Research And Implementation Of Personalized Recommendation Algorithm Based On User Location Distribution

Collaborative filtering is one of the most widely used personalized recommendation algorithms since the development of e-commerce,not only saves a lot of time and energy for users to choose items that meet their preferences,but also greatly improves the economic benefits of e-commerce platforms and sellers.However,on the one hand,there are still many problems to be solved in the personalized recommendation system,such as the sparseness of data,the cold start of new users or new products,and the accuracy and diversity of recommendations.Data sparseness,cold start for new users,the low of accuracy and diversity in recommendations,etc.On the other hand,e-commerce websites have accumulated a large amount of time and space data on user behavior,but algorithms such as collaborative filtering have not fully considered the impact of the information,making the accuracy of recommendation results unable to be further improved.This paper considers the influence of the characteristics of the user's location distribution on the similarity of the items,and proposes an improved collaborative filtering algorithm.With the help of experiments,it shows that the algorithm can generate more accurate recommendation lists for users,and the variety of recommended items also has excellent performance.Our research not only proves that the user distribution characteristics of products are an important factor for measuring similarity of products in collaborative filtering algorithms,but also provides valuable information on how e-commerce platforms can effectively use user location information to improve the accuracy of recommendation algorithms.The main research work is as follows:(1)We analyzed based on the of the distribution of users of real products,by comparing the differences in the distribution of users of different products,summarizing the internal laws and tentatively exploring the possible mechanism of this difference.In different geographical environments,users have different behaviors and preferences.This difference in preferences will affect the distribution of users in different products.Reasonable use of this rule will helps us to understand the mechanism of user preference differences and provide effective ideas of high accuracy recommendation algorithms(2)In order to study the impact of the differences in the distribution of product locations on the accuracy of the recommendation algorithm,this paper proposes an index of distance distribution coefficient to describe the distribution of user locations,and uses the coefficient as a special attribute to characterize items.After the distance similarity is calculated by the distance distribution coefficient,we can calculate the mixed similarity by using a balance factor to fuse the scoring information.Based on the mixed similarity,we use the K-Means algorithm to divide the user into different areas according to the user's position,then use the collaborative filtering algorithm to recommend in each area,and finally all the recommended lists need to be obtained through a weight Weighted to get the final recommendation list.With the help of experimental comparison on multiple real data sets shows that the algorithm proposed in this paper can effectively improve the accuracy and diversity of recommendations.(3)By supplementing and improving the experimental data set and integrating the user's location information,we designed a movie recommendation system and applied our algorithm to the system.The system designed has several major features.First,the system can use the geographical location of new users to generate a recommendation list,which alleviates the problem of cold start for new users.In addition,the recommendation algorithm used by the system can fuse the spatial distribution characteristics of items and User's location information,which enables more accurate recommendations.These characteristics indicate that the movie recommendation system can effectively enhance the user experience,while also having a strong promotion and application value.