SRV function-based collaborative filtering recommendation algorithm

Abstract

The present invention discloses a SRV function-based collaborative filtering recommendation algorithm. The method comprises the following steps of classifying all perceived attributes of users according to the data of all users: firstly classifying all users by specifically adopting the K-means clustering method; extracting category features to obtain the representative elements of corresponding categories, and then obtaining a corresponding spider diagram; firstly calculating the distance of each representative element of a newly added user and judging whether the user belongs to one category or not; if yes, adding the user to the above category; if not, turning to the next step; rotating each newly added user till an optimal rotation angle is found out, wherein each rotating angle can be determined according to the number of categories. The method is high in efficiency and high in precision.

Description

technical field [0001] The invention belongs to the technical field of the Internet, and in particular relates to a collaborative filtering recommendation algorithm based on an SRV function. Background technique [0002] In the Internet era, the e-commerce-based sales model has opened up a new shopping model, which allows people to freely choose their favorite items online, but what follows is a wealth of products, various promotional advertisements, whether it is a web page The pop-up window is still the red dot recommendation of WeChat, or there are so many recommended items in the email. Basically, it is impossible to chat online normally and choose items on the website. It is often interrupted many times. The main reason is online shopping. However, as a user, we cannot browse and view the information one by one to judge the pros and cons, and make a comparison before purchasing, because the general consumer’s consumption tendency is not very clear and is easily affected...

AI Technical Summary

Problems solved by technology

[0004] The first type is a user-based recommendation system called collaborative filtering method. Its main method is to recommend products to users based on the user's close surroundings or the purchasing habits of the user's circle of friends. Friends have a certain degree of shopping similarity or the user’s friend recommends the usage of the product to the user, so it is a very useful method. However, with the continuous expansion of the user range, especially the WeChat circle of friends is constantly being used by WeChat business. The occupation of such merchants leads to a gradual decrease in user satisfaction, especially when there are Wechat merchants in the user's circle of friends, which will seriously damage the user's trust in the product, and with the growth of the post-90s generation, gradually They are called the main force of consumption. They are the real digital generation. It can be said that they cannot live without mobile phones. But at the same time, this group of people is different from the previous generation. It is the so-called non-mainstream, so no matter what it is, they like to have their own unique side and are unwilling to follow the crowd, so this is why we will see a phenomenon called long tail in the sales of many large websites, that is, In addition to the best-selling products in the sales volume of products, although the sales volume of some products is not as much as the best-selling products, the sales of these products can be as much as the best-selling products, and sometimes even exceed the sales volume of the most popular products. Therefore, the sales recommendation for long-tail products is the focus of the website, but how to make the recommendation accurate in the existing situation? The method of collaborative filtering is obviously not suitable, because it ignores the needs of each user. Individual characteristics, so it is necessary to carry out refined operations on the basis of the original

[0005] The second type is item-based recommendation system. Its main method is to recommend related products or the same product of different brands to the user based on the user’s past search records, or peripheral items related to the searched product. It mainly focuses on The point is that the product itself is not related to the user's usage habits. Most websites now use recommendation algorithms that collect users' browsing history through the background. Of course, for users, some can help users find the products they need in time. Either the price is more appropriate, or peripheral products that the user has not considered. These are useful for the user's real-time search, but the user's search is random, and may be searched on this website or in other sites. Therefore, it cannot be fixed for a long time. Therefore, for a website, if it can grasp the user's consumption habits and give users a very good shopping experience, it is a good recommendation system to retain users for a long time.

[0006] Based on the above two types of recommendation systems, we can see that both the collaborative filtering method and the item-based recommendation system have shortcomings. The former ignores the user's personality characteristics, which is especially obvious in the current user group, while the latter cannot be retained for a long time. Users only exist randomness and contingency, so a new algorithm must be redesigned to improve the recommendation system to adapt to new changes and situations, and the recommendation system algorithm based on SRV function is to recombine the advantages of the two Designed to fit while eliminating some of the shortcomings of the two

Images