*3.4. Training and Testing of SVM 4*

The training and testing of SVM 4 were purposed to categorize between T1 and T2 fault classes. SVM 4 was trained by a total of 8 data samples with a T1 fault condition and a total of 7 data samples with a T2 fault condition. Consequently, a total of 10 data samples were used in testing SVM 4 and the results showed that 9 data samples were classified fittingly. The complete BCSVM was tested using 26 DGA data samples, and 24 samples were pigeonholed rightly. The precision of the overall BCSVM was 92%. The selection of responses and predictors selected by SVM 4 using the ingested data are demonstrated in Figure 14.


**Figure 14.** The dataset variables were selected by the SVM 4 classifier.

The dissolved gases were automatically placed as the predictors and the fault type as the response. In the SVM 3 algorithm, H2 and CH4 were considered to acquire the desired output. The characterization of H2 versus CH4 is illustrated in Figure 15.

**Figure 15.** H2 versus CH4. Blue and orange (•)—raw data, orange and blue (×)—predicted data.

Further, the ML algorithms were compared in terms of accuracy (%), prediction speed (objects/sec) and training time (sec), as tabulated in Table 7.


**Table 7.** SVM 4 DGA classification outcomes.

In the SVM 4 algorithm, quadratic SVM, fine Gaussian SVM, and medium Gaussian SVM provided the highest degree of accuracy and the corresponding configuration matrix shown in Figure 16 was imported to the workspace to predict the new testing data samples.

**Figure 16.** Confusion matrix for SVM 4.
