Enhancing University Selection for Admission Seekers a Data Mining-Driven Recommender System for Undergraduate and Graduate Programs using PCA

Research areas:

• Volume 12, Issue 6, May 2024

Year:

2024

Type of Publication:

Article

Keywords:

Recommender System, Academic Profile Matching Graduate University Selection, Data Mining Techniques, K nearest Neighbors KNN, Machine Learning Algorithm

Authors:

Ahmed M. Abdelrazek; Omar M. Elzeki; Hazem El-Bakry

Journal:

IJAIM

Volume:

12

Number:

6

Pages:

34-45

Month:

May

ISSN:

2320-5121

Abstract:

This paper introduces a recommender system tailored for prospective undergraduate and graduate students, providing essential assistance in selecting the most fitting graduate university in line with their academic profiles. Utilizing various data mining techniques, we transformed a diverse student information database into a standardized universal format, drawing insights from the academic experiences of successful students who pursued studies abroad. To enhance the recommendation system, a machine learning algorithm was developed to assess the similarity between training and test data using weighted scores. While the primary algorithm employed was K-nearest Neighbors (KNN), alternative approaches, including Nearest Neighbors and cKD Tree from the scikit-learn library, and a custom implementation of K-nearest Neighbors, were considered. Importantly, Principal Component Analysis (PCA) was integrated for dataset dimension reduction to refine the model. The core process involved computing the top N similar users for each test user, utilizing the K-nearest Neighbor algorithm and a feature-weighted approach to identify users whose academic profiles closely aligned with those of admission seekers. Subsequently, the system recommended the top K universities to users from the N identified similar users. The notable consistency and alignment of results across all algorithms with the dataset underscore the reliability and effectiveness of the recommendation system. Evaluation using a sample dataset further reinforces its capability in guiding admission seekers through the university selection process.

Full text: IJAIM_675_FINAL.pdf