摘要
文中基于小波多尺度分析进行网络论坛话题热度趋势的预报。该方法主要是对由帖子的点击数(或回复数)所形成的原始时间序列进行小波分解与重构,得到一个低频信号和多个不同尺度的高频信号;对具有近似平稳特征的低频信号建立ARIMA预测模型;对变化较多的各高频信号分别建立神经网络预测模型;然后分别对各信号进行一步预测并组合预测结果,获得网络论坛话题热度的最终预测。实验表明:将本方法用于网络论坛话题的热度趋势预测,可得出良好的预测精度。
In this paper, a method was proposed, which combined wavelet multi- resolution analysis to forecast network forum topic hot trend. That was first, using wavelet multi - resolution to decompose and reconstruct the original time series formed from click numbers or reply numbers of forum post, to create one low frequency signal and several high frequency signals. Then, the approximate low frequency signal is predicted using ARIMA model and the high frequency signals are forecasted respectively using neural networks that have different parameters. After one - step - ahead prediction, the predicted results of these signals are combined into the final predicted result of network forum topic hot trend. This proposed method was tested to make better prediction accuracy for network forum topic hot trend.
出处
《计算机技术与发展》
2009年第4期76-79,共4页
Computer Technology and Development
基金
国家高技术研究发展计划(863计划)研究资助项目(2005AA147030)
