Futubull Content Recommendation Algorithm

The algorithm extracts users’ preference features by comprehensively analyzing and statistically processing your in-app behavior (including self-selected stocks, searches, views, taps, follows, comments, shares, collections, "not interested" selections, and blacklists), personal settings, IP address, and device information (including device model and environment). Based on content filtering, retrieval, and ranking models, content is scored and sorted to recommend high-quality content that you might be interested in, thereby improving your content experience.

1. Based on users’ interactions while browsing market data and content, and by extracting tags from that content, the algorithm statistically determines the market and content tags you are interested in, building a weighted preference profile for users.

2. By tracking the stocks and content users follow and browse over the long and short term, the algorithm builds diverse retrieval strategies and creates an index of content to retrieve items users may find interesting.

3. The retrieved content is then filtered based on its review status, blacklisted authors, and content marked as "not interested," as well as users’ browsing history. This process retains high-quality and relevant content.

4. The retrieved content is scored and ranked using mainstream linear and deep learning ranking models. The content users are most interested in receives the highest score, making it more likely to be recommended to users.

5. After ranking, a diversification strategy is applied across multiple dimensions such as author, topic, and content type. This ensures content diversity while still recommending the most relevant content to users.

6. We will continuously adjust and optimize the recommendation results by using your interactive behavior during app usage to provide offline or real-time feedback on users and content features and the model itself.

Futubull App

When the "Settings - Privacy - Personalized Content" switch is turned on, the algorithm is applied in the following scenarios:

The algorithm is designed to help users get valuable information more conveniently and efficiently, to help us better understand users’ personalized content needs, and to improve users’ overall content experience.