Machine learning-driven maze behavior analysis: a novel paradigm for learning and memory evaluation

Maze behavior analysis, pivotal for assessing learning and memory, is advanced through a novel machine learning paradigm integrating LSTM and MobileNetV3. LSTM models temporal dependencies in Morris Water Maze trajectories, predicting memory decline via escape latency, while MobileNetV3 extracts behavioral biomarkers from navigation sequences for early detection of AD/ADHD. Theoretical validation highlights LSTM’s superior RMSE and MobileNet’s feature precision, despite challenges in data scale and ethical translation. This approach bridges neurobehavioral insights with computational efficiency, offering a scalable framework for cognitive evaluation.