Table 1 Overview of previous research.

1

Tavakolian et al. (2024)¹

Prediction of CH₄ and CO₂ Adsorption Capacity in Tight Reservoirs

Utilization of ML Methods Including RF and Hyperparameter Tuning with Optuna

The RF method demonstrated the best performance for predicting adsorption capacity.

CH₄: MAE = 0.0864, RMSE = 0.1520; CO₂: MAE = 0.0529, RMSE = 0.2308

2

Zhou et al. (2024)³³

Modeling of CH₄ Adsorption in Shale Using GPR

Development of GPR Model and Comparison with XGBoost

GPR was the most accurate method; TOC was the most influential variable.

Reduction of prediction error to less than 3%

3

Wang et al. (2024)³⁴

Prediction of Competitive CO₂-CH₄ Adsorption in Shale Porous Media

Integration of Molecular Simulation, Boltzmann Network, and ANN

Computational limitations were addressed; the impact of mineral type was examined.

-

4

Alanazi et al. (2023)³⁶

Prediction of CO₂ Adsorption Capacity in Coal

ML Models Including RF, ANN, and ANFIS

RF provided the most accurate predictions.

Low RMSE and AAPE at high pressures

5

Amar et al. (2022)³⁸

Modeling of CH₄ Adsorption Capacity in Shale

Application of GEP and GMDH

GEP was more precise with mathematical relationships.

R² = 0.9837; Moisture has a greater impact than TOC

6

Kalam et al. (2023)³⁷

Prediction of Hydrogen Adsorption in Shale

Use of Gradient Boosted Regression

Data-driven models were more accurate and faster.

Coefficient of determination: 99.6% (training), 94.6% (testing)

7

Meng et al. (2020)³⁹

Prediction of CH₄ Adsorption for Shale Production Planning

Comparison of ML with Classical Models

XGBoost showed the best performance.

Accurate predictability for TOC, temperature, and moisture

8

Alqahtani et al. (2024)³⁵

Prediction of CH₄ and CO₂ Adsorption in Shale and Coal Reservoirs

Optimized GRNN, RBFNN, and CatBoost Models

CatBoost-GWO was the most accurate model.

CO₂: RMSE = 0.1229; CH₄: RMSE = 0.0681