Analysis of Trainee Satisfaction with Mentorship in Early and Preschool Education Using Artificial Neural Networks

Effective mentoring is grounded in reciprocal collaboration between mentors and trainees, where mutual respect and open, reflective dialogue are essential from the very beginning of the internship. The aim of this study was to examine which factors predict trainees’ satisfaction with their mentors by applying a predictive model. The sample consisted of 104 trainee preschool teachers from Croatia. To test the model, perceptron artificial neural networks were used (with one hidden layer and four neurons), employing the hyperbolic tangent activation function. The results indicated a low discrepancy between the training and test datasets, high classification accuracy, and balanced TPR and TNR values, with an AUC above 0.95, confirming excellent predictive model accuracy. The analysis further revealed the relative predictive strength of individual factors related to satisfaction with the quality of mentor cooperation. The mentor–trainee relationship emerged as the strongest predictor, while net salary proved to be a more influential predictor of satisfaction with mentor cooperation than support from principals or professional associates.