CodeGen’s Personalization Engine is a combination of several Personalization and Recommendation tools, with B2B and B2C solutions to comply with end user requirements.

The Personalization Engine is powered with a social media analyzer which derives customer interests once access is granted through the login process. The Engine generates insights that can be used to recommend personalized products even before the customer initiates the first search. Personalization Engine provides an API (Application Programming Interface) for business users, which provides services such as customer profiling and product recommendation. The API facilitates integration of personalization as a service in customer websites and  other online applications. The Personalization Engine is enriched with contemporary technologies to provide the best personalized experience for users.

Images collected through social media profiles are analyzed using neural networks to identify the interest and score related tags. Posts in social media profiles are analyzed using sentiment analysis. These derived tags are mapped into user interests using expert recommendation system. Generated user interests are used to recommend most suitable products to users.

Since user interests are varying with time, Codegen Personalization Engine uses several time decay functions such as logarithmic and exponential to model user interest variation over time. Parameters used in those functions are specified to user profiles, since different users get different time decay models. Ability to recommend products to upcoming dates is a special feature in Personalization Engine. This is done through time series predictions. ARIMA model is used with previously modeled time series to calculate the future interest variations of a given user.

Event identification model of the Personalization Engine is capable of classifying special events of users, such as birthdays, anniversaries etc. User relevant recommendations are manipulated according to forecasted future events and interest pattern predictions.

The Personalization Engine uses number of machine learning and data mining techniques to determine recommendations. User based collaborative filtering is one of them. To perform collaborative filtering, users wish list and other actions are feeded to the system through the API. Using those user actions, ratings to items are generated. Those are used in collaborative filtering to suggest new items to users. And clustering methods are used to identify similar products and users to recommend products to cater customer interests and requirements. Along with those algorithms, an expert recommendation algorithm is running to give cold start for newer users or newer products.