Research On Commodity Recommendation Based On Massive Data And Business-circle Interest Model

Under the "Internet +" environment,market competition has become increasingly fierce,and retailers' marketing activities have gradually extended to consumer terminals.Terminal retailers have become a very important intermediate point between enterprises and consumers,which is directly related to enterprise development.At the same time,in the face of the massive data generated by 8 million retailers nationwide,how to tap potential value has become the key to achieving accurate marketing.This paper focuses on the problem that the number of enterprise terminal retailers is huge and widely distributed,and it is difficult to directly extract potential interest from retail data and make accurate recommendations.The location-based recommendation algorithm is studied,and finally the consumption capacity and regional economic development are introduced.As a unit of analysis,the representative business-circle proposes the commodity recommendation algorithm based on the business-circle.The method can alleviate the cold start problem to a certain extent,and transforms into a market-based recommendation and delivery model based on the business-circle to strengthen the pre-judgment and control of the market environment change.The research work of the thesis is mainly divided into the following points:1)Collection and processing of massive data.Through the terminal collection,market visits and system filing,the enterprise collects massive data,and uses the network crawler technology to obtain the POC point data in the Meituan network and the POI data in the Baidu map api.For mass retailer data issues,the Spark distributed processing framework was introduced for processing,data cleaning,integration and conversion were completed,and realize pre-processing of business-circle point of interest data and enterprise multi-source data.Integration.2)Establishment of a business point interest point model.In view of the instability of regional boundary classification in the existing business-circle evaluation model and the incomplete data of the business-circle in the map api,combined with the current effective regional boundary recognition method,the center point of the business-circle is taken as the center,and the radiation distance is the radius.The POI data is pre-partitioned to obtain the initial business-circle.Then,using the proposed partition-based DBSCAN improved algorithm to eliminate the noise points in the initial trading circle,using the kernel density estimation method to determine the Eps and MinPts parameter values in the initial trading circle,and delineating the local clustering of POI data.The area of the business-circle,and finally the establishment of the business-circle point of interest model.3)Business-circle Popularity Based Algorithm.Aiming at the problems of insufficient recommendation accuracy and cold start in the big data environment,this paper proposes a product recommendation based on the business-circle interest point model.By analyzing the characteristics of the retailer's order data,the business-circle interest point model is used to divide the retail household data into each business-circle to realize the conversion of point data to surface data.At the same time,taking into account the consumer's preference timeliness of the goods,combined with the time decay factor to calculate the commodity popularity weights in the business-circle,the recommendations based on the product popularity ranking in each business-circle are generated.4)Business-circle Similarity Based Collaborative Filtering.Aiming at the problems of lack of surprise and long tail effect in the Business-circle Popularity Based Algorithm,the Business-circle Similarity Based Collaborative Filtering is proposed.Considering the time decay factor,the retailer recommends the similar business-circle in the business-circle.Hot items in the middle.The experimental results show that the proposed algorithm is superior to the traditional recommendation algorithm in recommendation accuracy,and it alleviates the cold start problem,which has research significance.5)Implementation of a commodity recommendation system based on the business-circle interest point model.In order to realize the demand for accurate marketing of enterprises,the recommendation system integrates the interest point model of the business-circle,and proposes hot-spot recommendation based on the popularity of the business-circle,relevant recommendation based on the similarity of the business-circle,and new product placement.The application results show that the recommendation method and the new product delivery strategy proposed in this paper have good results,improve the overall sales volume of the enterprise and increase the sales profit.