Research/Blog
Recommender Systems in Artificial Intelligence
- February 20, 2020
- Posted by: CellStrat Editor
- Category: Artificial Intelligence Deep Learning Machine Learning Natural Language Processing
Recommender Systems demonstration was done by AI Lab member – Gurumoorthy Loganathan in CellStrat AI Conclave on 8th. Feb’20 in Bengaluru.
Abstract
Do you want to influence your users’ choices while they are on your site? Do you want to serve them what’s on their mind without letting them search? Then, add a good Recommender system to your digital tool kit.
Recommender systems are excellent Marketing tools that give insights based on user’s choice or interest. Recommendation engines are a set of powerful & intuitive AI tools and techniques which analyze huge volumes of product and user information, to provide relevant suggestions based on data-mining approaches. It processes data in 3 stages – Collection, Storage and Filtering. Finally, recommend the result set which matches user preferences.
Algorithm
Collaborative filtering: Predict user’s interest by collecting preferences
Content-based Recommender Systems: Content/ features of item and user profile
Hybrid Recommender Systems: Combines various Recommender systems to build a robust system
Context-aware recommender systems: Personalized Recommender systems based on present state of the user considering location, time, weather etc., we can build context aware recommendations as a three-dimensional problem recommendation = User x Item x Context
Model Based Recommender Systems: This involves building a model based on dataset of ratings. This helps build a scalable model with high performance compared to memory based recommender systems. Techniques like Matrix Factorization, KNN and Deep Learning models are used to build these systems
Implementation
- Recommender Systems are widely used across the industry. Amazon, Netflix and Google have very advanced AI based recommender systems.
- Refer Demo of a movie recommender system built using Singular value decomposition (SVD) and Collaborative filtering. This application takes users history as input and recommends a new movie that the user can watch.
View complete Poster here.