Neural Network Techniques in Analysing the TVET Instructor Competency Assessment

This study examines the influence of year of working experience on the competency assessment of technical vocational education and training (TVET) instructors using neural networks (NN) and the Curriculum Development Based on Ability Structure (CUDBAS) method. However, despite the widespread use of competency assessments, there remains a lack of a methods that includes year of working experience on instructor competency, which is crucial for training needs analysis (TNA). Experienced TVET instructors are pivotal in training delivery and development, making competency and ability assessments essential. To address this gap, this study applies the combination of CUDBAS Ability Map and NN method to evaluate TVET instructors' competency. Data from three certification courses were analysed using feedforward NN (FFNN) and cascade forward NN (CFNN) models. Neurons in the models ranged from 10 to 50, with performance assessed via regression, mean square error (MSE), mean, and standard deviation. Results show that FFNN and CFNN perform comparably for Courses 1 and 2, while FFNN slightly outperforms CFNN in Course 3, with a 0.2% higher regression value and lower MSE with 11.4% in Course 1, 1.01% in Course 2, and 2.25% in Course 3. Both FFNN and CFNN successfully identified the influence of year of working experience on TVET instructors’ Ability Map assessments, highlighting their potential in enhancing competency evaluations.