An approach for total organic carbon prediction using convolutional neural networks optimized by differential evolution

Rodrigo Oliveira Silva, Camila Martins Saporetti, Zaher Mundher Yaseen*, Egberto Pereira, Leonardo Goliatt*

*Corresponding author for this work

Research output: Contribution to journalArticlepeer-review


One of the most reliable indicators of the amount and condition of organic matter in a basin is the total organic carbon (TOC) content of rock samples. The manual estimation primarily involves inspecting rock samples for organic carbon analysis. However, this method is costly and time-consuming due to the necessity of collecting samples from a wide range of well intervals within the source rocks. Consequently, studies have been conducted to facilitate this process. Emerging alternative methods for estimating TOC involve utilizing well logs and stratigraphic analysis data in conjunction with machine learning (ML) algorithms. Recent studies advocate for the application of ML algorithms in estimating TOC. The model parameters were selected using a metaheuristics approach and cross-validation to enhance the model’s flexibility. This computational technique enables the identification of models with greater generalization potential. Convolutional neural networks (CNNs), extreme learning machines, Elastic Net, and Extreme Gradient Boosting (XGB) were employed for the estimation of TOC. The suggested data intelligent framework was validated using samples from various sedimentary basins. In several metrics investigations, the CNN model exhibits notable distinctions against the other models, highlighting its potential to assist geologists in predicting concentrations of total organic carbon.

Original languageEnglish
Pages (from-to)20803-20817
Number of pages15
JournalNeural Computing and Applications
Issue number28
StatePublished - Oct 2023

Bibliographical note

Publisher Copyright:
© 2023, The Author(s), under exclusive licence to Springer-Verlag London Ltd., part of Springer Nature.


  • Convolutional neural network
  • Differential evolution
  • Geology
  • Total organic carbon

ASJC Scopus subject areas

  • Software
  • Artificial Intelligence


Dive into the research topics of 'An approach for total organic carbon prediction using convolutional neural networks optimized by differential evolution'. Together they form a unique fingerprint.

Cite this