Feature extraction of user interests based on hidden semi-Markov model
Min Zhang · Jisuanji gongcheng yu sheji · 2011
For the extraction of users interests behavior feature,a method of user interests feature extraction based on hidden semi-Markov model is proposed,which can control the user's browsing behavior through by using the probability of state stay time,and combine the hidden state of described interest feature with the relevance of time tightly.According to the characteristic that hidden semi-Markov model can generate multiple sequences of observations,the text information is divided into several sub-regions,so that the feature of each sun-regions and the sequence of observations can correspond one to another.Experiments show that using HSMM has higher accuracy and recall than the HMM method for feature extraction.