feature_selection_univariate_testing.py

# import packages
import pandas as pd

from sklearn.feature_selection import SelectKBest, f_regression


# import data
my_df = pd.read_csv("feature_selection_sample_data.csv")


# Regression Template

X = my_df.drop(["output"], axis = 1)
y = my_df["output"]

feature_selector = SelectKBest(f_regression, k = "all")
fit = feature_selector.fit(X,y)

p_values = pd.DataFrame(fit.pvalues_)
scores = pd.DataFrame(fit.scores_)
input_variable_names = pd.DataFrame(X.columns)
summary_stats = pd.concat([input_variable_names, p_values, scores], axis = 1)
summary_stats.columns = ["input_variable", "p_value", "f_score"]
summary_stats.sort_values(by = "p_value", inplace = True)

p_value_threshold = 0.05
score_threshold = 5

selected_variables = summary_stats.loc[(summary_stats["f_score"] >= score_threshold) &
                                       (summary_stats["p_value"] <= p_value_threshold)]
selected_variables = selected_variables["input_variable"].tolist()
X_new = X[selected_variables]



 
# Classification Template

from sklearn.feature_selection import SelectKBest, chi2

X = my_df.drop(["output"], axis = 1)
y = my_df["output"]

feature_selector = SelectKBest(chi2, k = "all")
fit = feature_selector.fit(X,y)

p_values = pd.DataFrame(fit.pvalues_)
scores = pd.DataFrame(fit.scores_)
input_variable_names = pd.DataFrame(X.columns)
summary_stats = pd.concat([input_variable_names, p_values, scores], axis = 1)
summary_stats.columns = ["input_variable", "p_value", "chi2_score"]
summary_stats.sort_values(by = "p_value", inplace = True)

p_value_threshold = 0.05
score_threshold = 5

selected_variables = summary_stats.loc[(summary_stats["chi2_score"] >= score_threshold) &
                                       (summary_stats["p_value"] <= p_value_threshold)]
selected_variables = selected_variables["input_variable"].tolist()
X_new = X[selected_variables]