Entry 8: Recommender Systems and How it's Approached...

This week's content includes, as can be guessed from the title of this post, recommender systems, collaborative filtering, content-based filtering, and hybrid approaches and reasons for which we use recommender systems.

Recommender Systems

Recommender systems are algorithms that provide users a basis for decision-making by providing suggestions pertaining to each user. Meaning platforms that use recommender systems give different options to different users (most of the time). By providing a basis for making decisions, these systems ensure the quality of the decisions made by the users. Many online services including Amazon, Netflix, Youtube and Noon use such recommender systems.

Recommender systems can be approached in three different ways:

• Collaborative filtering

• Content-based filtering

• Hybrid filtering

Collaborative Filtering

This recommendation algorithm works based on either of 2 things:

• a single user's behaviour, or

• the behaviour of other users who have similar traits (that is, group knowledge)

Some of the issues that come up when using Collaborative Filtering include the following:

• Cold Start

• Sparsity

• First Rater

• Popularity Bias

Content-Based Filtering

This recommendation algorithm works based on the following:

• Information about the content that a specific user has watched, read, or selected.

• User's behaviour and not on other's users' opinions.

This algorithm does not have any issues such as cold start or sparsity and is able to recommend new and unique content to the users.

Some of the issues of content-based filtering algorithms include the following:

• It requires content from which you can get meaningful features.

• These features, in the form of a learnable function, should represent the users' tastes.

• Unless the content features include other users' judgments, this algorithm cannot exploit them.

Hybrid Filtering

As the name suggests, Hybrid filtering is a combination of collaborative filtering and content-based filtering. This approach helped increase the efficiency and complexity of the recommender systems. By integrating both these algorithms, you get more accurate recommendations.

Why Do We Use Recommender Systems?

Recommender systems provide value to both the customer and the providers.

Value for the customer includes the following:

• Customers can find new and more interesting content.

• It helps them make better decisions by shortening their list of choices.

• It helps them explore more content.

• And it provides customers with a source of entertainment.

Value for the providers includes the following:

• It adds to the list of services offered to customers.

• This increases the loyalty and trust between the customers and the providers.

• This also increases sales by making the customers' decision-making easier.

• These recommender systems also help providers in collecting data about their respective users.

References:

• AnalyticSteps (2021) What is Collaborative Filtering? Types, Working and Case Study Available at: What is Collaborative Filtering? Types, Working, and Case Study | Analytics Steps (Accessed: 6 March 2024)
  

• LupaSearch (2023) Understanding Content-Based Filtering Available at Understanding Content-Based Filtering (lupasearch.com) (Accessed: 7 March 2024)