American
International
University-
Bangladesh

Students & Supervisors

Abstract

Recruiters and hiring managers often face challenges in effi-ciently evaluating large volumes of job applications. Traditional manual screening methods are time-consuming and prone to bias, leading to sub-optimal hiring decisions. This research presents an automated approach for job application evaluation using Natural Language Processing (NLP) techniques and machine learning regression models. The methodology involves data preprocessing, feature extraction using techniques like TF- IDF, Word2Vec, and Latent Dirichlet Allocation, and sentiment analysis for candidate profiling. Regression models, including Random Forest, Support Vector Regression, and LightGBM, are applied for predictive scoring. The results indicate that text similarity metrics, particularly co-sine similarity, yield the highest correlation with job matching scores, while ensemble learning models, especially LightGBM, demonstrate su-perior predictive accuracy with an MSE of 0.0096 and an R² value of 0.6358. The findings suggest that optimizing feature selection and em-ploying advanced ensemble learning techniques significantly enhance job application screening. The impact of this research is substantial, as it provides an efficient, unbiased, and scalable approach to CV screening, reducing recruiter workload and improving the precision of candidate selection.

Keywords

Publication Details

• Type of Publication: Conference 
• Conference Name: International Conference on Data Science, AI and Applications
• Date of Conference: 18/07/2025 - 18/07/2025
• Venue: Dhaka, Bangladesh
• Organizer: EATL Innovation Hub, University of Salford Manchester, DSAI Hub