Genetic algorithm to optimize the SVM and K-means algorithms for mapping of mineral prospectivity
- Publisher:
- SPRINGER LONDON LTD
- Publication Type:
- Journal Article
- Citation:
- Neural Computing and Applications, 2023, 35, (1), pp. 719-733
- Issue Date:
- 2023-01-01
Closed Access
| Filename | Description | Size | |||
|---|---|---|---|---|---|
| Genetic algorithm to optimize the SVM and K-means algorithms for mapping of mineral prospectivity.pdf | Published version | 17.11 MB |
Copyright Clearance Process
- Recently Added
- In Progress
- Closed Access
This item is closed access and not available.
Unsupervised clustering (e.g., K-means) and supervised machine learning [e.g., support vector machines (SVMs)] methods can be used in data-driven classification and predictive mapping of mineral prospectivity. In this study, the conceptual model of porphyry-type Cu deposits in Varzaghan district, Northwest Iran, was utilized to translate mineralization-related processes into mappable targeting criteria, resulting in six evidence layers derived from geochemical, geological, structural and remote sensing data. Then, traditional K-means clustering (KMC) method and a supervised SVM were employed to construct GIS-based mineral prospectivity maps. One of the challenging issues for generation of mineral prospectivity maps using clustering algorithms is to select suitable cluster centers (called centroids) in order to circumscribe the similar spatial features into meaningful clusters. Machine learning methods are strongly sensitive to hyperparameter values that contribute to prediction, as the prediction accuracy can significantly enhance when the optimized hyperparameters are calibrated to training procedure. To reach these goals, a genetic algorithm was incorporated into K-means and SVM in order to automatically select the optimized cluster centroids and tuned hyperparameters, respectively, and two new prospectivity models namely GKMC and genetic-based SVM were then generated. To evaluate the performance accuracy in training procedures of SVM and genetic-based SVM, K-fold cross-validation and confusion matrix were employed. Moreover, success-rate curves were plotted to compare the overall performance of unsupervised clustering models comprising KMC and GKMC and also supervised machine learning models comprising SVM and genetic-based SVM for detection of favorable areas associated with porphyry-type Cu mineralization. The results suggested the superiority of genetic-based SVM model over other models.
Please use this identifier to cite or link to this item: