Collaborative filtering is a method of processing user interests
Visitors: 394
Collaborative filtering - a technology that allows you to predict the preferences of a particular user of the Internet resource, comparing its interests with the interests of other visitors to the resource. Based on this information, the recommendation section offers those products that were of interest to the audience of the site, and a particular user - not yet.
This method of processing user interests is not new
It is widely used on such major services as Amazon, Netflix, or in large social networks. Thanks to intelligent algorithms companies and organizations are much better understand the needs of their customers, and customers buy the most relevant products.
The principle of collaborative filtering is quite simple: if a user has made purchases or simply went to a product page, the system finds other users with similar requests. After that, the system recommends to the user those products that other customers were interested in, and the user - not yet. Otherwise, this algorithm is called User-Based (literally translated from English as "user-based").
The use of collaborative filtering can be conventionally divided into two approaches:
- User-Based (literally translated from English as "user-based") - applies to users with similar interests;
- Item-based - applies to similar products.
As a rule, users leave in system of estimation of objects which can be both obvious (for example, a rating on a 5-point scale, "likes" or an increase in the status), and implicit (for example, quantity of views of one card). In both cases, the algorithm compares the actions of a particular user with the actions of the site audience.
The method of collaborative filtering is applied.
YouTube applies a selection of recommended videos based on the videos you have watched, the service recommends you to view similar content. The service also offers to get acquainted with materials or sources of other users, whose interests coincide with the active user to the maximum.
Conclusion
Of course, collaborative filtering is not an ideal solution for building user recommendations. However, this method has demonstrated a very special approach to the selection of recommended products.
The previous algorithms of recommendations took into account only the interests of a particular user, thus closing it in the circle of their own preferences, isolating it from new, unfamiliar content. Now the user has the opportunity to familiarize himself with the full range of products resource and choose something fascinating and useful.
Comments 0