■suumoの物件分析 #5 「賃貸のパラメータの重要度を調べてみる ~東京編~」

from sklearn.model_selection import train_test_split
from sklearn import linear_model
from sklearn.metrics import mean_absolute_error as MAE
import numpy as np
clf = linear_model.LinearRegression()

select_columns = ['築年数', '部屋の向き', '部屋の広さ', '階数', '駅からの距離', '近い駅', 'ユニットバスかどうか']

train_data = pd.get_dummies(room_shape_date[select_columns], drop_first=True)
y_data = room_shape_date['家賃']

X_train, X_test, y_train, y_test = train_test_split(train_data, y_data, test_size=0.2, shuffle=True)

clf.fit(X_train, y_train)
y_pred_train = clf.predict(X_train)
mae_train = MAE(y_train, y_pred_train)
print(mae_train)

y_pred_test = clf.predict(X_test)
mae_test = MAE(y_test, y_pred_test)
print(mae_test)

from sklearn.model_selection import train_test_split
from sklearn.metrics import mean_absolute_error as MAE
import numpy as np
from sklearn.ensemble import RandomForestRegressor
from sklearn.model_selection import GridSearchCV

select_columns = ['築年数', '部屋の向き', '部屋の広さ', '階数', '駅からの距離', '近い駅', 'ユニットバスかどうか', '室内洗濯機置場' ]

train_data = pd.get_dummies(room_shape_date[select_columns], drop_first=True)
y_data = room_shape_date['家賃']

X_train, X_test, y_train, y_test = train_test_split(train_data, y_data, test_size=0.2, random_state=1234, shuffle=True)

frf = RandomForestRegressor()

frf.fit(X_train, y_train)
y_pred_train = frf.predict(X_train)
mae_train = MAE(y_train, y_pred_train)
print(mae_train)

y_pred_test = frf.predict(X_test)
mae_test = MAE(y_test, y_pred_test)
print(mae_test)

print(frf.score(X_test,y_test))

from sklearn.model_selection import train_test_split
from sklearn.metrics import mean_absolute_error as MAE
import numpy as np
from sklearn.ensemble import RandomForestRegressor
from sklearn.model_selection import GridSearchCV

select_columns = ['築年数', '部屋の広さ', '場所', '階数', '駅からの距離', '近い駅', 'ユニットバスかどうか', '室内洗濯機置場' ]

train_data = pd.get_dummies(room_shape_date[select_columns], drop_first=True)
y_data = room_shape_date['家賃']

X_train, X_test, y_train, y_test = train_test_split(train_data, y_data, test_size=0.2, random_state=1, shuffle=True)

parameters = {  
    'n_estimators': [50, 100, 300, 500, 1000],     # 決定木の数
    'max_features': ['sqrt', 'log2','auto', None],  # 利用する特徴量
    'max_depth':    [10, 20, 30, 40, 50, None],     # 決定木の深さ
    'random_state': [1]
}

frf = RandomForestRegressor()

# ハイパーパラメータチューニング(グリッドサーチのコンストラクタにモデルと辞書パラメータを指定)
gridsearch = GridSearchCV(estimator = frf,          # モデル
                          param_grid = parameters,  # ハイパーパラメータ
                          scoring = "r2"            # スコアリング
                         )

gridsearch.fit(X_train, y_train)

# グリッドサーチの結果から得られた最適なパラメータ候補を確認
print('Best params: {}'.format(gridsearch.best_params_)) 
print('Best Score: {}'.format(gridsearch.best_score_))

from sklearn.model_selection import train_test_split
from sklearn.metrics import mean_absolute_error as MAE
import numpy as np
from sklearn.ensemble import RandomForestRegressor
from sklearn.model_selection import GridSearchCV

select_columns = ['築年数', '部屋の広さ', '場所', '階数', '駅からの距離', '近い駅', 'ユニットバスかどうか', '室内洗濯機置場' ]

train_data = pd.get_dummies(room_shape_date[select_columns], drop_first=True)
y_data = room_shape_date['家賃']

X_train, X_test, y_train, y_test = train_test_split(train_data, y_data, test_size=0.2, random_state=1234, shuffle=True)

frf = RandomForestRegressor(max_depth=40, n_estimators=50, random_state=1)

frf.fit(X_train, y_train)
y_pred_train = frf.predict(X_train)
mae_train = MAE(y_train, y_pred_train)
print(mae_train)

y_pred_test = frf.predict(X_test)
mae_test = MAE(y_test, y_pred_test)
print(mae_test)

print(frf.score(X_test,y_test))

labels = X_train.columns
importances = frf.feature_importances_

dictionary = {k: v for k, v in zip(labels, importances)}
top_items = sorted(dictionary.items(), key=lambda x: x[1], reverse=True)[:20]
bottom_items = sorted(dictionary.items(), key=lambda x: x[1])[:20]

for key, value in top_items:
    print(key,": ", value)
    
print("-----------")

for key, value in bottom_items:
    print(key,": ", value)