def create_stats(x_train, x_test, y_train, y_test, enc="labelencoder"):
#Call basic model (Linear, Lasso and Ridge)
lm_model, lm_y_pred, lm_stats = linear_model(x_train, x_test, y_train,
y_test, 0.01)
la_model, la_y_pred, la_stats = lasso(x_train,
x_test,
y_train,
y_test,
alpha=0.1)
ri_model, ri_y_pred, ri_stats = ridge(x_train,
x_test,
y_train,
y_test,
alpha=0.1)
#Filter Best feature using K percentile
feat_sel = feature_selection(x_train, y_train, k=50)
x_train = x_train[feat_sel]
x_test = x_test[feat_sel]
#Call basic model with selected features only (Linear, Lasso and Ridge)
lm_model_fs, lm_y_pred_fs, lm_stats_fs = linear_model(
x_train, x_test, y_train, y_test, 0.01)
la_model_fs, la_y_pred_fs, la_stats_fs = lasso(x_train,
x_test,
y_train,
y_test,
alpha=0.1)
ri_model_fs, ri_y_pred_fs, ri_stats_fs = ridge(x_train,
x_test,
y_train,
y_test,
alpha=0.1)
#Concate the returned response.
complete_stats = pd.concat(
[lm_stats, lm_stats_fs, la_stats, la_stats_fs, ri_stats, ri_stats_fs])
complete_stats.index = [
'lm_score', 'lm_features_score', 'la_score', 'la_features_score',
'ri_score', 'ri_features_score'
]
#Rmse and mse for certain models are NaN. need to fill with zeros.
complete_stats[['rmse', 'mse']] = complete_stats[['rmse', 'mse']].fillna(0)
complete_stats.mse += complete_stats.rmse
complete_stats = complete_stats.drop(['rmse'], axis=1)
return complete_stats
def create_stats(X_train, X_test, y_train, y_test, enc="labelencoder"):
_, _, linear_model_scores = linear_model(X_train, X_test, y_train, y_test,
0.01)
_, _, lasso_scores = lasso(X_train, X_test, y_train, y_test)
_, _, ridge_scores = ridge(X_train, X_test, y_train, y_test)
selected_features = percentile_k_features(X_train, y_train, k=50)
X_train_features = X_train[selected_features]
X_test_features = X_test[selected_features]
_, _, linear_model_scores_features = linear_model(X_train_features,
X_test_features, y_train,
y_test, 0.01)
_, _, lasso_scores_features = lasso(X_train_features, X_test_features,
y_train, y_test)
_, _, ridge_scores_features = ridge(X_train_features, X_test_features,
y_train, y_test)
complete_stats = pd.concat([
linear_model_scores,
linear_model_scores_features,
#chain_stats_with_features_selection ,stats_chain,
lasso_scores,
lasso_scores_features,
ridge_scores,
ridge_scores_features
])
complete_stats.index = [
'linear_model_scores',
'linear_model_scores_features',
#chain_stats_with_features_selection ,stats_chain,
'lasso_scores',
'lasso_scores_features',
'ridge_scores',
'ridge_scores_features'
]
#complete_stats.columns=['Name', 'cross_validation','rmse','mae','r2']
complete_stats.mse = complete_stats.mse.fillna(0)
complete_stats.rmse = complete_stats.rmse.fillna(0)
complete_stats.rmse = complete_stats.mse + complete_stats.rmse
complete_stats = complete_stats.drop(['mse'], axis=1)
return complete_stats
def create_stats(X_train, X_test, y_train, y_test, enc="labelencoder"):
a, b, lm_score = linear_model(X_train, X_test, y_train, y_test, '')
c, d, lasso_score = lasso(X_train, X_test, y_train, y_test)
e, f, ridge_score = ridge(X_train, X_test, y_train, y_test)
best_features = feature_selection(X_train, y_train, k=50)
a, b, lm_score_bf = linear_model(X_train[best_features],
X_test[best_features], y_train, y_test,
'')
c, d, lasso_score_bf = lasso(X_train[best_features], X_test[best_features],
y_train, y_test)
e, f, ridge_score_bf = ridge(X_train[best_features], X_test[best_features],
y_train, y_test)
complete_stats = pd.concat([
lm_score, lasso_score, ridge_score, lm_score_bf, lasso_score_bf,
ridge_score_bf
],
ignore_index=True)
del complete_stats['rmse']
return complete_stats #lm_score,lasso_score,ridge_score,lm_score_bf,lasso_score_bf,ridge_score_bf