Unveiling Student Trial-and-Error Patterns through a Optimized Deep Learning for Accurate Profiling

Student trial-and-error behavior plays a crucial role in cognitive development, yet traditional profiling methods often fail to capture the underlying patterns accurately. This study introduces a novel approach to uncovering these patterns through a deep learning framework designed for accurate profiling. The framework utilizes a hybrid model combining DBSCAN clustering, Multilayer Perceptron (MLP) classification, and Genetic Algorithm (GA) for training optimization, enhancing model accuracy. One of the key challenges addressed by this framework is the invisible distribution of trial-and-error tendencies in the data, as the class distribution is not readily apparent, small samples, and the dataset suffers from class imbalance. The results demonstrate significant improvements in profiling accuracy compared to traditional methods, with accuracy reaching 86.67%, precision at 89.45%, recall at 86.67%, and F 1-score of 85.78%. This research provides a comprehensive solution for predictive cognitive characteristic, laying the foundation for personalized learning strategies and improved educational outcomes.