Categorical variables contain distinct data values to represent different groups. Some machine learning algorithms, for example, decision trees, are designed to handle categorical data structure, however, a majority of machine learning algorithms, including support vector machine (SVM), neural network (NN) only process numeric data. Therefore, the categorical variables require conversion before our data training process. The one-hot encoding can transform the categorical variables into binary columns, which represent one category in one column.  In python pandas package, the one-hot encoding can be attained using the function "get_dummies".