Converting Labels to Indicator Matrix After Dividing a Dataset: Best Practices for Machine Learning