Python: float()参数必须是字符串或数字，而不是'pandas._libs.interval.Interval'。

Question

我正在尝试完成Analytics Vidhya的贷款预测机器学习实践问题。在使用随机森林分类器时，出现以下错误：

TypeError:float() argument must be a string or a number, not 'pandas._libs.interval.Interval'

代码：

train['Loan_Status']=np.where(train['Loan_Status']=='Y', 1,0)

train_test_data=[train,test]

#Gender Feature
for dataset in train_test_data:
  dataset["Gender"]=dataset["Gender"].fillna('Male')
for dataset in train_test_data:
  dataset["Gender"]=dataset["Gender"].map({ "Female" : 1 , "Male" : 0}).astype(int)

#Married Feature 
for dataset in train_test_data:
  dataset['Married']=dataset['Married'].fillna('Yes')
for dataset in train_test_data:
  dataset['Married']=dataset['Married'].map({"Yes" : 1 , "No" : 0}).astype(int)

#Education Feature
for dataset in train_test_data:
  dataset['Education']=dataset['Education'].map({'Graduate' : 1 , 'Not Graduate' : 0}).astype(int)

#Combine Applicant income and coapplicant income
for dataset in train_test_data:
dataset['Income']=dataset['ApplicantIncome']+dataset['CoapplicantIncome']
train['IncomeBand']= pd.cut(train['Income'] , 4)
print(train[['IncomeBand' , 'Loan_Status']].groupby(['IncomeBand'] , as_index=False).mean())

for dataset in train_test_data:
  dataset.loc[dataset['Income'] <= 21331.5, 'Income'] =0
  dataset.loc[(dataset['Income'] > 21331.5) & (dataset['Income'] <= 41221.0), 'Income'] =1
  dataset.loc[(dataset['Income'] > 41221.0) & (dataset['Income'] <= 61110.5), 'Income'] =2
  dataset.loc[dataset['Income'] > 61110.5, 'Income'] =3
  dataset['Income']=dataset['Income'].astype(int)

# Loan Amount Feature
fillin=train.LoanAmount.median()
for dataset in train_test_data:
  dataset['LoanAmount']=dataset['LoanAmount'].fillna(fillin)
train['LoanAmountBand']=pd.cut(train['LoanAmount'] , 4)
print(train[['LoanAmountBand' , 'Loan_Status']].groupby(['LoanAmountBand'] , as_index=False).mean())

for dataset in train_test_data:
  dataset.loc[dataset['LoanAmount'] <= 181.75, 'LoanAmount'] =0
  dataset.loc[(dataset['LoanAmount'] >181.75) & (dataset['LoanAmount'] <= 354.5), 'LoanAmount'] =1
  dataset.loc[(dataset['LoanAmount'] > 354.5) & (dataset['LoanAmount'] <= 527.25), 'LoanAmount'] =2
  dataset.loc[dataset['LoanAmount'] > 527.25, 'LoanAmount'] =3
  dataset['LoanAmount']=dataset['LoanAmount'].astype(int)

#Loan Amount Term Feature
for dataset in train_test_data:
       dataset['Loan_Amount_Term']=dataset['Loan_Amount_Term'].fillna(360.0)

Loan_Amount_Term_mapping={360.0 : 1 , 180.0 : 2 , 480.0 : 3 , 300.0 : 4 , 84.0 : 5 , 240.0 : 6, 120.0 :7 , 36.0:8 , 60.0 : 9, 12.0 :10}

for dataset in train_test_data:
              dataset['Loan_Amount_Term']=dataset['Loan_Amount_Term'].map(Loan_Amount_Term_mapping)

# Credit History Feature
for dataset in train_test_data:
  dataset['Credit_History']=dataset['Credit_History'].fillna(2)

# Property Area Feature
for dataset in train_test_data:
 dataset['Property_Area']=dataset['Property_Area'].map({'Semiurban' : 0 , 'Urban' : 1 , 'Rural' : 2}).astype(int)

# Feature Selection
features_drop=['Self_Employed' , 'ApplicantIncome' , 'CoapplicantIncome', 'Dependents']
train=train.drop(features_drop, axis=1)
test=test.drop(features_drop, axis=1)
train.drop(['Loan_ID' , 'IncomeBand' , 'LoanAmountBand'] , axis=1)

X_train=train.drop('Loan_Status' , axis=1)
y_train=train['Loan_Status']
X_test=test.drop('Loan_ID' , axis=1).copy()

X_train.shape , y_train.shape , X_test.shape

clf = RandomForestClassifier(n_estimators=100)
clf.fit(X_train, y_train)
y_pred_random_forest = clf.predict(X_test)
acc_random_forest = round(clf.score(X_train, y_train) * 100, 2)
print (acc_random_forest)

我不知道浮点数错误是从哪里来的。如果有任何建议，将不胜感激。

- ALT

---

尝试运行 X.train.dtypes 并在此处添加结果。如果您最小化代码以说明问题，那也会很有帮助，参见 [mcve]。 - Shaido

浮点错误发生在哪一行？ - doctorlove

@doctorlove，浮点数错误发生在clf.fit（X_train，y_train）上。 - ALT

@Shaido 谢谢您的建议。我已经添加了 X_train 的结果类型。 - ALT

Answer 1

问题在于类别数据类型的列。这些列可以使用pd.cut函数创建。随机森林分类器不能将其作为输入，因此需要将它们转换为数字。最简单的方法是使用cat.codes。

在上面的代码中，需要将两个列IncomeBand和LoanAmountBand从类别更改为数字。

train['IncomeBand']= pd.cut(train['Income'] , 4).cat.codes
train['LoanAmountBand']=pd.cut(train['LoanAmount'] , 4).cat.codes

- Shaido

---

谢谢，错误已经修复，但现在我遇到了这个错误：无法将字符串“LP002990”转换为浮点数。 LP002990是训练数据集中的Loan_ID。我已经在上面的代码中删除了Loan_ID。你有什么建议吗？ - ALT

@ALT：从您的 X_train.dtypes 中可以看到该列仍然存在。在调用 drop 时，您可以添加 inplace=True，否则您需要将输出分配给一个新变量。 - Shaido