Figure 3: Unsupervised learning techniques (e.g., hierarchical clustering, tree classification) can be used for clustering analysis. This hierarchical clustering was constructed based on Euclidean distance (between T1 relaxation and T2relaxation) and its´ quantitative linkages (e.g., inter/intra cluster similarity) shown in a heat map. (A, A´, B, B´, C, C´, C´´, D) were the variants of the same oil content taken from different manufacturers. Tree classification method is shown for comparison (Supp. Fig. 2).
Using unsupervised learning, the relationship between each objects were rapidly constructed using clustering analysis (e.g., tree classification, hierarchical clustering) and its´ quantitative linkages (e.g., inter/intra cluster similarity) were shown on a dendogram and a heat map (Fig. 3). Supervised learning models (i.e., neural network, kNN, logistic regression, naïve Bayes, and random forest) can be used to train the datasets and the best model with the highest accuracy can be chosen to predict the object classification (e.g., oil classification, infection/non-infection) using pre-trained datasets (Fig. 4 and Table 1).