The Research And Implementation Of A Hybrid Recommendation

As a result of the combination of traditional retailing and information consumption, online shopping conforms to the trend of upgrading to new consumption, and has become the normalized shopping mode for residents. At the same time, the information in the electronic business website is growing explosively. The advent of "mobile e-commerce era" indicates that the number of users and commodities in the future will continue to grow, commodity recommendation system is thus facing a more severe test. Most of the existing recommendation algorithms do not consider whether the price of the recommended commodity is in line with the purchasing power of the user and whether the discount meets the user's expectation. This paper argues that commodity price and discount are one of the important factors influencing the user's purchasing decisions. Therefore, they are the influential factors that the commodity recommendation algorithm should focus on.Therefore, this paper presents a hybrid recommendation algorithm based on user purchasing power and collaborative filtering. The method considers whether the price of the commodity to be recommended is within the range of the user's acceptance and in line with the purchasing power of the user; and whether the discount of the commodity to be recommended is in line with the expectation of the user and whether it can promote the sale of the commodity. The main contents are as follows:1). A method to calculate the influence coefficient of commodity price on purchasing power is proposed. For a given commodity, the method first calculates the price level factor based on its price and the average price of the type to which it belongs. The purchasing power factor of the user is then calculated for a given user based on the price level factor of the commodity in the user's purchase history. Finally, for a given commodity and the user, a quantitative price influence coefficient function which can quantify the matching degree between commodity price and user purchasing power level is proposed by compromising two factors mentioned above.2). A method to calculate the influence coefficient of commodity discount is proposed to quantify the attractiveness of commodity's discount to users. According to the theory of commodity discount, this method calculates the matching degree between commodity price and user's expectation from the view of consumer behavior.3). A hybrid recommendation algorithm which combines the influence coefficient of commodity price, commodity discount coefficient and collaborative filtering recommendation score is proposed. The method assumes that the price of the commodity to be recommended needs to satisfy the purchasing power level of the user to the maximum extent, and the discount of the commodity to be recommended needs to satisfy the expectated discount of the user to the maximum extent, and the commodity to be recommended should be similar to the commodity once purchased by the user; recommending commodity which is similar to the commodity the user likes and of which the price and discount accord with the user's expectation.4). This paper crawls the standard dataset from Amazon which contains information such as commodity price, category average price, commodity discount, user-commodity ratings and so on. In this dataset, five benchmark algorithms are implemented and compared with the method in this paper. The experimental results show that the algorithm proposed in this paper has not only improved the score prediction (MAE, RMSE) or the ranking index (NDCG). It also shows that the proposed algorithm considering the user's specific purchasing power is feasible, and can improve the recommendation success rate of the recommendation system.