Introduction

AI or Artificial Intelligence is the basis for the imitation of human intelligence by developing and using algorithms designed into a complex computer environment. In simple words, AI is trying to make computer think and act like humans. The way we do business is greatly revolutionized by artificial intelligence. From decision making to self-governing, AI is slowly changing how the business activities is carried out. Recent progress has led to the availability of artificial intelligence technologies for a broader audience as the number of applications in areas such as intelligent decision support systems has increased. In the recent years, AI has been used in many sectors though the amount of AI used may vary.

One of the most popular and most used form of AI is a recommendation system. Recommender systems are algorithms that recommend items to users that are relevant to them. With the ever-increasing amount of information available on the internet, recommendation systems have become a typical tool for assisting Internet users in finding what they need. The recommender system works with a vast amount of data by filtering the most important information based on the data provided by the user and other criteria such as the user's preferences and interests.  Recommendation systems are able to determine which items are of interest to a certain user by leveraging a variety of information connected to users and their behaviors. A recommendation system goes through four stages:

• Data Collection

• Data Storing

• Data Analyzing

• Content Filtering

  
  
  

  
  
  

Collaborative Filtering

The process of collaborative filtering is based on collecting and evaluating data on user activities. This involves tracking the user's online behavior and forecasting what they would like based on their similarity to other users. One of the key advantages of this recommendation system is that it can make exact recommendations for complex objects without having to understand the object itself.

To visualize and quantify these similarities, collaborative filtering employs a matrix-style technique. Collaborative filtering has the virtue of not necessitating material analysis or comprehension (products, films, books). It just chooses things to suggest based on the user's profile.

Content-Based Filtering

Based on other users' previous activities or explicit input such as product ratings, the content-based filtering method uses product features/attributes to recommend other products comparable to what the user enjoys.  

Content-based recommendation systems work on the premise that if a person like one item, he or she would enjoy another that is similar. This type of recommender makes suggestions to customers depending on how similar the products or the users' backgrounds are.

Hybrid Filtering

A hybrid recommendation system incorporates both content and collaborative filtering. Combining collaborative and content-based filtering may help us overcome the limits we face when using them separately, and it may even be more successful in some cases.

Netflix is an outstanding example of a hybrid recommendation engine. It employs collaborative filtering to consider the user's preferences, as well as content-based filtering to consider the movie or show's descriptions or features.

Some of the pros of a Recommendation system are listed below:

• The customers are constantly engaged in the system ultimately leading to increase in sales and average order value.

• The revenue for a company is increased drastically.

• The use of a recommendation engine allows for focused traffic to be sent to a website.

• Recommendation system can help create a consistent brand experience.

Some of the cons of a Recommendation system are:

• The system can be inefficient decreases if there is less data available.

• For a huge amount of data, it can take a lot of time to learn.

• The system's accuracy isn't always guaranteed.

KNN Algorithm

KNN is a lazy learning algorithm that is non-parametric. It makes inferences for new samples using a database in which the data points are divided into many clusters. KNN is a great go-to model for item-based collaborative filtering and a great starting point for recommender system development. KNN depends on item feature similarity rather than making any assumptions about the underlying data distribution. When KNN makes an inference on the target, it calculates the \"distance\" between the target and every other target in its database, ranks the distances, and returns the top K nearest neighbour recommendations as the most comparable.

The following methods describes how K-NN algorithm works:

• Decide on the number of neighbours.

• Determine the distance between the members.

• Find the closest neighbour based on the distance (Euclidean, Manhattan, Hamming distance or Cosine Similarity).

• Count how many data points there are in each category among the k neighbours.

• Assign the newly acquired data points to the category with the most neighbors.