Learning Effect Evaluation of Mobile Eye-Tracking APP Based on LSTM Deep Learning Model

With the fast growth of mobile learning and digital education, traditional ways of assessment can no longer reflect how learners think during the learning process. This study looks at how eye-tracking behavior relates to learning outcomes. It uses a mobile app that records eye movements and builds a new way to evaluate learning based on a Long Short-Term Memory (LSTM) deep learning model. The app collects real-time eye movement data, such as how long learners fixate on something, how far their eyes move, and how their pupil size changes. It also records behavior data like how long tasks take, how often learners pause, and how many times they rewind. These data are turned into sequences and fed into the LSTM model to predict learning outcomes based on the learning process. The experiment shows that task time, fixation time, and pause frequency have strong links to learning outcomes. The model reaches a high level of accuracy, with a prediction score of R² = 0.832. This study shows that using deep learning to work with cognitive process data is possible and effective. It provides useful ideas and tools for personalized teaching and the design of smart education systems.