Okhowvata, H.R., Riahib, M.A. and Abedi, M.M., 2022. The effect of supervised feature extraction techniques on the facies classification using machine learning. Journal of Seismic Exploration, 31: 563-577. The widely accepted supervised machine learning classification algorithms are used for the semi-automating of the feature extraction process. In the machine learning facies classification process, each wireline log is a feature in the feature space. Since features are important in classification decisions, using suitable features improves the performance of a classification algorithm. Three feature sets were compared containing the original conventional features (well-logs), and the extracted features from the unsupervised PCA and supervised FDA methods, using two classifiers algorithms, namely SVM and RF. The FDA showed that improvement in the performance of facies classifiers while PCA can even deteriorate the results. An Fl score of 0.61 averaged over the available 20 folds for the combination of FDA feature extractor and RF classifier is achieved. This represents about a 5% improvement in the prediction accuracy, compared to the conventional use of wells information as features. Moreover, the Fl score was achieved without the usage of feature extraction. This value is 0.56 by using all 7 conventional features (well-logs), thus 5 percent lower than using FDA with only 3 first features.