Characteristics prediction of hydrothermal biochar using data enhanced interpretable machine learning.

Area of Science:

• Environmental Science
• Soil Science
• Materials Science

Background:

• Hydrothermal biochar shows potential for sustainable soil remediation and enhancing plant growth.
• Varied soil environments necessitate tailored biochar properties, posing prediction challenges.
• Existing machine learning models for biochar property prediction often lack interpretability.

Purpose of the Study:

• To develop accurate and interpretable machine learning models for predicting hydrothermal biochar properties.
• To enhance prediction accuracy through data augmentation techniques.
• To identify critical factors influencing biochar property predictions.

Main Methods:

• Established and evaluated data-enhanced machine learning models, including support vector machine, artificial neural network, and random forest.
• Conducted sensitivity analysis to determine the influence of various parameters on model predictions.
• Utilized data enhancement to improve model performance and overcome interpretability issues.

Main Results:

• Data enhancement led to an average accuracy increase of 5.8% to 15.8% compared to traditional models.
• The optimal random forest model achieved an average accuracy of 94.89%.
• Sensitivity analysis identified reaction temperature, reaction pressure, and specific biomass feedstock elements as crucial predictors.

Conclusions:

• Data-enhanced interpretable machine learning offers a promising approach for predicting hydrothermal biochar characteristics.
• Accurate prediction of biochar properties can facilitate its effective application in soil remediation.
• Understanding key influencing factors aids in optimizing biochar production for specific soil conditions.