muRecommend™
A Meta-software that integrates cross-industry experience in the recommendation, personalization and
CRM landscape
The combined strength of individual modules in the framework multiplies effectiveness at scale

Contextual Synergy
-
Recommendations based on given features and Identifying latent features for users and products
-
Ensemble of many recommendation algorithms that handles both variety and similarity

Scalability
-
Low dependency on environment / architecture
-
Fits with any analytical process flow seamlessly
-
Supports big data through python server

Data Driven Class Selection
-
Input data driven algorithm selection
-
Distribution fitting, variance and correlation checks for determining quality of variables
-
Rule based decision tree will determine the algorithms that need to be run based on the above checks

Speed vs Accuracy
-
Full blown factorization algorithms that focus on accuracy
-
Hybrid models with hashing that focus on speed
-
Multiple algorithms enable the user to pick between quasi real time algorithms or accuracy focused algorithms
The framework accounts for three dimensions – Data, Context and Algorithms
-
Content based recommendation
-
Collaborative filtering
-
Hybrid model
-
Association rules

-
“People who are similar have bought this item”
-
“People who have bought this item have also bought the following”
-
Product bundling
-
Transactional
-
Product ranking (Implicit/Explicit)
-
Click stream data
-
Cached data
-
User/Item features data