Scholarlink

<p>This paper aims to investigate the use of machine learning (ML) regression techniques to consider information from multiple variables in the estimation of density. The idea is to build a new variable that combines information from multiple variables related to density using ML regressors. The multiple variables may be either continuous or categorical. The new variable obtained by ML is then used as secondary information to estimate density using multivariate geostatistical techniques. The impact of this new variable on the precision and accuracy of the density estimates is investigated. A case study with data coming from an iron deposit is presented. Eight different regression techniques were initially investigated. The best machine learning (ML) regressor was used to build a new secondary variable. This secondary variable is considered to estimate density using multivariate geostatistical techniques. Three multivariate geostatistical methods were tested: (i) standardized ordinary cokriging, (ii) simple kriging with local varying mean, and (iii) ordinary kriging with variance of measurement error. The methodologies were compared against traditional ordinary kriging and standardized ordinary cokriging using a super-secondary variable as secondary data. The results show that a secondary variable obtained by ML regressors that combines multiple variables does enhance the density estimates. This new variable resulted in density estimates that are more accurate and precise than methods that used the original variables as secondary information.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Machine Learning Regressors: An Alternative to Compact Grades Information, Generate Secondary Information, and Improve the Density Block Models

  • 期刊论文

  • 发表日期：

    2025年8月7日

查看全文
  • W. Emilio G. Moreno,
  • Áttila Leães,
  • Marcel Antonio Arcari Bassani,
  • Diego Marques,
  • João Felipe Coimbra Leite Costa

摘要

This paper aims to investigate the use of machine learning (ML) regression techniques to consider information from multiple variables in the estimation of density. The idea is to build a new variable that combines information from multiple variables related to density using ML regressors. The multiple variables may be either continuous or categorical. The new variable obtained by ML is then used as secondary information to estimate density using multivariate geostatistical techniques. The impact of this new variable on the precision and accuracy of the density estimates is investigated. A case study with data coming from an iron deposit is presented. Eight different regression techniques were initially investigated. The best machine learning (ML) regressor was used to build a new secondary variable. This secondary variable is considered to estimate density using multivariate geostatistical techniques. Three multivariate geostatistical methods were tested: (i) standardized ordinary cokriging, (ii) simple kriging with local varying mean, and (iii) ordinary kriging with variance of measurement error. The methodologies were compared against traditional ordinary kriging and standardized ordinary cokriging using a super-secondary variable as secondary data. The results show that a secondary variable obtained by ML regressors that combines multiple variables does enhance the density estimates. This new variable resulted in density estimates that are more accurate and precise than methods that used the original variables as secondary information.