Scholarlink

<p>The capabilities of the pilot classifier model trained to recognize traces of microbial activity on solid surfaces identifying the formation of soil-like bodies were preliminarily assessed. A database of 500 samples described by the authors and taken in open sources from 1988 to the present time was collected for machine learning; among them, 59 samples were soil horizons, 146 samples represented parent rocks and soil-like bodies; there were also rock-forming minerals, accompanying components of soil formation, and xenobiotics common in technogenically transformed landscapes of the world. The samples included in the database differed in their size, coverage with biofilms and films of other nature, and chemical and physical processing. The array of sample features significant for machine learning included quantiles of the wetting contact angle distribution and generalized categorical indicators of surface geometry, mineral composition, and state of organic matter. The target function of the classification was the presence of stable traces of microbial activity on a solid surface. Missing data were reconstructed using Monte Carlo procedure and bootstrapping. As a result of numerical experiments on optimizing the machine learning, a balanced training dataset containing 1233 pseudo-sample elements was obtained. Six classifier models with parameter variations were trained and evaluated. The most productive classifier—a five-layer neural network with randomly dropout neurons—demonstrated the prediction accuracy of 0.74 and the ROC AUC of 0.80 on the test sample, which is higher than that of simpler and faster classifiers (accuracy and ROC AUC of 0.70). Based on the disagreement between the classifications of a human expert and a trained algorithm, common features of samples that are difficult for machine classification were established: traces of life activity, carbonates, and high degree of dispersion. This is important for determining the direction of collecting information to improve the performance of the classifier. The development of an algorithm for recognizing traces of microbial activity is useful for clarifying the mechanisms of biogeochemical and biogeotechnological processes in soils of various origins, including soil formation and terraforming.</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

Application of Machine Learning Algorithms to Classify Soil Components with Different Hydrophilicity

  • 期刊论文

  • 发表日期：

    2025年2月27日

查看全文
  • O. A. Sofinskaya,
  • F. A. Mouraviev,
  • D. Rakonjac,
  • L. M. Mannapova

摘要

The capabilities of the pilot classifier model trained to recognize traces of microbial activity on solid surfaces identifying the formation of soil-like bodies were preliminarily assessed. A database of 500 samples described by the authors and taken in open sources from 1988 to the present time was collected for machine learning; among them, 59 samples were soil horizons, 146 samples represented parent rocks and soil-like bodies; there were also rock-forming minerals, accompanying components of soil formation, and xenobiotics common in technogenically transformed landscapes of the world. The samples included in the database differed in their size, coverage with biofilms and films of other nature, and chemical and physical processing. The array of sample features significant for machine learning included quantiles of the wetting contact angle distribution and generalized categorical indicators of surface geometry, mineral composition, and state of organic matter. The target function of the classification was the presence of stable traces of microbial activity on a solid surface. Missing data were reconstructed using Monte Carlo procedure and bootstrapping. As a result of numerical experiments on optimizing the machine learning, a balanced training dataset containing 1233 pseudo-sample elements was obtained. Six classifier models with parameter variations were trained and evaluated. The most productive classifier—a five-layer neural network with randomly dropout neurons—demonstrated the prediction accuracy of 0.74 and the ROC AUC of 0.80 on the test sample, which is higher than that of simpler and faster classifiers (accuracy and ROC AUC of 0.70). Based on the disagreement between the classifications of a human expert and a trained algorithm, common features of samples that are difficult for machine classification were established: traces of life activity, carbonates, and high degree of dispersion. This is important for determining the direction of collecting information to improve the performance of the classifier. The development of an algorithm for recognizing traces of microbial activity is useful for clarifying the mechanisms of biogeochemical and biogeotechnological processes in soils of various origins, including soil formation and terraforming.