Teaching Machine Learning Concepts to Engineers: A Practical Approach With Free Software

ABSTRACT In recent years, materials engineering education has undergone significant reform due to the transformative impact of computer simulations. This shift towards a more digital curriculum requires accessible and intuitive software for both educators and students. This study evaluates the use of a free and user‐friendly data mining and machine learning platform (Orange Data Mining) as a possible pedagogical tool to teach basic concepts of artificial neural networks (ANNs) in materials engineering education. The platform enables the creation of customized workflows and simplifies complex data analysis without requiring programming skills. As a practical application, the study investigates the prediction of inorganic compound solubility in water (412 compounds) and the dissolution rates of inorganic glasses (295 entries from 19 compositions) using experimental data obtained from the literature. Input features comprised elemental composition, temperature, and pH conditions. The best‐performing model consisted of an ANN with 30 neurons in a single hidden layer, using the ReLU activation function—selected for its computational efficiency and suitability for nonlinear modeling—and 400 iterations, chosen to balance model capacity and generalization in a teaching‐oriented scenario. The results demonstrated high predictive accuracy, with a mean squared error (MSE) of 0.5 for solubility and 0.03 for dissolution rates. Additionally, the study illustrates both classical and modern interpretations of ANN behavior, including overparameterization effects. Overall, the findings indicate that Orange is an effective and accessible tool for teaching machine learning concepts in introductory classrooms, allowing the exploration of complex chemical and physical relationships using simple computational resources.