如何在现有的朴素贝叶斯算法（Python 3）中测试新数据

抱歉这里有任何明显的错误-我是一个真正的新手。我将一个数据集拆分为训练/测试，并成功应用了一个贝叶斯算法，结果为0.8888（见下面的代码）。现在，我想将第二个数据集应用到现有算法中——相同的特性和标签，但结果未知。我怎样才能做到这一点

import pandas as pd
import numpy as np

testdf = pd.read_csv("train_predictions.csv")

#change output settings
pd.set_option("display.width", 400)
pd.set_option("display.max_columns", 20)
pd.set_option("display.max_rows", 200)

# print data types of each column
print(testdf.dtypes)

# transform str data to numerical
from sklearn import preprocessing
le = preprocessing.LabelEncoder()
testdf["ID"] = le.fit_transform(testdf["ID"])
testdf["THAL"] = le.fit_transform(testdf["THAL"])
print(testdf.head())

# ID is not relevant to model, HEART DZ will be our target
cols = [col for col in testdf.columns if col not in    ["ID","HEART DZ"]]
data = testdf[cols]
target = testdf["HEART DZ"]
print(data.head())

from sklearn.model_selection import train_test_split
# split dataset
data_train, data_test, target_train, target_test = train_test_split(data, target, test_size=0.30, random_state=10)

# Gaussian Naive Bayes
from sklearn.naive_bayes import GaussianNB
from sklearn.metrics import accuracy_score

gnb = GaussianNB()
pred = gnb.fit(data_train, target_train).predict(data_test)
print("Naive-Bayes accuracy : ",accuracy_score(target_test, pred, normalize=True))

更新代码：

testdf = pd.read_csv("train_predictions.csv")
predictdf = pd.read_csv("export_dataframe.csv")

#change output settings
pd.set_option("display.width", 400)
pd.set_option("display.max_columns", 20)
pd.set_option("display.max_rows", 200)

# print data types of each column
#print(predictdf.head())

# transform str data to numerical
from sklearn import preprocessing
le = preprocessing.LabelEncoder()
testdf["ID"] = le.fit_transform(testdf["ID"])
testdf["THAL"] = le.fit_transform(testdf["THAL"])
predictdf["ID"] = le.fit_transform(predictdf["ID"])
predictdf["THAL"] = le.fit_transform(predictdf["THAL"])
#print(predictdf.head())

# ID is not relevant to model, HEART DZ will be our target (drop them)
cols = [col for col in testdf.columns if col not in ["ID","HEART DZ"]]
data = testdf[cols]
target = testdf["HEART DZ"]
pred_cols = [col for col in predictdf.columns if col not in ["ID","HEART DZ"]]
pred_data = predictdf[cols]
pred_target = predictdf["HEART DZ"]
#print(pred_data.head())

from sklearn.model_selection import train_test_split
# split dataset
data_train, data_test, target_train, target_test = train_test_split(data, target, test_size=0.30) #random_state=10)

# Gaussian Naive Bayes
from sklearn.naive_bayes import GaussianNB
from sklearn.metrics import accuracy_score

gnb = GaussianNB()
pred = gnb.fit(data_train, target_train).predict(data_test)
predictions = gnb.predict([predictdf])
#print("Naive-Bayes accuracy : ",accuracy_score(target_test, pred, normalize=True))
print(predictions)

更新代码2

testdf = pd.read_csv("train_predictions.csv")
testlabelsdf = pd.read_csv("train_labels.csv")
predictdf = pd.read_csv("export_dataframe.csv")
#print(testlabelsdf.head())

# transform str to int
from sklearn import preprocessing
le = preprocessing.LabelEncoder()
testdf["ID"] = le.fit_transform(testdf["ID"])
predictdf["ID"] = le.fit_transform(predictdf["ID"])
testdf["THAL"] = le.fit_transform(testdf["THAL"])
predictdf["THAL"] = le.fit_transform(predictdf["THAL"])

# ID is not relevant to model, HEART DZ will be our target (drop them)
cols = [col for col in testdf.columns if col not in ["ID"]]
data = testdf[cols]
target = testlabelsdf["HEART DZ"]

from sklearn.model_selection import train_test_split
# split dataset
data_train, data_test, target_train, target_test = train_test_split(data, target, random_state=10) #test_size=0.30,random_state=10)

# Gaussian Naive Bayes
from sklearn.naive_bayes import GaussianNB
from sklearn.metrics import accuracy_score

gnb = GaussianNB()
gnb.fit(data_train, target_train)
target_pred = gnb.predict(data_test)
ac = accuracy_score(target_test, target_pred, normalize=True)

yNew = gnb.predict(predictdf)
#print(yNew)

for i in range(len(predictdf)):
    print("Predicted: ", yNew[i])

scikit学习库中的类包含一个名为predict的方法，您的GaussianNB也不例外，因此，为了预测您可以使用的新数据的标签

...your codes here
gnb = GaussianNB() #again your code
...your codes here as well
data_to_predict = ... # load your new data to predict here 
predictions = gnb.predict[data_to_predict]
print(predictions)

---

有关详细信息，请参阅

谢谢您的回复。我按照您的建议执行了操作，并收到如下错误消息：无法将大小为90的序列复制到尺寸为15的阵列轴。我假设出现这个错误是因为我试图预测的数据集有90行，而训练数据集的输出有15行。我不知道如何纠正这个问题。上面是完整的编辑代码：您必须对预测数据应用相同的预处理操作（对trainig数据所做的操作）。