基于改进的隐马尔可夫模型的网页预取

Web Pre-Fetching Based on Improved Hidden Markov Model

为提高基于隐马尔可夫模型(HMM:Hidden Markov Model)的网页预取精度,对经典隐马尔可夫模型的两个前提假设进行了扩展,推导出新模型中计算观测序列概率的公式。由此构建出可用于网页预取的高阶隐马尔可夫模型,同时为降低高阶隐马尔可夫模型的空间复杂度,给出了构建树状状态空间存储访问序列的算法。介绍了将改进的隐马尔可夫模型应用于网页预取的具体方法,通过对比实验证实该方法的预取准确度提高了7%。

In order to improve the accuracy of the predictions, we expand two premises of the traditional hidden markov model, and calculate new formulas of the observing sequences＇probabilities in new high-order modal used for web pre-fetching. To reduce the space complexity of the high-order model, an algorithm used to store the visited sequences by tree-like structure is given. Finally the method of using expansive hidden markov model in web pre-fetching is shown. The experiment confirms that the accuracy of new method is 7% higher than old one.