Research On Indoor Personalized Recommendation Based On Location Perception

With the development of urbanization,people spend more and more time in indoor spaces such as large shopping malls,supermarkets and office spaces.Understanding people's indoor activity patterns and interests and preferences,and providing personalized recommendation services not only meet the individual needs of users and facilitate their daily lives;at the same time,it also helps optimize the layout of indoor facilities and scientific management,and improve service quality.Therefore,the research on indoor personalized recommendation has important practical significance.Different from online shopping,in these indoor spaces,due to the lack of means and methods to obtain objective feedback information from customers to a certain store or facility,it is difficult to accurately analyze the user's behavior preferences,consumption habits,and personalized needs,and thus cannot provide Personalized recommendation service.In recent years,with the development of indoor high-precision positioning technology,it is possible to easily perceive the user's indoor location information.Through the user's location perception,combined with the indoor environment,mining the trajectory data of their indoor movement,it can objectively analyze the user's spatio-temporal behavior,obtain personalized needs such as shopping habits,interest preferences,etc.,so that user interests are no longer limited to questionnaire surveys and mobile apps The highly subjective feedback such as comments lays the data foundation for the indoor personalized recommendation,and also makes the personalized recommendation based on location perception a hot spot in the current GIS research.Based on the user trajectory data in a large shopping mall,the social media data of indoor shops,combined with the physical characteristics of the indoor space,this paper studies the indoor personalized recommendation method from two levels of customers and merchants to provide convenience for customers' shopping activities;at the same time,it also allows businesses to understand Customers who are interested in our store can further optimize products and improve services based on these customer groups.The main work of this paper has the following three points:(1)Research on indoor POI recommendation algorithm fusing location perception and gravity model.First,according to the characteristics of indoor trajectory data,expand the POI attributes to improve the gravity model to obtain the attractiveness between different POIs;propose the use of structural factors to optimize the weighting ratio of attractiveness and cosine similarity,and optimize the calculation method of similarity between indoor POIs;Secondly,mine the length of time the user stays on different POIs in the trajectory,and convert it into user interest;finally,the recommendation task is calculated according to the principle of collaborative filtering algorithm,and the optimal recommendation model is obtained through multiple iterations.The entire recommendation task uses the Hadoop big data platform and Spark.The distributed computing framework realizes point-of-interest recommendation in complex indoor situations.(2)Research on indoor POI recommendation algorithm based on knowledge graph.First,design a personalized indoor POI knowledge graph specifically for recommendation based on the characteristics of indoor POI distance,floor,type,and per capita consumption;then improve the Ripple Net water wave network model by adding a gravitational neural layer to optimize user interest representation;finally;The attention mechanism is introduced to calculate the weight value between candidate POIs and user preferences,and accurately predict the probability of users accessing candidate POIs.The algorithm improves the network structure of Ripple Net,makes it suitable for indoor POI recommendation,and provides a reference for indoor personalized recommendation based on knowledge graph.(3)Research on indoor customer group recommendation algorithm based on tag similarity.First,user comment data is segmented and cleaned,and name matching is performed with indoor POI to obtain user tag information;then TF-IDF algorithm is used to calculate tag similarity,and the concept of user tag distance is proposed for indoor customer group extraction;finally based on tag similarity Design the indoor customer group recommendation algorithm,select the customer group with the highest recommendation index for recommendation,improve the indoor personalized recommendation system,and help shops carry out targeted publicity and sales.The paper has 22 pictures,15 tables,and 81 references.