Research On E-commerce Nearest Neighbor Collaborative Filtering Recommendation Algorithm Using Data Field Clustering

With the rapid development of Internet technology,the level of product information on e-commerce platforms has increased exponentially.In this context,product recommendation technology has emerged as an active information filtering method.Among the many recommendation techniques,the clustering-based neighbor collaborative filtering recommendation algorithm has attracted wide focus from scientific research workers at home and abroad because of its advantages of simplicity and efficiency.However,with the continuous expansion of e-commerce scale and the continuous reduction of data value density,the impact of data sparseness on the performance of the existing clustering-based neighbor collaborative filtering recommendation algorithm has become more and more significant,and it has been unable to meet the current background of commodity recommendation for recommendation quality.To this end,we proposed an improved data field clustering-based Neighbor Collaborative Filtering from an interdisciplinary perspective,in order to achieve a win-win situation for users and businesses.Firstly,we comprehensively compare 11 mainstream e-commerce platforms at home and abroad,sums up 17 heterogeneous preference data collection feature points that can be used for clustering nearest neighbor collaborative filtering recommendation algorithm optimization,and explained the selection basis of some feature points.And then,a data processing program that meets the requirements of subsequent recommendation algorithm optimization is proposed.The program includes data integration,content security library construction and semantic data extraction.Secondly,in order to solve the problem of low performance of the recommendation algorithm caused by the high dimension of the rating matrix and the distortion of the rating matrix,we propose a processing method for the rating matrix from two aspects of commodity reduction and rating correction and processes the original rating matrix.Thirdly,in order to solve or alleviate the low efficiency of the existing clustering algorithm when users are grouped,we combine the traditional mean clustering algorithm with the formal cognitive model data field and proposes an improved mean clustering algorithm.And then,the users in the revised user-commodity rating matrix are clustered and grouped by means of the virtual active field.Fourthly,in order to solve or alleviate the problem of distortion of similarity calculation results caused by insufficient historical scoring data,we propose an improved Pearson correlation coefficient which combines user semantic similarity and scoring similarity.Then,we calculate the similarity between users and other users in the same cluster based on the clustering results and fills in the missing scores in the rating matrix based on the historical rating data of the nearest neighborhood users,so as to make personalized product recommendation lists for users.Fifthly,10 groups of control experiments are planned and constructed in this paper.The first 6 groups of experiments are used to determine the optimal value of algorithm parameters,and the last 4 groups of experiments are used to verify the algorithm optimization idea and algorithm performance.The experimental results show that the optimization idea of the improved data field clustering-based neighbor collaborative filtering recommendation algorithm proposed in this paper is correct and effective.In particular,compared with the results of each test set of the state-of-the-art recommendation algorithm,the precision of the optimization algorithm in this paper is improved by about 18%on average,the recall rate is increased by about 12%on average,the proportion of prohibited goods is reduced by about 11.8%and diversity decreased by about 15.3%.In addition,the clustering results of the UCI benchmark datasets prove that the improved mean clustering algorithm proposed in this paper is more stable and efficient than the traditional mean clustering algorithm and has good scene migration capabilities.