Implementation and Comparative Analysis of Supervised Machine Learning Methods for Domain Modeling and Grade Estimation Techniques

Aug 4, 2026

Upload by Miles

Mining operations depend on reliable mineral resource characterization to improve extraction decisions and ensure safer, more profitable outcomes. Traditional resource modeling is time-intensive and vulnerable to errors from manual interpretation, while geostatistics relies on assumptions that may fail; practices like variogram interpretation and early data filtering can propagate uncertainty. This thesis proposes geospatial estimation methods validated on geological domain case studies using supervised machine learning, compares modeling and grade estimation approaches, and highlights the need for geographic validation and visualization beyond statistics. Future work recommendations include targeted data feeding, better handling of sample relationships with variable orientation, anisotropy and anisotropy ratio, and integration of geochemical data to strengthen predictive performance.

LanguageEnglish

CategoryResearch & Report

FormatPDF

Pages185

Reading Time7 h 46 min

AccessFree to download

Abstract

Table of Contents

List of Figures

List of Tables

Acknowledgements

Chapter 1 Introduction

Problem Statement

Thesis Objectives

Thesis Outline

Chapter 2 Literature Review

Chapter 3 Methodology

Geostatistical Characteristics

Data Validation

Mineralization

Drilhole Compositing

Data Declustering

Variography

Variogram Modeling

Defining Variogram Directions

Defining Lag Parameters

Lag Angle Tolerance

Estimation Criteria

Search Ellipsoid

Geostatistical Methods

IMPLEMENTATION AND COMPARATIVE ANALYSIS OF SUPERVISED MACHINE LEARNING METHODS FOR DOMAIN MODELING AND GRADE ESTIMATION TECHNIQUES by Erkan Ugur Aslan Copyright by Erkan Ugur Aslan 2023 All Rights Reserved A thesis submitted to the Faculty and the Board of Trustees of the Colorado School of Mines in partial fulfillment of the requirements for the degree of Master of Science (Mining and Earth Systems Engineering) . Golden, Colorado Date Signed: Erkan Ugur Aslan Signed: Dr. Kadri Dagdelen Thesis Advisor Golden, Colorado Date Signed: Dr. M. Stephen Enders Professor and Department Head of Mining Engineering ABSTRACT In the face of dwindling economic mineral resources, this study addresses the critical need for reliable characterization of the mineral resources for improved extraction methods and safer, more profitable mining operations. Current resource modeling techniques are time-consuming, and prone to errors due to manual interpretation of detailed data leading to economic viability issues due to uncertainties associated with the estimates. Geostatistics, while a predominant method in resource modeling, depends on assumptions that may not always hold true, and common practices, such as variogram interpretations, and data filtering, can introduce errors early in the modeling process. Supervised Machine learning (ML) offers a promising alternative, capable of handling complex data sets for domain analysis and grade estimation. This research introduces novel geospatial estimation methods, validated through a case study on geological domains using Supervised Machine Learning methods. A comprehensive comparison analysis of various modeling and grade estimation methods is presented, highlighting the potential of supervised ML algorithms as an alternative to traditional geostatistical methods. However, the study also acknowledges the limitations of ML, emphasizing the importance of geographic validation and visualization, apart from statistical analysis, in ensuring methodological rigor. Recommendations for future work include enhancing ML algorithms through specific data feeding, improving sample relationships with variable orientation data, anisotropy, and anisotropy ratio, and integrating geochemical data to enhance predictability. The thesis serves as a foundational guide for future resource estimation endeavors using not only ML algorithms but also geostatistical methods, underscoring the necessity of methodological rigor and validation in both geostatistical and machine-learning applications. TABLE OF CONTENTS ABSTRACT ........................................................................................................................................... iii TABLE OF CONTENTS ....................................................................................................................... iv LIST OF FIGURES............................................................................................................................... vii LIST OF TABLES ................................................................................................................................. xi ACKNOWLEDGEMENTS .................................................................................................................. xii CHAPTER 1 INTRODUCTION............................................................................................................. 1 1.1 Problem Statement .................................................................................................................. 2 1.2 Thesis Objectives .................................................................................................................... 5 1.3 Thesis Outline.......................................................................................................................... 6 CHAPTER 2 LITERATURE REVIEW..............................................................

FAQs

Why does the thesis focus on reliable mineral resource characterization?

It addresses economic and operational risks caused by uncertain grade and resource estimates, which affect both extraction effectiveness and profitability while also supporting safer mining operations.

What limitations of geostatistics does the thesis highlight?

It emphasizes that geostatistics depends on assumptions that may not hold, and that common steps like variogram interpretation and early data filtering can introduce errors and uncertainty.

How does supervised machine learning contribute to domain modeling and grade estimation in this work?

The thesis introduces geospatial estimation methods using supervised ML, validates them through geological case studies, and compares their performance with traditional geostatistical approaches while stressing the importance of geographic validation and visualization.

Aug 4, 2026

Upload by Miles