价格进行建模预测。在输入因子相同的条件下,以PSO-BP与QPSO-BP模型分别预测未来锌矿价格行情,以预测精度(MAPE)和泛化能力指标(ARV)评定两种模型的优劣。结果表明,改进的QPSO-BP模型的预测精度和泛化能力明显高于PSO-BP模型,更能适用于锌价格预测,对项目投资决策和风险评估有一定的参考价值。
In order to improve the prediction accuracy of zinc prices, this paper optimizes the BP neural network and constructs the prediction model of zinc prices with optimizing BP network weights and threshold by the improved QPSO algorithm. The particle position vector is obtained by optimization search decoding as network weights and threshold method. Under the conditions of same input factors, PSO-BP and QPSO-BP models were used to predict zinc prices, and the merits of the two models were assessed by predict accuracy （MAPE） and the generalization ability index （ARV）. The results show that the improved QPSO-BP model is better than PSO-BP model in the prediction accuracy and generalization ability significantly. The method provides some reference value for the study of investment decision-making and risk assessment.
出处
《有色金属（矿山部分）》
2014年第4期101-106,共6页
NONFERROUS METALS(Mining Section）
关键词
价格预测
量子粒子群算法
QPSO-BP模型
price forecast
quantum particle swarm optimization （QPSO） algorithm
QPSO-BP model
