Research And Implementation Of Agricultural Product Recommendation System Based On Time Effect

With the continuous expansion of the scale of e-commerce in agricultural products,agricultural products have seen rapid growth in both species and quantity.In the face of massive agricultural product information,consumers find it difficult to find the agricultural products they need.Therefore,it is urgent to introduce a recommendation system to solve the problem of "information overload" of agricultural products.The traditional recommendation method seldom takes into account the seasonal characteristics of agricultural products,resulting in unsatisfactory results.In this thesis,based on the full consideration of the impact of time on agricultural products,the changes in user interests and seasonal changes in agricultural products are combined into traditional collaborative filtering recommendations to improve the quality of agricultural products.The main research content is as follows:(1)The impact of reaction time on the recommendation of agricultural products from changes in user interests.First of all,it is discussed that the change of user interest is influenced by the short-term and long-term time.Therefore,users' long-term interests cannot be ignored when they are concerned about the short-term interests of users.Secondly,the user interest weighting function fu(u,i)is constructed by combining the user short-term interest function fs(u,i)and the user long-term interest function fl(u,i)by the scale factor ?.Finally,the user interest weighting function fu(u,i)is integrated into the traditional similarity calculation of the merchandise so as to improve the accuracy of the merchandise similarity.(2)The influence of reaction time on the recommendation of agricultural products from the change of commodity popularity.First of all,taking into account that agricultural products are different from other common commodities,they have distinctive seasonal features.Therefore,the recommendation for agricultural products should not only consider the timeliness of the goods,but also take into account the seasonality of the goods.Second,Commodity Popularity Function Pop(i,ti.)and Commodity Seasonal Function Season(i,s)are combined by a scale factor ? to construct a commodity popularity weighting function Y.Finally,the commodity popularity weighting function f(i)is integrated into the product scoring prediction calculation,thereby improving the accuracy of commodity rating prediction.(3)The impact of time on agricultural product recommendation is discussed based on changes in user's interests and commodity popularity,and a dynamic time weighting algorithm based on user's interest and commodity popularity,T-UICF,is proposed.By adjusting the values of the proportional factors ? and ?,the appropriate weights are determined,and the similarity and the scoring prediction calculation method are optimized,thereby improving the accuracy of the recommendation.(4)Using the agricultural product data collected by Jingdong Mall,the accuracy and reliability of the algorithm T-UICF were verified by predictive accuracy evaluation indicators.Comparing the score prediction results of the algorithm T-UICF with the results of the traditional recommendation algorithm score prediction.Experimental results show that the accuracy of the improved recommendation algorithm T-UICF is obviously higher than the traditional recommendation algorithm.(5)Jingdong Mall is used to design and implement a recommendation system for e-commerce in agricultural products.On the basis of improving the traditional recommendation algorithm,this article considers the particularity of agricultural products,increases the time effect on the recommendation system,makes the recommendation system from a static system into a dynamic system,and improves the accuracy and real-time of the recommendation system.Sex is of great significance in promoting the development of e-commerce in agricultural products and the construction of agricultural information.