Study On Cross-category Recommendation Algorithm With Consideration Of Time Interval And Price

With the rapid development of Internet technology and the increasing extent of the user participation,all kinds of information on the network gradually become rich and the amount of them is increasing rapidly.However,too much information in a specific application scenario is increasing the cost and difficulty of a user to obtain the real needed information,resulting in the problem of information overload.Recommender system,as a kind of techniques solving information overload problem effectively,has been widely studied.Various recommendation algorithms have been put forward,and widely used in different fields,especially the electronic commerce domain.An e-commerce site tends to have a lot of different kinds of products,and the number of products in each category is numerous.When a user does shopping on a website,usually only those products of specific categories that the user is interested in or needs were browsed.However,when recommending the items to users,most of the existing recommender systems usually choose to ignore the inter-purchase time intervals among different types,with the items purchased before only considered.Thus,the generated recommended items at different time points present high homogeneity,and item categories possess a high degree of overlap with the items purchased before,reducing the users’ satisfaction of the recommendation system.On the other hand,a lot of researches in the field of purchase decision have shown that price is one of the main factors influencing a user’s purchase decision for a specific product.A users tends to hold different price preferences for different kinds of products,while the existing recommender systems rarely consider the price factor;even in the utility-based recommender systems,price factor is only viewed as a common product attribute,not fully using the users’ price preferences for different types of products.Considering the shortcomings of existing recommender systems stated above,this paper proposes a cross-category recommendation algorithm.First,by considering the multi-category inter-purchase time intervals,based on sequential pattern mining technology,we model the drift of a user’s interest for different types of products for a particular user.Secondly,based on the fuzzy set theory,we develop a method to model the price preferences of different types of products for each user.At last,by combining the methods mentioned above together that consider the time intervals and price,the proposed algorithm generates a unique recommendation list at a particular time point.Based on purchase records from two real websites,we do some experiments to demonstrate the effectiveness of the two factors in improving the recommendation accuracy as well as the proposed algorithm in combining the two factors.Experimental results show that the price is really an important factor affecting the users’ purchase decisions,and can significantly improve the recommendation accuracy.However,the time interval factor,with regard to the improvement of recommendation accuracy,shows different performance in different product fields.In the non-frequently purchased product area,such as electronics,the time interval factor can effectively improve the accuracy of recommendation.Nevertheless,in the field of products and services that are purchased regularly,the effect the time interval factor plays in improving the recommendation accuracy is not significant.In addition,no matter in which product areas,our proposed recommendation algorithm can significantly improve the accuracy of recommendation compared with the traditional recommendation algorithms.