One of the most important applications of data science is the recommendation system. Every organization requires a good recommendation system to express a large range of items. This research focuses on the creation of a book recommender system using collaborative filtering.

The recommendation engine is defined as a machine learning (ML) class that offers the right suggestions or recommendations to the user or organization. The recommendation system helps the user or the company to create a trust for the services and products. These systems are classified into collaborative filtering (CF), content-based filtering, and a combination of both. The book recommendation system recommends the books to the readers based on their styles and interests. The online sites use book recommenders that provide several electronic book categories [1, 2]. This research focuses on the creation of a book recommender using the CF approach.

This session focuses on developing a powerful book recommender using the CF approach [3, 4, 5, 6, 7, 8]. The datasets are taken from the URLs:

1. https://www.kaggle.com/datasets/rxsraghavagrawal/book-recommender-system
2. https://thecleverprogrammer.com/2020/05/23/book-recommendation-system-with-machine-learning/

The datasets are BX-Users, BX-Books.csv and BX-Book-Ratings.csv. These consists of the fields: book-title, publication year, publisher, user id (users), ISBN (books), and book-rating.

The steps in the creation of the book recommender are as follows:

1. Import the necessary libraries and load the datasets.
2. Remove the unnecessary columns and rename the required columns.
3. Check the reliability of the dataset using the command: books.head()
4. Apply the matrix factorization for the CF.
5. Create the matrix with users as columns, books as indices, and values as ratings.
6. Consider the users who rated at least 250 books and book ratings as 50 for each user.
7. Perform the exploratory data analysis using ratings['user_id'].value_counts().
8. Get the users those who rated at least 250 books.
9. Combine the ratings with the books.
10. Filter the books with ratings of at least 50.
11. Define the pivot table with users as columns, books as indices, and values as ratings.
12. Develop the model using KNN training and convert the pivot entries into spare form.
13. Apply the nearest neighbor and check the top 10 books.

The performance measures are shown in figure 1 and the sample recommendations using the KNN model are shown in Figure 2.

Figure 1

Performance measures

Figure 2

Sample recommendations

The book recommender is developed using the CF approach as unsupervised learning. This model required some preprocessing steps. In the future, the proposed model can be integrated with the other soft computing strategies [9, 10, 11, 12].