연속 변수의 이산화

Python으로 배우는 예측 분석 입문

Nele Verbiest, Ph.D

Data Scientist @PythonPredictions

Python에서의 이산화

variable = "max_gift"
number_bins = 3
basetable["disc_max_gift"] = pd.qcut(basetable[variable], number_bins)

basetable.groupby("disc_max_gift").size()

disc_mean_gift
[2, 84.25]         33330
(84.25, 96.833]    33330
(96.833, 197]      33330
dtype: int64

이산화해야 할 변수

variables_model = ["income_average","mean_gift","gender_M","min_gift","age"]
def check_discretize(basetable, variable, threshold):
    return(len(basetable.groupby(variable))>threshold)

check_discretize(basetable, "mean_gift",5)

True

check_discretize(basetable, "income_average",5)

False

모든 변수의 이산화

variables_model = ["income_average","mean_gift","gender_M","min_gift","age"]
def check_discretize(basetable, variable, threshold):
    return(len(basetable.groupby(variable))>threshold)

threshold = 5
number_bins = 5
for variable in variables_model:
    if check_discretize(basetable, variable, threshold):
        new_variable = "disc" + variable
        basetable[new_variable] = pd.qcut(basetable[variable], number_bins)

명확한 구간 설정

basetable["disc_age"] = pd.qcut(basetable["age"], 5)
basetable["disc_age"].unique()

[(38, 49], (68, 110], [19, 38], (49, 59], (59, 68]]

basetable["disc_age"] = pd.cut(basetable["age"],[18,30,40,50,60,110])
basetable.groupby("disc_age").size()

disc_age
(18, 30]     10017
(30, 40]     14448
(40, 50]     19002
(50, 60]     24684
(60, 110]    31849

연습해 봅시다!

Python으로 배우는 예측 분석 입문