超市管理系统-毕业论文.pdf

毕业设计（论文） 题 目 超市管理系统 　 姓 名 学 号 专业班级 指导教师 分 院 信息科学与工程学院 完成日期 2013 年 5 月 28 日 摘 要 本文建立了一种基于粒子群（PSO）算法优化BP神经网络的改进股票预测模 型。该模型采用时间序列作为输入数据，以此来构造BP神经网络拓扑结构，利用 粒子群算法优化BP神经网络的权值和阈值。同时，考虑到由于股票市场受到多种 因素的影响，故股票价格形成的时间序列是非平稳的且具有强噪声的，本文用小 波降噪来处理样本数据，使得数据更加真实地反映股票趋势。 结果表明，经过小波降噪处理后的时间序列数据更加平稳；粒子群算法优化 后的BP神经网络预测模型能够有效弥补BP神经网络的不足，改善BP神经网络易陷 入局部最优的问题，在一定程度上达到了提高算法性能的目的。 关键词：小波降噪；BP神经网络；股票预测；PSO优化 I Abstract Based on the analysis and study of the artificial neural network (ANN), this paper proposes the improved stock forecasting model of BP neural network optimized on the foundation of Particle Swarm Optimization. This model uses time series as the input data to establish BP’s (Back Propagation) neural network topology, and uses Particle Swarm Optimization to optimize the its weight and threshold. In the meantime, taking the various factors that affect stock market into consideration, it can be found that the time series of stock price is unstable and strong-noisy. Wavelet de-noising is used to deal with sample data so that the data can more authentically reflect the stock tendency. The results show that the time series data after wavelet de-noising are more stable; BP neural network forecasting model after Particle Swarm Optimization can make up the weakness of BP neutral network and mend the problem that BP neural network can easily get stuck into local optimum. As a result, this can obviously improve arithmetic performance to certain extend. Keywords: Wavel