基于Bayes网的时间序列预测

Time-Series Forecast Based on Bayes Network

时间序列预测问题在气象、天文、电力、医学、生物、经济、金融和计算机等各个领域有着广泛的应用。本文将Bayes网模型用于该领域,提出并建立了一套基于Bayes的时间序列预测模型——静态Bayes网预测模型,动态Bayes网预测模型和分类静态Bayes网预测模型。实验表明,这三个模型能更准确地描述用户在Web上的浏览特征,在预测准确率和存储复杂度方面都显著地优于传统的时间序列预测模型。

Time Series Forecast problem is widely used in weather, astronomy, electric power, medicine, biology, economy, finance, computer etc. Now we use Bayes Network in Time Series Forecast, put forward and established a set of based on Bayes time-series forecasting model -- static Bayes network forecasting model, dynamic Bayes network forecasting model and the classification static ]3ayes network prediction model. Experiments have shown that these three models can more accurately describe users browsing features on Web. They are significantly better than traditional time-series forecasting model in the forecast storage complexity and the accuracy rate.