Analysis And Research Of Music Personality Recommendation Based On Mixed Mode

Nowadays,the Internet has become an inseparable part of our lives.With the development of network-related technologies,various applications have emerged.A large amount of user data information is transmitted through the network,and the rapid expansion of data and information eventually occurs.In the face of these blowout data,we did not care at the outset and we were overwhelmed.In fact,these data are a treasure trove of value and contain a lot of valuable information.Personalized recommendation algorithm came into being,which greatly reduced the user resource search overhead.The recommendation algorithm is mainly to search for the user's preference,and to filter out unnecessary data information through the user's interest preference so as to obtain the final desired result.Although the recommendation system to a large extent allows the user to find the expected information in a relatively short period of time,there are certain problems in the recommendation quality,recommendation efficiency,and scalability.This article mainly analyzes the mainstream content-based recommendation and collaborative collaborative filtering technology in the personalized recommendation system.It aims at the sparseness and slow-start problem that are commonly found in the general personalized recommendation system and starts to improve and adjust accordingly.Try to mix on the basis of the above two algorithms to form a hybrid recommendation algorithm.This article tries to use the music recommendation as a background.The traditional recommendation system is recommended for related interest preferences.However,there are deficiencies in the definition of the so-called user preference model,which is often based on the user's historical rating.This article bases user preferences on three parts: user background attributes,user activity,and user history ratings.Recommend data sparseness problems common to the system.We fill in relevant gaps through predictions.For the missing values of user ratings,the common mode method and average method are not used.Instead,more reasonable predictions of missing values are used to increase the accuracy rate.The user's rating comes from the user's actual rating,as well as the above-mentioned missing value prediction.Although the missing value prediction is more reasonable,it still cannot represent the user's true inner choice very accurately.Therefore,the user is treated differently when performing the calculation.Use different weight values.Based on the relatively mature K-means algorithm,the user is clustered offline,thereby reducing the recommendation time and improving the user's online recommendation experience.Newly added music,no relevant history rating of the user.The newly registered user has no corresponding activity.This article deals with the problem of cold start of the algorithm by handling the corresponding project attributes and user background attributes.The algorithm of this paper is verified through experiments.This paper tries to simulate the algorithm and test the performance of the algorithm with some music data in Ali Tianchi data.Firstly,based on the evaluation index such as accuracy,average absolute error,etc.,the similarity calculation method that is most suitable for this data set is selected.Then,based on the same indicators,the experimental method of multiple cross-validation was used to verify that the improved algorithm compared with the traditional collaborative filtering recommendation algorithm and the traditional content-based recommendation algorithm has higher accuracy and lower error rate.Users have a better user experience.