def rf_grid(df_train):
# Run random forest, find best parameters
RF = nfl.rf(df_train, 'Pos')
params = {
"n_estimators": range(1, 420, 20),
"criterion": ["gini", "entropy"]
}
df_cv = nfl.grid_search(df_train, RF, 'Pos', params, save=True)
nfl.display_parameter_curve(df_cv, 'param_n_estimators', 'mean_test_score',
'# Estimators', 'Mean Test Score',
'param_criterion')
def rf_grid_maxfeature(df_train):
# Run grid search multiple times for max features
RF = nfl.rf(df_train, 'Pos')
params = {
"n_estimators": [220],
"criterion": ["gini"],
"max_features": range(1, df_train.shape[1] - 1, 1)
}
for i in range(0, 10):
print(i)
try:
df_cv = df_cv.append(
nfl.grid_search(df_train, RF, 'Pos', params, save=False))
except:
df_cv = nfl.grid_search(df_train, RF, 'Pos', params, save=False)
df_cv = df_cv.groupby(['param_max_features']).mean().reset_index()
nfl.display_parameter_curve(df_cv, 'param_max_features', 'mean_test_score',
'n_estimators', 'Mean Test Score')
nfl.display_parameter_curve(df_cv, 'param_max_features',
'mean_train_score', 'n_estimators',
'Mean Train Score')
def svm_grid(df_train):
# Run support vector machine, find best parameters
SVM = nfl.svm(df_train, 'Pos')
params = {
"C": np.power(10.0, np.arange(-6, 5, 1)),
"kernel": ["linear", "rbf"]
}
df_cv = nfl.grid_search(df_train, SVM, 'Pos', params, save=True)
nfl.display_parameter_curve(df_cv, 'param_C', 'mean_test_score',
'Penalty Param', 'Mean Test Score',
'param_kernel', True)
nfl.display_parameter_curve(df_cv, 'param_C', 'mean_train_score',
'Penalty Param', 'Mean Train Score',
'param_kernel', True)
def dt_grid(df_train):
# Run decision tree, find best parameters
DT = nfl.decision_tree(df_train, 'Pos')
params = {
"min_samples_leaf": range(1, 40, 1),
"criterion": ["gini", "entropy"]
}
df_cv = nfl.grid_search(df_train, DT, 'Pos', params, save=True)
nfl.display_parameter_curve(df_cv, 'param_min_samples_leaf',
'mean_test_score', 'Min Samples Leaf',
'Mean Test Score', 'param_criterion')
nfl.display_parameter_curve(df_cv, 'param_min_samples_leaf',
'mean_train_score', 'Min Sample Leaf',
'Mean Train Score', 'param_criterion')
def rf_grid_est(df_train):
# Run grid search multiple times for RT n_estimators
RF = nfl.rf(df_train, 'Pos')
params = {
"n_estimators": range(1, 420, 20),
"criterion": ["gini", "entropy"]
}
for i in range(0, 1):
print(i)
try:
df_cv = df_cv.append(
nfl.grid_search(df_train, RF, 'Pos', params, save=False))
except:
df_cv = nfl.grid_search(df_train, RF, 'Pos', params, save=False)
df_cv = df_cv.groupby(['param_n_estimators']).mean().reset_index()
nfl.display_parameter_curve(df_cv, 'param_n_estimators', 'mean_test_score',
'# Estimators', 'Mean Test Score',
'param_criterion')
nfl.display_parameter_curve(df_cv, 'param_n_estimators',
'mean_train_score', '# Estimators',
'Mean Train Score', 'param_criterion')