Natural disasters, such as floods and earthquakes, posed a significant threat to educational institutions, particularly in high-risk areas like Peru. This study reviewed the application of Convolutional Neural Networks (CNNs) for predicting natural disaster risks in secondary educational centers. CNNs enabled the processing of large volumes of geospatial data and satellite imagery, providing accurate, real-time predictions of vulnerable areas. The systematic review examined the potential of CNNs to predict natural disaster risks specifically in secondary educational institutions located in high-risk zones in Peru. The study assessed the effectiveness of CNNs compared to traditional risk assessment methods based on historical data, highlighting the advantages of real-time predictions for improving disaster response. Additionally, it explored the impact of these technologies on enhancing school safety and disaster preparedness. Through the implementation of a CNN-based web system, the research aimed to strengthen the resilience of educational institutions and contribute to improved emergency planning and response. The review also addressed the limitations and challenges of current implementations, offering recommendations for future developments in disaster risk prediction.