from sklearn.preprocessing import StandardScaler
from sklearn.impute import SimpleImputer
from sklearn.decomposition import PCA
from sklearn.feature_selection import RFE
from sklearn.ensemble import RandomForestRegressor
from sklearn.linear_model import LinearRegression

def select_features_pca(df, n_components):
    df_numeric = df.select_dtypes(include=[float, int])
    
    imputer = SimpleImputer(strategy='mean')
    df_imputed = imputer.fit_transform(df_numeric)
    
    scaler = StandardScaler()
    df_scaled = scaler.fit_transform(df_imputed)
    
    pca = PCA(n_components=n_components)
    components = pca.fit_transform(df_scaled)
    
    feature_names = df_numeric.columns
    important_features = feature_names[:n_components]
    
    return df[important_features], important_features

def select_features_rfe(df, n_features_to_select):
    df_numeric = df.select_dtypes(include=[float, int])
    
    imputer = SimpleImputer(strategy='mean')
    df_imputed = imputer.fit_transform(df_numeric)
    
    scaler = StandardScaler()
    df_scaled = scaler.fit_transform(df_imputed)
    
    model = LinearRegression()
    rfe = RFE(estimator=model, n_features_to_select=n_features_to_select)
    rfe.fit(df_scaled, df_scaled[:, 0])
    
    important_features = [df_numeric.columns[i] for i in range(len(rfe.support_)) if rfe.support_[i]]
    
    return df[important_features], important_features

def select_features_rf(df, n_features_to_select):
    df_numeric = df.select_dtypes(include=[float, int])
    
    imputer = SimpleImputer(strategy='mean')
    df_imputed = imputer.fit_transform(df_numeric)
    
    scaler = StandardScaler()
    df_scaled = scaler.fit_transform(df_imputed)
    
    model = RandomForestRegressor()
    model.fit(df_scaled, df_scaled[:, 0])
    
    importances = model.feature_importances_
    indices = importances.argsort()[-n_features_to_select:][::-1]
    
    important_features = [df_numeric.columns[i] for i in indices]
    
    return df[important_features], important_features