Watch brand ranking and price: Four important features of website SEO optimization model recommendation

The first feature is the popularity feature, which covers the popularity of various aspects, such as the popularity of the website theme, the popularity of the post content classification, and the popularity of the keywords in the website. The popularity information of those contents will have some unexpected effects on some large websites.

The second feature is the relevance feature. Relevance is to match the content attributes of the article with the needs of users. The matching here is divided into explicit matching and implicit matching. Matching of some article categories, matching of keywords in web pages, matching of page topics, etc. all belong to explicit matching. The matching in some FM models is implicit matching. We can see this from the vector distance between the content and the user.

The third feature is what we often call the collaborative feature, which to a certain extent solves the problem of the algorithm becoming more and more difficult. Collaborative features do not look at the content viewed by users, but rather determine whether there are similarities between different users through analysis of each user, such as similar hobbies, similar topics, or similar clicks. From these aspects, we can expand the search capabilities of the model.

The fourth feature is the environmental feature, which includes the publication time and place of the article, etc. This is both a bias feature and a matching feature.

In terms of model training, most websites recommend that their products can be trained in real time. Real-time practice here not only saves resources but also makes users respond to the product very quickly. You must know that in our website, current website processing requires sample data. This includes click-through rate, number of impressions, sharing level, action type, etc.