• Nearest neighbor search is an optimization problem for finding close points.

• It is defined as follows: given a set S of points in a space M and a query Q E M, find the set of closet points in S to Q.

• There are many applications of nearest neighbor techniques as:

  • Optical character recognition.

  • Content based image retrieval.

  • Collaborative filtering.

  • Document similarity.

• Collaborative filtering is a process whereby we recommend to users items that were linked by other users who have exhibited similar tastes.

Shortcomings:

1. Large search space to find nearest neighbors.

• We can use various MR algorithms to tackle this.

2. Use of distance measure.

• Which particular distance measure is most suitable for any particular application is difficult to determine.

3. Scalability.

• Scalability issue arises if there are many more users than items like Amazon.com.

• This problem can be solved by using adjusted cosine similarity.

• As well we can give weighted to users and highly weighted users can only be considered.

4. High sacristy problem.

• This problem arises when there are very few common ratings between the users.

5. Cold start.

• For a new item when no ratings are available then how to and whom to recommend.

• To solve this we may use user’s demographic or non-personalized data to recommend this item.