Your customers are looking for relevant products and services. With our data scientists, you can develop a recommender system that offers personalized recommendations.
A recommendation engine is an intelligent information filtering system that predicts the rating or the preference a user would give to a particular item. Recommendation engine can keep track of casual browsing done by a user, including price and product comparison. This allows the engine to make personalized recommendations and suggest the next best action to the company. It is built using one of the three models:
Our data scientists will study your business needs, understand your industry and then work with you to give you a recommendation engine or recommender systems as a service that provides useful recommendations.
Our client is a major online bookstore in Europe. The bookstore has hundreds of thousands of books and in order to compete with other online bookstores, the company needed to know user preferences and how the ratings that they give affect the overall sales.
The data our data science team worked on consisted of three files - users (demographic data), books (book title, author name, publication year etc.), and book ratings (user IDs along with ISBNs and book ratings for all the books).
The first step in our project was uploading all the CSV files to the Azure Machine Learning Service since it allows for an easy and quick way to build, test and deploy predictive analytics solutions on datasets. After this, our team implemented the Train Matchbox Recommender module to train the recommendation engine.
The recommendation engine created in this manner now gives the bookstore three kinds of predictions which we will explain through the example of two members whom we will call John and Jane.
Recommends items to a given user: If John likes the Harry Potter series and has also read The Lord of the Rings, then Jane will also be recommended The Lord of the Rings when she checks out any of the Harry Potter books. This is because the recommendation engine recognizes the common pattern between the two users in their choice of books.
Predict ratings for a given user and item: The system is even able to predict what rating Jane will give to a book based on the rating given to it by John.
Find items related to a given item: This model involves giving book recommendations to Jane based on the genre that she likes. If she is a fan of science fiction stories, the system will recommend science fiction literature.
The recommendation engine that our client now uses has a NDCG (normalized discounted cumulative gain) score of 0.77 (on a scale of 0 – 1) when it comes to recommending books to a user.