Research And Realization Of Rangking Based On Multi-attributes For E-commerce Search Engine

With the development of internet technology, the scale of electronic commerce is in rapid growth, it needs specialized e-commerce oriented search engine to help users searching and screening commodities. E-commerce search engines would rank the search results, and the ranking result is directly related to the quality of user’s consumption experience. At present, the common manner of ranking the search results is to sort based on single commodity attribute (such as price, sales volume), which is monotonous and can not satisfy users’ increasing diversity and personalized demand. This paper would analyze the possibility of sorting search results based on multiple commodity attributes. The common method is to construct an aggregate function to convert the multi-attributes problem into single attribute problem. However, as all the users take the same aggregate function, the ranking result could not reflect user’s individual preferences for different commodity attributes; and it is lack of the measure for objective factor that is commodity itself’s quality on each attribute. Meanwhile, with the rapid growth of the number of users and commodities in e-commerce, the performace of the commodity sorting is also faced a lot of challenges.To deal with these problems, this paper studies sorting the search results personalized based on multiple commodity attributes. Firstly, we construct a user preference model, it could mine user’s preference for different commodity attributes by analyzing user’s shopping behavior on e-comerce website and make the ranking result to better meet user’s personalized demand. Secondly, we propose a Skyline ranking solution based on user preference. The solution constructs personalized weight vector based on user preference model, to achieve sorting the search results personalized based on multiple commodity attributes; It improves the influence of objective factor that is the commodity itself’s quality on each attribute when ranking the search results based on Skyline query mechanism, to increase the ranking rationality. And to compute the commodity Skycube in the Skyline query mechanism, we propose a better Skycube calculation method CSBSC; At last, we will compute the commodity Skycube on Hadoop platform because of its very large demand for computing power and storage space, and we have realized CSBSC’s MapReduce implementaion CSBSC-MR, to more effectively improve the computational efficiency of commodity Skycube.Based on the above research work, we construct a commodity ranking system SLine to achieve to personalized rank the search results based on multiple commodity attributes, and SLine would be deployed to the cloud computing center of Southeast University. Finally, we will evaluate CSBSC by synthetic data sets and test SLine by real data sets accessed from Jingdong mall.