WebSphere Commerce Recommendation Engine powered by LikeMinds

WebSphere Commerce contains an optional, dynamic recommendation system powered by LikeMinds. LikeMinds is a collaborative filtering engine that analyzes user interactions that occur on your Web site and generates real time predictions and recommendations to your Web site users, which are displayed using Web activities.

WebSphere Commerce Recommendation Engine collects profile information to develop mentor groups. These groups are the foundation for subsequent product recommendations. Customers that fit the profile of a particular community are presented with recommendations based on the behaviors of others in the community.

When a user visits your Web site, WebSphere Commerce captures clickstream data which is stored in your database. For example, the following types of interactions may be recorded:

Using pre-defined rules, LikeMinds surfaces results through a recommendation engine, called the Clickstream engine. The Clickstream Engine tracks clickstream (or rating) behavior and generates recommendations based on mentors who exhibit similar content and product affinities. The Clickstream Engine tracks the pages that users have looked at. As this data is collected, it is analyzed to identify users' traffic patterns. Finally, the Clickstream engine, then makes content recommendations for each specific user, using data from relevant subsets of the user base.

Typically, after a user has completed a minimum number of transaction activities, that user is assigned a set of mentors. A mentor is a specially designated user who has visited the Web site a number of times, and whose profile is similar to the user's. LikeMinds uses a technique called collaborative filtering to build a mentor's profile for each user to predict how much a user will like particular items and which items that user will enjoy, buy, or add to their shopping cart.

The Clickstream engine

You can configure your LikeMinds Clickstream recommendation engine to control the predictability of the recommendations returned.

The engine generates recommendations based on a record of user shopping behavior as the user navigates through a Web site; that is, a history of user "clicks" during the user's Web site visit. It uses data such as the items the users view, click on, and add to their shopping carts. The engine relies on the sifter utility to assign mentors to users.