dc.description.abstract | This paper presents a validated conceptual model for evaluating graduates using machine learning techniques by mapping their
problem solving skills to industry jobs’ competence requirements. This is because, for college graduates, knowing the right
industry role that suits them based on their competences has remained critical when searching for jobs after graduation. Indeed,
as thousands of university students graduate each year and enter the market to search for jobs that are limited so then are they
exposed to a high risk of not only long search but also job mismatch on employment. In order to enhance both their quick
employability and optimal performance in the job, evaluation of graduates’ possession of relevant skills is necessary by not only
employers but also training institution. In fact, problem solving is one of the skills acquired by graduates during training and
strongly sought for by employers during evaluation, yet it is not clear which of its predictor attributes are related to enhanced
performance in the job. Besides, evaluation is supposed to be predictive by matching skills possessed by graduates with those
required by the job. Thus, predictive evaluation using data-driven techniques such as machine learning may greatly promote
graduates’ performance in the job. However, we do not have a validated conceptual model for machine learning-based
predictive evaluation of graduates skills towards industry roles that can be used by both employers and learning institutions. As
a result, there is a mismatch between skills possessed by graduates and those required in the job whose impact is evidenced by
high employee turnover, poor productivity and low job motivation. This paper focusses on this gap by addressing two
objectives: 1) to outline theoretical conceptual development 2) to develop experimental conceptual validation methodology.
Theoretical development was approached through two cognitive dimensions, namely knowledge and skills, and were derived
from three cognitive theories. A total of 13 concepts were revealed as follows: 4 independent and 9 confounding. Validity of
these concepts was investigated empirically where 5 concepts were confirmed as valid, namely relevant content knowledge,
cognitive skills, technical skills, academic capacity, and age. The machine learning implementation of the validated conceptual
model recorded an average accuracy of 88.6% on a carefully selected benchmark dataset. | en_US |