Research And Implementation Of Page Recommendation Model Based On Web Usage Mining And Associate Rule

Web usage mining is the hot research issue of web mining. It plays an important role in the discovery of visitors' browsing behaviors, the improvement of Web system and the implementation of personalized service of web system. Both web usage mining and personalized recommendation technlogy are used for the personalized service based on web usage mining(PSWUM). The normal PSWUM is the page recommendation based on web usage mining(PRWUM), and the normal recommendation method of the PRWUM is associate rule.In the paper, the key algorithms of page recommendation model based on web usage mining and associate rule (PRWAR) were studied in detail. The PRWAR consists of three parts of web usage mining which are data preprocessing, pattern discovery and pattern analysis. It includes offline part and online part. The process of the offline part is preprocessing the web log and discovering frequent access patterns. The process of the online part is acquiring users' access sequences, obtaining associate rules based on frequent access patterns and attaining the back items of associate rules as the recommended page by selecting associate rules. The main work of the thesis is showed as following:1. A algorithm(SDHP) which is for the use of discovering frequent access patterns is proposed. It uses the technology which is used for linking the transactions in AprioriAll and utilizes hash method and pruning method which are used for processing the the transactions in DHP. The difference is that DHP processes no order data, but SDHP and AprioriAll process the order data.2. A method(TPM) which is used for page recommendation based on transaction Partition. It recommends pages based on transaction Partition after processing users' access sequences by MFP algorithm, but the method(SWM) which is used for page recommendation based on Sliding Window recommends pages after acquiring users' access child sequences by Sliding Window.3. A page recommendation model based on transaction(PRBT) is proposed. It is based on SDHP and TPM, but the page recommendation model based on sliding window(PRBSW) is based on AprioriAll and SWM. Both PRBT and PRBSW belong to PRWAR.4. Designing and developing a page recommendation prototype system(PRPS) based on .NET platform. This system throughs three parts of web usage mining and consists of six function modules. It has realized PRBT and PRBSW and their relevant algorithm by C#.The experimental results in the PRPS show that SDHP not only reduces candidate set generation, but also compresses the size of transaction database, so it comparing with AprioriAll can reduce the computational cost significantly; and TPM comparing with SWM can acquire especially all-round and accurate recommended results. In a word, synthetical function of the PRBT is better than the PRBSW.