Clusters of users

Another way to use community detection to improve a recommendation engine is to try and create groups of customers. In that way, you can create groups of customers with similar habits, which can reflect similar interests. On an e-commerce website proposing sport-related articles, you can create a community of fishermen and a community of football players, without any prior knowledge about your users except their purchases: if the purchased products are mostly about fishing, you know this customer is likely a fisherman. This knowledge can be used to extend the list of relevant recommendations. For instance, some new football socks that have not been bought by anyone yet can probably be recommended to the users identified as being football players.