Development of a hybrid machine learning model for classification of soil types based on geophysical parameters

Abstract

In this paper, a hybrid model based on RandomForestClassifier and MLPClassifier is presented, achieving an accuracy of 96.07% in the task of soil classification based on geophysical parameters. The results demonstrate the advantages of the proposed approach over selected classical algorithms, indicating a high practical value for precision agriculture and environmental monitoring. A dataset containing key soil parameters such as electrical conductivity, density, P-wave velocity, and depth was utilized. Prior to training, the data were preprocessed: the target variable was converted to numeric format using LabelEncoder, and the features were standardized using StandardScaler to bring them to a common scale. Data were divided into training and test samples using the train_test_split method (80% training, 20% test).