E-commerce Site Personalized Recommendations Research And Implementation

With the continuous rapid development of Chinese e-commerce, e-commerceknowledge of the whole society has entered a new level. Advantage of thecharacteristics of electronic commerce, also caused widespread concern in thegovernment, the healthy development of e-commerce, China’s economicdevelopment has been referred to the Government’s work agenda. The traditionalinformation service model can’t provide personalized information service for thedifferent Web users. In order to improve the competitiveness of the electricitybusiness, the Web user’s potential information needed to fully tapThis thesis mainly focuses on Web Mining in e-commerce site recommendationsystem application. A merge clustering algorithm to generate page clustering isstudied firstly. This algorithm can be provided to the recommendation system basedon a new page clustering generated by the user’s model, page structure, and contentschanges of the page. Then a page recommendation algorithm comes out. Duringrecommendation, the score of node is firstly computed except current nodes.Secondly a minimum threshold recommendation is set, then the cluster nodes areplaced in the stack in addition to the current node, and then breadth-first traversalmethods is carried out to build up recommended collection according to therecommended score more than the minimum recommended threshold nodes. Finally,the personalized recommendation system based on the above mentioned algorithm isdesigned and implemented.Under laboratory environment, the preliminary experimental results of therecommendation algorithm coverage and accuracy show that: the betterrecommendation results are the minimum threshold between0.2and0.3. Load therecommended system APACHE server does not delay significantly, which showsbetter overall performance of recommendation system.