This work-in-progress research paper describes a study of different categorical data coding procedures for machine learning (ML) in engineering education. Often left out of methodology sections, preprocessing steps in data analysis can have important ramifications on project outcomes. In this study, we applied three different coding schemes (i.e., scalar conversion, one-hot encoding, and binary) for the categorical variable of Race across three different ML models (i.e., Neural Network, Random Forest, and Naïve Bayes classifiers) looking at the four standard measures of ML classification models (i.e., accuracy, precision, recall, and F1-score). Results showed that in general, the coding scheme did not affect predictive outcomes as much as ML model type did. However, one-hot encoding - the strategy of transforming a categorical variable with$k$possible values to$k$binary nodes, a common practice in educational research - does not work well with a Naïve Bayes classifier model. Our results indicate that such sensitivity studies at the beginning of ML modeling projects are necessary. Future work includes performing a full range of sensitivity studies on our complete, grant-funded project dataset that has been collected, and publishing our findings.

This Work-In-Progress research paper presents the investigation of a decision tree model that was trained to predict engineering students' first-semester math performance using demographic and survey data. This is a small step in a larger project that will develop a predictive AI model that can identify students at risk of leaving engi-neering. Ultimately, we will pair a predictive model with an explanation method to identify targeted interventions that can be implemented within the first year of engineering school. Our findings from this project indicate that we may be able to successfully identify students at risk of low performance in first-semester math courses, and design effective individualized interventions.

This Research to Practice Full Paper presents an investigation of engineering retention using machine learning models. We use random forests and artificial neural networks in the form of multilayer perceptrons to analyze the interaction between different factors, such as demographic information, standardized test scores, first semester grades, and surveys to predict student retention in engineering. We find that obtained models can predict with good accuracy if students will remain in engineering, with F1 scores of at least 75 percent. We find that each model places different levels of importance on distinct factors.