Abstract:
Now a day’s e-commerce widely uses recommendation system to recommend
the customer by mining the historical data of purchases made by its customers.
The recommendation system tries to analyze user and item tends do
recommended relevant items to the customer. This helps them to increase their
sales. The recommended system is either item based, user based or have
combined approach. However, it does not automatically evaluate the accuracy
of those recommendations. It has been observed that some of predicted
recommendations have been irrelevant to the user. There is no mechanism to get
the accuracy of those recommended items. There is also limitation with ranking
of recommended items. A user clicks on the recommended items and clicks only
on the most relevant item irrespective of its rank. This is a kind of inherent
feedback given by the user by not clicking on irrelevant items. In this
dissertation work improved relevant ranking and reduce the mean absolute
error.