نموذج SLEUTH-CNNsاقتراح لتعديل مفاهيمي في نموذج الذكاء الاصطناعي الجغرافي المعتمد على الخلايا الذاتية متعددة الأنظمة

The cellular automata-based SLEUTH is an independent fully operated intelligent dynamic simulation and prediction model for urban growth. It was released simultaneously with the beginning of the second millennium by the geographical scientist "Keith Clarke" and his colleagues in the U.S.A. This model infused computer science and programming into the science of urban geography. Although it was first designed to simulate and predict an American urban environment, it was universally acclaimed by many researchers around the world. Thus, many different applications of the model spread across the globe, resulting in hundreds of published scientific papers that used it world widely. Yet, the geographical environments are varied from study area to another in most researches and its final output results were mostly insignificant and not reflecting the reality. Being an open-source code model, SLEUTH has gone through several evolutionary stages of programmatic modifications either were done by few modelers or by Clarke himself with his co-modelers. The most challenging and critical modeling modification is concerning the improvement of the algorithm's performance in terms of calibration and reducing computation time while maintaining its basic structure. The SLEUTH-GA model is one modification of the model that aims to replace the Brute Force (BF) calibration method with a Machine Learning based Genetic Algorithms (GA). From this respect, this research is a new attempt to offer a novel modification methodology for creating a complete modeling integrated multi-system which consists of a set of artificial intelligence subsystems that are implemented together in a parallel programming approach. This proposed modification transformed SLEUTH from being an applicable intelligent model to be an individual complete intelligent software program. The novelty of this research also comes in proposing a full integration of the Deep Learning (DL) based Convolutional Neural Networks (CNNs) in both processing and calibration modeling modes. This promising attempt of this suggested modification offers a new Geo-Artificial Intelligence approach that can be used by many researchers not only to simulate and predict future urban growth, but also it can be applied for other geographical growing phenomena.