Using Collaborative based (User based and Item based) Filtering Techniques
In recent years, recommender systems have been in great demand in every field, and so has it been in the world of readers. There have been a lot of recommendation techniques that have been developed in the past years like Content based filtering, Collaborative based filtering and Hybrid filtering. In this project we have used different collaborative filtering techniques like Item-based and User-based collaborative techniques. The User-based collaborative filtering uses correlation factors as the similarity metric.We have performed multiple clustering techniques to build the Item-based recommender models like the KNN, KMeans, DBSCAN and agglomerative clustering. The end results also show a comparative analysis of how well these models perform for three different subset of the entire dataset.
https://www.kaggle.com/code/hilalmleykeyuksel/book-recommender/data
- Exploratory Data Analysis
- User-Based Collaborative Filtering using correlation matrix
- Item-Based Collaborative Filtering using K-means, DBSCAN and Hierarchical Clustering
Finally, the model is able to recommend books by giving either a user as an input or a book title. For all three subsets of the dataset, the Hierarchical Clustering performed the best. This conclusion was drawn by taking the silhouette scores, calinski harabasz scores and the davies bouldin scores into account.