Dependency of the learning technique on the problem nature

Abstract

Indicators perception and prediction based on available datasets have recently gained increasing importance. Artificial intelligence (AI) is the backbone of perception and prediction; learning techniques are being used by most researchers to achieve these goals while ontologies are being used to collect, represent, understand and use input data. Using a comprehensive ontology can improve the process of incrementally learning a visual concept detection model. The problem nature may be in healthcare, Transportation, etc. Applying AI to different environmental sectors like solar irradiation, agriculture, water domain and other natural disasters has increased in recent years due to weather changes and human activities. Achieving high accuracy and high efficiency have always been challenges for researchers for faster natural disaster management or natural phenomena exploitation in economic development. With inflating data, there is a direction to deep learning models and hydride methods that enhance the outcome. This paper reviews how artificial intelligence applied in different environmental applications and the development stages of AI models until now. It shows the advantages and disadvantages of each model and provides appropriate recommendations for each application to achieve the best forecasting.