“Recommendation System”

Generally, we people are having lots of confusion in choosing the things that what actually we want.

Then we go for the suggestion from others. At this point, the recommendation was overwhelmed

Recommendation system was developed to reduce the information overloaded for the internet users and to make ease information retrieval more efficient

For example, if you want to know about the product details there are two types of preferences, Explicit and Implicit.

Explicit preferences: It having the “Ratings, Review, Vote” for the particular product.

Implicit preferences: In this, we are having the “Click, Purchase, Follow”.

These two will influence the value of the product.

Types of Recommendation system

The most popular recommendation system is

1. Content-based Filtering
2. Collaborative Filtering
3. Knowledge-based Filtering
4. Content-Based: In the content-based, we take account of both users and items information.

For example, generating “user profiles, item profile”. This will helps in building the content based recommendation system.

2. Collaborative Filtering:

Collaborative recommender systems combine the ratings or recommendations of objects, recognize commonalities between the users on the basis of their previous history, search, ratings, and generate new recommendations based on inter-user comparisons.

In collaborative filtering, there are two types

a) Item-based similarity

b) User-based similarity

a) Item-based similarity:

1)For each item, measure how similar they are in terms of having received similar ratings by similar users

2) For each item, identify the k-most similar items

3) For each user, identify the items that are most similar to the user’s purchases.

b) User-based similarity:

1). Measure the similarity of every user to the new user.

2. Identify the most similar users by the widely used options like

a) Taking account of the top k users (k-nearest neighbors).

b) Taking the account of the users whose similarity is above a defined threshold.

3. Rating the items purchased by similar users. The rating is the average rating among similar users and the approaches are:

a) Average rating.

b) Weighted average rating, using the similarities as weights.

4. Pick the top-rated items.

3.Knowledge-based Filtering:

There are situations where collaborative and content-based filtering doesn’t work. In these contexts, we will use explicit knowledge about the users, products, and recommendation criteria.

This branch of techniques is called knowledge-based.

There are a variety of techniques, and they depend on the data and on the business problem.

It can be employed in specific domains where the purchase history of the users is smaller.

In such systems, the algorithm takes into consideration the knowledge of the items, such as features, user preferences asked explicitly, and recommendation criteria, before giving recommendations.

The accuracy of the model can be judged based on how useful the recommended item is to the user.

We take into consideration the following questions:

• What kind of information about the items is taken into the model?

• How are user preferences captured explicitly?