Synthetic Data Generation for Engineering Education: A Bayesian Approach

Engineering education research often requires large amounts of data that can be time-consuming and costly to collect. In response to these challenges, synthetic data generation has emerged as a pragmatic solution across various industries, allowing researchers to sidestep the complexities associated with data collection and preprocessing, thereby streamlining the focus on actual model implementation. Despite the widespread adoption of synthetic data in diverse domains, its integration into engineering education research remains relatively underexplored.This paper aims to fill this gap by proposing a novel method for synthetic education data generation through the application of Bayesian approaches. The proposed approach involves leveraging a domain knowledge base to construct a Bayesian network and its corresponding Conditional Probability Tables (CPTs). The subsequent use of Gibbs sampling facilitates the generation of synthetic data, enabling a direct comparison between the characteristics of the original and synthetically generated datasets.