Research And Development Of Context-Aware Recommender System In The Large Shopping Mall

With the development of RFID (Radio Frequency Identification) andother sensing technologies, the ubiquitous IOT (Internet of Things) iscoming into reality. Among the various intelligent applications built on IOT,the attempt of context-aware recommender system in large shopping mall isa promising research field and significant practice. Recommender systemsanalyze the implicit and explicit users’ feedback, thus help to find the users’favourite entities and recommend them to individual users. The effects ofrecommender systems can be divided into two aspects. First, users will bemore loyal, because the system can help them quickly find their favourites.Second, recommener systems help to improve transaction conversion rateand increase the cross-selling abilty of sellers.In this paper, we analyzed the domestic and foreign research status, andalso the characteristics of the recommender system context-awareapplications in a large shopping mall, then propose the design andimprovements of a set of shopping mall oriented recommender algorithms or solutions. The main work of our research could be divided into three parts.First, we focused in the application scenario of large intelligent shoppingmall, and extracted both the implicit and explicit user feedback fromshopping records for analyzing users’ preference. We based on the traditionalsimilarity algorithms and comibine with the charactristics of the shoppingmall data, normalized the information like the number of the consumingtimes. By doing those, we could eliminate and weaken the influence of toopopular merchants and active users may bring to the algorithms. Second,when coping with “cold start” which meant insufficient records to utilize, wemade use of user demographic characteristics and decision-tree methods torecognize both high and low granularity of users’ preference. At last, whenconsidering integrating the contextual information into recommender model,we focused on time, extracted it from RFID data and put it into preferencecalculation, so the users’ preference is time-related and context-sensitive.The paper is organized as follows: firstly, we describe the researchbackground, including related concepts of IOT, Contex-Awareness andRecommender System, and then we analyze the research status ofContext-Aware Recommender System. The context fusion process can begrouped into three groups: the the context of the pre-filtering, post-filteringand context modeling; the recommendder algorithm system can be dividedinto: recommendation based on collaborative filtering, content-based andmixed. We propose and design universal contextual recommender systemarchitecture. At last, give the specific implementations of variouscomponents of context-aware recommender system.