Dear Wassim Diai,

In its most simple type SVM is applied on binary classification, dividing data points either in 1 or 0. your case is a multiclass classification, for which, the same principle is utilized. The multiclass problem is broken down into multiple binary classification cases, which is also called one-vs-one. for instance, in Python, scikit-learn one-vs-one is not the default case and needs to be selected explicitly using the parameters. One-vs-rest is set as default. It basically divides the data points in class x and rest. Consecutively a certain class is distinguished from all other classes.

Please take a look at, https://www.kaggle.com/pranathichunduru/svm-for-multiclass-classification

You can set a threshold. If the classification result is over the threshold, you can think it is right. Additionally, you can use cross-validation method to test the accuracy of classification reslut.

You can refer:

https://www.baeldung.com/cs/svm-multiclass-classification

Moreover, can design an optimized model by finding the optimal values of hyperparameters used in SVM and compare the performance.

Error correcting output code based multiclass SVM using the different coding scheme

You can use Error correcting output code based multiclass SVM. If you use Matlab, there is built-in function that does the job

https://www.baeldung.com/cs/svm-multiclass-classification