Design of personalized recommendation system based on optimized fuzzy mean algorithm

Intelligent algorithm is an effective technical means for the personalized recommendation problem of big data. The new retail industry has accumulated a large amount of user data, and how to use these data efficiently has become the key to improve the quality of service. The traditional single-dimensional analysis method cannot accurately define the user's portrait, it is necessary to portray the user from multiple levels and dimensions in order to achieve a more accurate marketing strategy. In order to build a more accurate new retail user portrait and improve the accuracy of personalized recommend, a user portrait model containing four dimensions: commercial social networking, retail marketing, domain and basic attributes is proposed. In view of the of fuzzy mean algorithm in dealing with unbalanced data samples, the density function and sample influence value are introduced to optimize the iterative process of clustering center and improve the stability and accuracy of clustering. The of the improved fuzzy mean algorithm is improved from 0.77 to 0.95 in accuracy, the recall rate remains stable, and the average absolute error is reduced from 0.88 to 0.78. It is shown that the improved fuzzy mean algorithm can effectively improve the construction quality of the new retail user portrait and the accuracy of the personalized recommend system, and also prove that the prediction accuracy of the designed recommend system is significantly improved.