Study On Time Weight Hybrid Recommendation Algorithm

With the development information age,computer network technology is widely used in the electronic commerce systems.Buying your favorite products online has become an integral part of contemporary life.However,the large and complex information in the system make it difficult for users to choose the goods and services quickly and satisfactorily,so the proposed algorithm is developed.In the current popular recommendation algorithm,the collaborative filtering algorithm gradually shows advantages in a large number of practices,but traditional collaborative filtering algorithms do not take into account the time effect of user preferences,the time information of user behavior is not recommended as a reference.In real life,however,user preferences are likely to change over time.So studying the time effect of users' interest is of great significance to improve the ability of recommendation algorithm to predict the current interest of users.In this paper,inspiring by the collaborative filtering algorithm,first of all,the traditional collaborative filtering algorithm based on the users of the nearest neighbor recommendation algorithm is combined with time decay factor,then we construct the time decay function,and using its reference value to give different weights according to the rating data of different period in the process of recommended,Then,aiming at the generally encountered problem of sparseness in score data matrix data of the current recommendation algorithm,this paper proposes a solution,it is a hybrid recommendation algorithm that combines the recommendation algorithm and the collaborative filtering algorithm based on time weight.The advantages of the hybrid strategy is making a success to avoid the disadvantages of the two algorithms on the recommendation accuracy,and it greatly improves the accuracy of the recommended to some extent,making the algorithm applied more widely.In this paper,the time weight collaborative filtering recommendation algorithm is calculated based on the cosine similarity and the average score deviation formula when calculating the user's prediction score for the unknown project.using the exponential function as the time decay function.Finally,using Movie Lens data set as a source of data for the algorithm simulation,this paper designs three experiments to implement the movie recommendation for the users.This paper illustrates that the proposed algorithm in this paper has good viability and effectiveness through comparing the accuracy,MAE and confidence indicators of collaborative filtering algorithm based on user's nearest neighbor,collaborative filtering recommendation algorithm based on time weight and the hybrid recommendation algorithm.