Research Of Path-Recommendation Algorithm Based On User Preference

Personalized recommendation is a process that recommends to the users the informationand products in which they are most interested according to their interests and purchase habits.As the scope of electronic commerce gradually enlarges and the number and type of productsincrease rapidly, the customers need to spend a lot of time to find out their target products.Doubtless, the number of customers is doomed to decrease, because there are too manyirrelevant information and products for glance during the process, which immerses thecustomers in information overload. To solve these problems, personalized recommendationemerges as the times require. Personalized recommendation systems are one kind of advancedbusiness intelligence platforms based on magnanimity data mining, which help electroniccommerce websites offer totally personalized decision making support and informationservice to the customers for shopping. Recommendation systems in shopping websitesrecommend products to the customers, automatically finish the process of personally selectingproducts, satisfy personalized requirement of customers, recommend popular products basedon websites, and speculate possible purchasing behaviour in the future with respect to thecities the customers live in, the purchasing behaviour and records of customers in the past.Path search is a process that searches for a path with lowest cost between the source and thedestination. The customers give out some constraints or preferences and recommendationsystems find out the optimal path based on these constraints. There are many paths betweenthe source and the destination, which makes it impossible for customers to evaluate all thesepaths by themselves. Therefore, recommendation systems will help customers find out themost appropriate paths.The method of this article calculates the path cost from source to destination for usersaccording to the shortest path algorithm, and then does skyline computation combining thepath cost with static attributes of available destinations. The static attributes of data pointsstay unchanged for different kinds of queries, so this article proposes a strategy that firstlyoffline calculates the dominant relationships and stores them with an efficient data structure,and secondly online calculates the path cost between the query point and each data point, and then does skyline computation combining the path cost with the dominant relationships forthe static attribute space. One method to store the dominant relationships for the staticattribute space is to allocate a dominant list for each data point to store its dominators, whichleads to the sequentially scanning algorithm based on dominant list by this article. However,this method needs very much memory space, because it stores all the dominant pairs.Therefore, this article proposes another method, called memory search algorithm with pruningstrategy based on dominant graph, which effectively reduces memory occupation of dominantrelationships and performs efficiently.