A Comparative Study of AI-based Learning Models for Crop Recommendation in Egypt

Abstract

The relationship between food availability, population density, and agricultural sustainability is the focus of research because it is an issue that challenges the entire world, as many countries, including Egypt, have been affected. According to projections by the Food and Agriculture Organization (FAO) and the Intergovernmental Panel on Climate Change (IPCC), the world's population could reach approximately 9.7 billion in 2050. This projected growth places significant pressure on agricultural systems to increase productivity and feed the growing population.; Therefore, agricultural systems have no choice but to increase their productivity if they are to meet the rising demands for food. Traditional farming methods are increasingly failing, due to people's lack of sufficient knowledge about climate change and the soil nature. So, technologies such as the Internet of Things are widely used for modern agriculture to facilitate agricultural processes such as irrigation and kill harmful pests. On the other hand, the use of sensors leads to an increase in cost, as well as the lack of clarity about the reliability of sensors. Also, these technologies ignore important factors, such as the suitability of the soil with its geographical weather properties for specific crops (i.e., crop recommendation). Although many studies have proposed such recommendation models using ML techniques, they do not provide complete data for crop nomination. They ignore essential factors such as soil quality (i.e., nitrogen for plant growth, phosphorus for root formation, and potassium for disease and drought resistance), climate change, and historical crop cultivation data. In addition, the results of these studies were conducted on generated data that may not reflect the actual situation. Some studies conducted in recent years proposed an ensemble learning technique that generated a crop recommendation model. Some of them depend on specific crops such as apples, rice, corn, grapes, bananas, oranges, and coffee. They revolve around improving production in the same soil. In this research, an ensemble crop recommendation model has been proposed to formulate the relationship between the production amount of crops and the previous factors. The proposed model has been generated based on actual data collected from reliable sources (e.g., the General Authority for Meteorology for the Minya Meteorological Station and the General Authority for Meteorology, unpublished data Collected from the General Authority of the Ministry of Agriculture in Minya, and NASA) for a region in Egypt different areas (i.e., Minya Governorate and Beheira Governorate) including factors such as, temperatures, thermal range, winds, rain, and type of soil. The results demonstrate the remarkable effectiveness of the proposed ensemble-based prediction model in recommending suitable crops to specific regions.