A Hybrid Machine Learning Approach for Career Prediction Using Academic Features and Psychometric Personality Models

Abstract

Selecting an adequate career requires taking into account various elements including academics, individual interest areas, and personality characteristics. Conventional career recommender models usually consider only one of these aspects, which is either academics or interest profiling, yielding relatively poor accuracy rates. This paper explores a multi-modal machine learning paradigm that utilizes academic characteristics, vocational interests, and personality traits in recommending careers in order to improve recommendation efficiency and robustness. Three major types of datasets will be applied for training: a Career Recommender Dataset with academic features, a massive RIASEC dataset that provides information about interests, and the Big Five personality dataset. Academic features will be normalized with TF-IDF and NLP techniques, and a logistic regression model will be first generated based on this data. It will then be improved by the addition of psychometric features, and a stacked ensemble will be constructed where a Random Forest meta-classifier will make final recommendations. Experimental analysis will reveal that the integration of diverse sources yields substantial prediction performance advantages in contrast to the use of academic-only data, and it may even provide valuable insight regarding personality-career associations.