Environmental engineering solutions for efficient soil classification in southern Syria: a clustering-correlation extreme learning approach

The classification of soil types for agricultural management systems and environmental engineering is extremely important for land use planning and environmental conservation. Generally speaking, soil classification involves time-consuming resource-intensive methodologies where uneconomical laboratory tests are necessarily needed. Artificial intelligence-based machine learning techniques are being leveraged to compensate for this. Before progressing into machine learning classification methods, statistical analysis are required to understand the type of data and compatibility of applied algorithms. This paper gives an examination of soil type dataset collected from southern Syria by conducting clustering behavior, correlation analysis, and artificial neural network-based classification. Five soil features were involved in the current approach, namely E/N coordinates, elevation values, slope percentage, and rainfall depth. The conducted analysis has leveraged a total of 25 types of soils regardless of the total quantity in each target. It is commonly rare to reach high classification accuracies in such adopted cases of work, however, the resulted 94.9% accuracy significantly advances soil recognition in all considered aspects. It is demonstrably concluded that high potentiality is noted in artificial neural networks where future proposals are written for algorithm optimization perspectives. This proposed methodology will minimize the resources requirements, consumed time of mapping, and manpower necessity which will collectively boost land management practices.