Research On Indoor Retail Location Based On Machine Learning And Wi-Fi Radio Signal

With the popularity of intelligent terminal and the rapid advancement of Internet technology,the modern lifestyle has changed a lot.Location Based Service(LBS)is a basic function that is necessary for today's intelligent terminals.The implementation of many software functions depends on the ability to accurately locate the specific location of the target user,such as weather forecast,recommendation for tourist attractions,car navigation and other software.The positioning service can be mainly divided into outdoor positioning and indoor positioning.In outdoor positioning,the traditional positioning technology can satisfy the basic positioning requirements.But in indoor positioning field,location results cannot be acquired by directly using the outdoor positioning technology.At present,the offline market is greatly affected by online shopping.In order to maintain competitiveness,shopping malls need to develop unique advantages of offline consumption,such as experiential shopping and personalized consumption.First of all,how to accurately locate the customer's location in the mall is an urgent problem to be solved.Based on the consumption records and the information about the activities of customers,this paper quantifies the relationship between the specific store and the real?time information collected by the web server in the mall,and proposed a set of indoor shop location algorithm.The works include the following aspects:(1)Analyze the traditional technology used in indoor positioning field,and select the benchmark method for indoor shop location problem.(2)Design a new shop location algorithm.Considering the application scenarios and complementarity of various methods,and make full use of the information recorded by the web server(3)Conduct experiment.Solving indoor store positioning prediction problem with a single machine learning algorithm,and carry out experiment comparison with the traditional k-NN fingerprint positioning method.(4)Design the k-NN correction framework for multi-classification model and the stacking fusion process for the two-category model,which can improve the accuracy of indoor retail location.