Laboratory with random forest, logistic regression and SVM.
The dataset used for this test is a set of points generated randomly with the following specification: • Number of Samples: 1200 • Number of Classes: 3 • Number of Features: 2 (Length and Width).
Comparing all the accuracies and classification reports of the three trained models, it is clear that the Random Forest is the best training method to the dataset chosen. Even though precision and recall are high in this method, not giving a balance. The F1-score is considered perfect when it is 1, and this is the case for the Random Forest algorithm.
Check out MNIST-database-TrainingModels repository if you are enjoying the laboratory reading.