Comprehensive study on the Python-based regression machine learning models for prediction of uniaxial compressive...

Area of Science:

• Geotechnical Engineering
• Rock Mechanics
• Machine Learning Applications

Background:

• Uniaxial Compressive Strength (UCS) is vital for geomechanical applications.
• Direct UCS testing is challenging for certain rock types.
• Indirect estimation methods are crucial for rock characterization.

Purpose of the Study:

• To develop and evaluate Machine Learning models for predicting Charnockite UCS.
• To compare the performance of Simple Linear Regression and Step-wise Regression models.
• To identify key factors influencing Charnockite UCS.

Main Methods:

• Implementation of Simple Linear Regression and Step-wise Regression models in Python.
• Utilizing Ultrasonic Pulse Velocity (UPV), Schmidt Hammer Rebound Number (N), Brazilian Tensile Strength (BTS), and Point Load Index (PLI) as input parameters.
• Evaluation using Coefficient of Regression (R²), Root Mean Square Error (RMSE), and Mean Absolute Error (MAE).

Main Results:

• Both models exhibited high predictive capabilities for Charnockite UCS.
• The Step-wise Regression model achieved R² values of 0.99 (testing) and 0.988 (training).
• Ultrasonic Pulse Velocity (UPV) was identified as a significant predictor of UCS.

Conclusions:

• Machine learning models provide accurate indirect estimation of Charnockite UCS.
• Step-wise Regression is a highly effective method for predicting rock strength.
• UPV is a critical parameter for assessing rock geomechanical properties.