Metaheuristics for Machine Learning

15Soil Aggregate Stability Prediction Using a Hybrid Machine Learning Algorithm

M. Balamurugan

Department of Computer Applications, Acharya Institute of Graduate Studies Bengaluru, India

Abstract

The concept of soil aggregate stability (SAS) is utilized as a criterion for soil composition because several soil ecosystem functions depend on the presence of stable aggregates. Predictive models have gained increased attention as a replacement for direct measurements because aggregate stability is rarely recorded in soil surveys. In order to anticipate soil aggregate stability, this work aims to develop a novel machine learning (ML) technique called the hybrid tree-based twin-bounded support vector machine (HT-TBSVM). Unexpected discoveries were made by examining soil records that provided information on soil properties such as structure, soil nutrient concentration, alkalinity, and moisture aggregates. This collection includes 109 soil samples from Chile’s hyperarid, arid, and semiarid regions, as well as humid areas, including cultivated fields, grasslands, and tree plantations. The most prevalent soil types in this dataset were clay loam, sandal loam, and loam, and the values for each soil attribute ranged widely. The mean absolute error (MAE), R², normalized root mean square error (nRMSE), and root mean square error (RMSE) are used as different indicators to determine the effectiveness of the given prediction. We evaluated these measurements in comparison to conventional approaches ...