Multi-Level Hybrid Machine Learning Approach for Early Identification of Poor Performers and Industry Readiness Assessment.

Abstract The growing need for graduates who meet academic and industry expectations has highlighted the importance of detecting learning challenges early. Traditional evaluation practices rely heavily on examination scores and often overlook technical skills, behavioural attributes and socio-economic influences. This study proposes a multi-level assessment framework supported by baseline machine-learning models and a hybrid KNN–RNN architecture to classify student performance and identify those requiring intervention. A dataset of 1,500 undergraduate learners was analysed across academic, skill-based and personal attributes collected over several years. The hybrid model achieved improved prediction accuracy compared to classical classifiers and offered clear insights into student preparedness for higher studies or employment. The findings demonstrate how multi-stage analytics can strengthen the academic support system and guide institutions in developing targeted strategies to reduce dropout risk and enhance career readiness. Keywords: Academic analytics, machine learning, KNN, RNN, dropout prediction, employability, higher education, skill assessment, performance correlation.