In order to improve the user experience, the user's favorite content is found from a large number of candidate content and given priority to the user.

For the platform itself, it can improve the user click conversion rate, increase the user viewing time (broadcast completion rate), and improve the user member recharge conversion rate.

invest

Starting from the user-scene-behavior of watching video, the core function of video recommendation scenario is that users use devices to enter the video platform to watch video, so the input mainly includes four blocks, user basic characteristics, scene characteristics, user behavior characteristics and video characteristics.

calculate

3. 1 tourist user recommendation-cold start recommendation strategy

The basic calculation logic is:

Recommend to users based on video characteristics, calculate the correlation with the currently watched video, and form a candidate content pool with high correlation; Sort by video score, video popularity and release time.

3.2 Using historical data to recommend users-personalized recommendation strategy

The basic calculation logic is:

Step 1, according to the basic characteristics of users, match the historical behavior characteristics of users with the current video characteristics, and recall the video to form a high-quality candidate content pool;

Step 2, making secondary selection according to the current login behavior characteristics of the user and the current scene characteristics;

The third step is to consider the user's payment, membership level, video rating, video popularity, release time, etc. If the user is a vip, recommend 4 VIP+ 1 free+1 charges; If the user is not a vip, recommend two VIPs+four for free.

(1) video recall

(2) Secondary selection

(3) Sorting recommendation

output

(1) Settings 1: Ordinary tourists.

The output results are as follows: Wukong Tale (free), Wukong Tale edited by netizens (free), high-rated movies with the same content (vip), high-rated movies with the same theme (vip), the latest movies created by the same author (free) and the same type of Xian Xia movies (free).

(2) Setting 2: Xiaoli, a 22-year-old female student, likes idolization and vip, and likes to talk about dramas at night and weekends.

The output results are as follows: Wukong's story (free), the latest movie created by the same author (charged), the latest movie created by the same author (vip), the latest movie with the same content (vip), the movie with high score on the same theme (vip) and the movie of the same genre (vip).