In this regard, it was decided to create a comprehensive recommendation system that would take into account not only the user's behavioral activity on the social network, but also the semantic content of the content with which he interacted.
Key Challenges
- Develop a concept for a recommendation system with a step-by-step implementation.
- Implement an ML module that analyzes content and generates personalized recommendations in a specialized section and search.
- Organize hypothesis testing, algorithm setup, and ongoing performance monitoring.
- Collect detailed metrics before and after implementation for an objective assessment of the result.
The project was carried out in two stages.
At the first stage, the ICL Services team built recommendations based on user activity.
Then, an individual feed was formed with a content balance: 80% - new publications, 10% - potentially interesting content, 10% - random sampling to expand the user's horizons. Here, we excluded closed accounts and outdated publications.
The second stage included:
· Connecting content analysis: text, photos and videos.
· Building publication vectors and clustering by interests.A more complex logic of recommendations was also built taking into account the subject matter and types of content, and not just the connections between authors.
At each stage, testing was carried out, algorithms were adjusted and the quality of issuance was improved based on real user behavior metrics.
Products and technologies
- Agile / Kanban
- User activity and content analysis (vectorization)
- Microservice architecture
- Azure DevOps
- Confluence
- MS Teams, Zoom
- Machine Learning
Results
- The social network feed has become customized – individual for each user.
- Increased engagement: users stay in the application longer and interact with the content more actively.
- An analytical base with metrics of the quality of recommendations for further optimization has been created.
- The system architecture is ready for scaling and adding new algorithms.
YOU CAN ASK ME ANY QUESTIONS YOU HAVE AND GET CLOSE CONSULTATION ON OUR SERVICES.
expert in Application and database migration service