Mutual Information

mutual_information.py

import matplotlib.pyplot as plt
import numpy as np
import pandas as pd
import seaborn as sns
from sklearn.feature_selection import mutual_info_regression

# Set Matplotlib defaults
plt.style.use("seaborn-whitegrid")
plt.rc("figure", autolayout=True)
plt.rc(
    "axes",
    labelweight="bold",
    labelsize="large",
    titleweight="bold",
    titlesize=14,
    titlepad=10,
)

#X는 DataFrame
#y는 Series
def make_mi_scores(X, y):
    X = X.copy()
    for colname in X.select_dtypes(["object", "category"]):
        X[colname], _ = X[colname].factorize()#ordered encoding 수행, codes(=결과 ndarray), unique 반환 
        #unique는 컬럼의 dType에 따라 unique value들을 담는 클래스 type을 다르게 줌, 
        #Categorical 일땐 Categorical 다른panda object일땐 Index, 나머지는 ndarray
    discrete_features = [pd.api.types.is_integer_dtype(t) for t in X.dtypes] 
    #pd.api.types.is_integer_dtype() 입력 dtype(혹은 string alias), array-like가  NumPy와 Pandas의 integer의 확장 타입일때만 True
    mi_scores = mutual_info_regression(X, y, discrete_features=discrete_features, random_state=0)#feature-target간의 mutual_info_score반환
    mi_scores = pd.Series(mi_scores, name="MI Scores", index=X.columns)#index로 column이름을 씌워줌
    mi_scores = mi_scores.sort_values(ascending=False)#score를 내림차순으로 나열.
    return mi_scores

def plot_mi_scores(scores):
    scores = scores.sort_values(ascending=True)#score를 오름차순으로 나열. plot은 가장 앞선 원소가 밑에 오기때문에.
    y_space = np.arange(len(scores))
    ticks = list(scores.index)
    plt.barh(y_space, scores) #수평바그리기
    plt.yticks(y_space, ticks) #수평바 라벨링하기
    plt.title("Mutual Information Scores")