This guide provides a practical approach for beginners to use Support Vector Machines (SVM) for classification. It outlines a simple procedure that usually gives reasonable results. SVM is a powerful technique for classification, but beginners often struggle with it due to the complexity of the underlying theory. The guide aims to help novices achieve acceptable results quickly and easily without requiring a deep understanding of the theory. The guide begins by explaining the basics of SVM, including the optimization problem that needs to be solved. It then introduces the four basic kernels used in SVM: linear, polynomial, radial basis function (RBF), and sigmoid. The guide recommends starting with the RBF kernel, as it is generally a good first choice and can handle nonlinear relationships between class labels and attributes. Data preprocessing is crucial for SVM. The guide emphasizes the importance of scaling data to avoid numerical issues and to ensure that attributes with larger numeric ranges do not dominate those with smaller ranges. It also discusses how to handle categorical features by converting them into numeric data. The guide then discusses model selection, focusing on the RBF kernel and the use of cross-validation to find the best parameters C and γ. It recommends a grid search approach for parameter selection, which involves trying different combinations of C and γ and selecting the one with the best cross-validation accuracy. The guide also provides examples of the proposed procedure, showing how it can improve accuracy on real-world datasets. It includes detailed steps for using the LIBSVM software to train and test models, as well as how to use an automatic script to perform the entire process. Finally, the guide discusses when to use the linear kernel instead of the RBF kernel, particularly when the number of features is large. It also compares the performance of LIBSVM and LIBLINEAR for different types of data, highlighting the advantages of LIBLINEAR for large-scale problems.This guide provides a practical approach for beginners to use Support Vector Machines (SVM) for classification. It outlines a simple procedure that usually gives reasonable results. SVM is a powerful technique for classification, but beginners often struggle with it due to the complexity of the underlying theory. The guide aims to help novices achieve acceptable results quickly and easily without requiring a deep understanding of the theory. The guide begins by explaining the basics of SVM, including the optimization problem that needs to be solved. It then introduces the four basic kernels used in SVM: linear, polynomial, radial basis function (RBF), and sigmoid. The guide recommends starting with the RBF kernel, as it is generally a good first choice and can handle nonlinear relationships between class labels and attributes. Data preprocessing is crucial for SVM. The guide emphasizes the importance of scaling data to avoid numerical issues and to ensure that attributes with larger numeric ranges do not dominate those with smaller ranges. It also discusses how to handle categorical features by converting them into numeric data. The guide then discusses model selection, focusing on the RBF kernel and the use of cross-validation to find the best parameters C and γ. It recommends a grid search approach for parameter selection, which involves trying different combinations of C and γ and selecting the one with the best cross-validation accuracy. The guide also provides examples of the proposed procedure, showing how it can improve accuracy on real-world datasets. It includes detailed steps for using the LIBSVM software to train and test models, as well as how to use an automatic script to perform the entire process. Finally, the guide discusses when to use the linear kernel instead of the RBF kernel, particularly when the number of features is large. It also compares the performance of LIBSVM and LIBLINEAR for different types of data, highlighting the advantages of LIBLINEAR for large-scale problems.