Recent advancements in machine learning (ML) have paved the way for its integration into mapping studies, aiming to refine the accuracy of predictive models. Researchers are increasingly turning to ML to develop mapping algorithms that bridge various datasets, comparing their effectiveness against traditional regression models (RMs).
Methodological Approaches in Mapping Studies
A comprehensive systematic review was conducted, encompassing 12 databases up to December 2023, to uncover studies where ML techniques were employed for mapping algorithm development. Data extraction focused on multiple facets, including dataset attributes, source and target measures, and the specific ML and RM methodologies utilized. The analysis differentiated between direct and indirect mapping types and evaluated the goodness-of-fit using metrics such as MAE, MSE, RMSE, R-squared, and ICC.
Performance Comparison Between ML and Regression Models
Thirteen mapping studies incorporated both ML and RM approaches. Bayesian networks emerged as the predominant ML technique, used in half of the studies, while LASSO featured in a third. Among RMs, the ordinary least squares model was most common. The findings indicated that ML methods offered modest improvements in goodness-of-fit metrics over RMs, with R-squared values seeing the most notable enhancement.
- ML techniques showed an average increase of 0.058 in R-squared compared to RMs.
- Bayesian networks and LASSO were the most effective ML approaches in the reviewed studies.
- The marginal gains in MAE and MSE suggest limited superiority of ML over traditional methods.
- Interpretation and external validation of ML models remain significant challenges.
The incremental benefits of ML in mapping algorithms highlight its potential, yet also underscore the necessity for more robust validation and interpretation frameworks to fully leverage these technologies.
As the adoption of machine learning in mapping studies grows, researchers must address the subtleties in model performance and application. Ensuring that ML models are not only accurate but also interpretable and generalizable will be crucial for their sustained integration into the field. Future research should focus on enhancing the transparency of ML algorithms and establishing standardized validation protocols to better assess their true advantages over traditional regression approaches.
This article has been prepared with the assistance of AI and reviewed by an editor. For more details, please refer to our Terms and Conditions. We do not accept any responsibility or liability for the accuracy, content, images, videos, licenses, completeness, legality, or reliability of the information contained in this article. If you have any complaints or copyright issues related to this article, kindly contact the author.
