An Efficient Web Information Retrieval Model using Machine Learning

Abstract

The rapid expansion of the World Wide Web (WWW) revolutionized how individuals access and interact with information. However, as web content grew exponentially, it became increasingly difficult for users to retrieve relevant information efficiently through simple browsing. This research focuses on the development and evaluation of a web information retrieval model, combining traditional Search Engine Optimization (SEO) techniques with data-driven analysis to improve website visibility, and applying machine learning (ML) for predictive decision-making in the context of higher education enrolment. Search Engine Optimization (SEO) emerged as a critical methodology for enhancing the visibility and discoverability of websites within Search Engine Results Pages (SERPs). SEO includes keyword research, indexing, and on-page and off-page strategies to help content rank well in search results. To assess SEO effectiveness, metrics like keyword ranking, user sessions, referrals, and social media engagement are analyzed over time, typically through tools like Google Analytics. In this research, SEO techniques were implemented in real-time for the website. The analysis was done quarterly to check for enhancement in various factors like keywords, users, sessions, referrals, and Facebook likes. The analysis shows that there is a rise in the website ranking using selected keywords, which led to substantial improvements across key parameters: keyword ranking (77%), user traffic (98%), sessions (97%), referrals (95%), and Facebook likes (52%). This demonstrates how targeted SEO interventions can significantly boost a website s online presence and engagement metrics. newlineHigher Education Institutions (HEIs) face another critical challenge: optimizing student enrolment. Enrolment decisions are influenced by a complex interplay of academic, financial, and demographic variables. Traditional analytic methods often fail to capture the non-linear relationships and interactions inherent in this data. This research presents the application of Machine Learning (ML) models, particularly a stacking ensemble approach, to predict and prioritize student enrolment using real-time data (generated after application of SEO techniques) from an HEI in Delhi NCR, India. The stacking model combines five base classifiers: Extra Trees Classifier, Light GBM Classifier, Support Vector Machine, K-Nearest Neighbor, and Gaussian Naive Bayes, integrated within a Multi-Layer Perceptron (MLP) as a meta-classifier. Hyperparameter tuning is done to enhance model robustness and generalizability. Influential features include State, City, Primary Traffic Channel, Payment Method, Program, Category, Discount, Origin, Publisher, and Form Completion Rate. With an accuracy of 85%, sensitivity of 84%, recall of 85%, precision of 85%, F1-score of 85%, Mean Square Error (MSE) of 15%, and an AUC of 94%, this ML-driven approach demonstrates improved predictive performance over previously reported stacking ensemble methods having the accuracy score of 80%, Recall score of 80%, Precision score of 79%, F1- Score of 79%. Beyond binary classification, the model assigns a probabilistic enrolment score to each prospective student, indicating their likelihood to enroll. The proposed approach was simulated and evaluated empirically and equips admission teams with actionable insights to prioritize high-conversion candidates. newline Together, these findings emphasize the transformative role of data-driven strategies- SEO for digital visibility, and ML-based enrolment probability scoring for targeted enrolment in achieving institutional goals. From strengthening online presence to optimizing student enrolment processes, leveraging data analytics and automation empowers Higher Educational Institutions (HEIs) to make informed decisions in an increasingly competitive landscape. newlineKeywords: Data-Driven Decision Making, Enrolment Prediction, Enrolment Probability Scoring, Enrolment rates, ExtraTrees Classifier, Gaussian Naive Bayes, HEI, Hyperparameter Tuning, Keyword Ranking, KNN, LightGBM Classifier, Machine Learning, MLP Classifier, SEO, Stacking ensemble, Stratified K-Fold, Support Vector Machine. newline newline newline newline newline