Job Recommendation System Using Machine Learning and NLP

In the era of digital employment, job portals generate massive volumes of unstructured textual data that make it increasingly difficult for candidates to discover roles that truly align with their skills and career goals. Traditional keyword-based and job-title-matching systems fail to capture the semantic depth of candidate profiles, resulting in poor recommendation quality and inefficiency for both candidates and recruiters. This research proposes an enhanced Machine Learning and Natural Language Processing (NLP) based Job Recommendation System that accurately extracts technical skills from candidate résumés and job descriptions, encodes them into structured feature vectors, and computes personalised similarity scores to generate ranked, relevant job recommendations. The system incorporates a multi-stage NLP pipeline — comprising text extraction, tokenisation, stop-word removal, Named Entity Recognition (NER)-assisted skill extraction, TF-IDF vectorisation, and cosine-similarity scoring — to produce high-accuracy, semantically grounded recommendations. As a novel contribution, a Hybrid Recommendation Module is introduced that combines content-based skill matching with collaborative-filtering signals, significantly improving personalisation. Experimental evaluation on four candidate résumés against a Kaggle job-description dataset of 50 postings achieves a Precision of 0.86, Recall of 0.86, F1-Score of 0.86, and overall Accuracy of 80%, outperforming both baseline keyword-matching and pure TF-IDF cosine-similarity approaches.