Digital Platform Recommendation System with Scalable model using Clustering

DOI

https://www.doi.org/10.59256/indjcst.20250403026

Published In

Volume 4, Issue 3 (2025)

Published Online

September-December 2025

Pages

160-166

Article Type

Original Article

Check for updates

A Python-based digital platform recommendation dashboard for multi-criteria evaluation of major websites is presented. The system computes privacy, design, collaboration, popularity, and innovation scores from various web metrics and applies K-Means clustering to group sites by their characteristics. This model employs a correlation-based multi-domain scoring system encompassing the above mentioned scores. An interactive Streamlit GUI enables users to select focus areas (e.g. design or privacy) and view recommended top-ranked websites accordingly. The dashboard uses feature scaling and the Elbow method to determine the optimal number of clusters and it also provides visualizations of cluster distributions and recommendations showing transparency and interpretability. This explainable approach contrasts with conventional ranking only by popularity offering transparent multi-dimensional insights into website quality and suggesting improvements. The results (Elbow plot, cluster scatter) demonstrate meaningful groupings (e.g. high-design vs. high-popularity sites) and validates the recommender logic. The system’s implementation leverages Pandas, scikit-learn, Matplotlib/Seaborn, and Streamlit, and its evaluation highlights both practical utility and areas for future enhancement.