Today's headline is a recommendation engine product based on data mining developed by Beijing ByteDance Technology Co., Ltd., which can recommend information for users and provide services to connect people with information. Created by Zhang Yiming in March 20 12, the first version was released in August 20 12.
Product characteristics
Based on personalized recommendation engine technology, personalized recommendation is made according to each user's interest, location and other dimensions. The recommended content includes not only news in a narrow sense, but also information such as music, movies, games and shopping.
According to their social behavior, reading behavior, geographical location, occupation, age, etc. Through social behavior analysis, the user's interest is calculated in 5 seconds; Through user behavior analysis, the user model is updated within 10 second after each user action.
Extract dozens to hundreds of high-dimensional features from each piece of information, and perform dimensionality reduction, similarity calculation, clustering and other calculations to remove duplicate information; Information is processed by machine classification, abstract extraction, LDA topic analysis and information quality identification.
Recommend according to the matching degree of human body characteristics, environmental characteristics and article characteristics.