Monday, 17 August 2015

Mahout: Recommendations

Recommendations are all about predicting user behavior and suggest recommendations relevant to him. you may observe in e-commerce sites, when you browsing it shows some recommendations to you.

In some cases, websites cluster users with nearly same behaviors and shows recommendations relevant to this cluster.
Primarily there are two categories of recommendation algorithms.
1.   User based collaborative filtering
2.   Item based collaborative filtering

User based Algorithms
Suppose a new user ‘PTR’ comes to your website and started browsing things. Based on his recent history, you matched him against your company database to discover neighbors (neighbors of PTR are other users who have historically had similar taste like PTR.). Once you identify the neighbors, you can recommend, items that the neighbors like.

Item based collaborative filtering
Item based collaborative filtering is a form of collaborative filtering based on the similarity between items calculated using people's ratings of those items. I will explain about each approach in details later.


Prevoius                                                 Next                                                 Home

No comments:

Post a Comment