Missforest fit_transform 缺失值插补耗时过长

Question

我这里有一个数据集，但是文件太大了，我不得不把它分成多个“output.csv”文件。上传到这个 github 上的部分大约是整个数据集的三分之一，但它可以让您了解数据：https://github.com/chongochoo/dataset

我正在尝试估算“suspect_age”的缺失值。我准备数据框进行插补，但是当我调用它时：

# Miss Forest
imputer = MissForest() #miss forest
X_imputed = imputer.fit_transform(df_fitted)
X_imputed = pd.DataFrame(X_imputed, columns = df_fitted.columns).round(1)

对 fit_transform() 的调用花费了非常非常长的时间，我什至无法到达获得估算值的地步。不知道如何解决这个问题，因为我不认为我的数据集如此复杂，实际上它是那些 Kaggle 玩具数据集之一。

这里是总代码：

import pandas as pd
import matplotlib.pyplot as plt
import numpy as np
import pandas as pd
import numpy as np
import matplotlib.pyplot as plt
import seaborn as sns
import random
#sys.modules['sklearn.neighbors.base'] = sklearn.neighbors._base
import sklearn.neighbors._base
import sys
sys.modules['sklearn.neighbors.base'] = sklearn.neighbors._base
#from sklearn.neighbors import  KNeighborsClassifier
from missingpy import MissForest
#from sklearn.impute import KNNImputer


df = pd.read_csv('C:/Users/Connie/PycharmProjects/507-FinalProject/data-preprocessed10.csv', sep=',')

#print(100*(df.isnull().sum())/len(df.index))


# drop all columns no info
df_fitted = df.drop(['incident_id','address','incident_url','source_url','incident_url_fields_missing',
                     'congressional_district', 'gun_stolen', 'gun_type', 'incident_characteristics',
                     'latitude', 'longitude', 'location_description', 'notes', 'participant_age',
                     'participant_age_group', 'participant_gender', 'participant_name', 'participant_relationship',
                     'participant_status', 'participant_type', 'sources', 'state_house_district',
                     'state_senate_district', 'zipcode', 'city_or_county', 'suspect_gender'],axis=1)

df_fitted['year'] = ''
df_fitted['month'] = ''
df_fitted['day'] = ''

for i,row in df_fitted.iterrows():
    date_parts = str(row['date']).split('-')
    df_fitted.at[i, 'year'] = date_parts[0]
    df_fitted.at[i, 'month'] = date_parts[1]
    df_fitted.at[i, 'day'] = date_parts[2]

df_fitted = df_fitted.drop(columns=['date'])
#df_fitted['suspect_gender'] = df_fitted.suspect_gender.map({'Male':0, 'Female':1, 'Gender Unknown':2})
df_fitted['suspect_age'] = df_fitted.suspect_age.map({'Adult 18+':0, 'Teen 12-17':1, 'Child 0-11':2, '':3})
df_fitted = pd.get_dummies(df_fitted, columns=['state'], drop_first=True)


# Miss Forest
imputer = MissForest() #miss forest
X_imputed = imputer.fit_transform(df_fitted)
X_imputed = pd.DataFrame(X_imputed, columns = df_fitted.columns).round(1)


#df_transform = imputer.fit_transform(df_fitted[['suspect_age']])

print(X_imputed.head(5))