To implement logistic regression in python with the scikit learn package, there are two logistic regression function classifiers namely, LogisticRegression and LogisticRegressionCV. For both of the two classifiers, we can opt for optimizers such as liblinear, newton-cg, and sag of lbfgs and regularization methods supporting the selected optimizer. The liblinear solver supports both L1 and L2 regularization with dual formulation only for the L2 penalty. However, the different between the two classifier is the best hyperparameter of LogisticRegressionCV can be selected by by the cross-validator StratifiedKFold, or it can be altered using the cv parameter. The results of both algorithms with different setups are shown below. In order to have the logloss and percent accuracy, we use the training set and divide it into 5 folds, and use 5-fold cross validation to validate each model. Moreover, we also test the model with the test set data to find which model can still predict the unseen data with the lowest logloss.