Sklearn KNN模型fit函数报错：n_neighbors需传入整数而非浮点数

问题描述 #

使用sklearn的KNN模型进行二分类（类别Y取值为1或2），特征为X1、X2、X3，运行模型训练代码时触发错误：

"n_neighbors does not take <class 'float'> value, enter integer value"

即使将数据集改为全整数类型，错误依然存在，代码如下：

#Importing necessary libraries
import pandas as pd
import numpy as np
#Imports for KNN models
from sklearn.model_selection import train_test_split
from sklearn.neighbors import KNeighborsClassifier
#Imports for testing the model
from sklearn.metrics import confusion_matrix
from sklearn.metrics import f1_score
from sklearn.metrics import accuracy_score

#Import the data file
data = pd.read_csv("/content/drive/MyDrive/Python/Colab Notebooks/Onlyinttest.csv")

#Split data
X = data.loc[:,['X1','X2','X3']]

Y = data.loc[:,'Y']

X_train, X_test, Y_train, Y_test = train_test_split(X,Y, random_state=0, test_size=0.2)

#Determine k by using sqrt
import math
k = math.sqrt(len(Y_test))
print(k)
#Make k uneven
k = k-1

#KNN Model
classifer = KNeighborsClassifier(n_neighbors=k, p=2,metric='euclidean')
classifer.fit(X_train,Y_train)

问题原因 #

错误和数据集的数值类型完全无关，核心问题出在k的计算逻辑：

• math.sqrt()返回的结果是浮点数类型，即使执行k = k-1，最终结果依然是浮点数
• KNeighborsClassifier的n_neighbors参数要求必须是正整数，不接受任何浮点数输入

解决方案 #

将计算得到的k转换为整数，同时保证k为正奇数（符合你想要避免投票平局的需求），可以采用以下两种方式：

方式1：强制转为整数 #

直接用int()截断浮点数的小数部分，快速得到整数k：

#Determine k by using sqrt
import math
k = math.sqrt(len(Y_test))
print(k)
#Make k uneven and convert to integer
k = int(k - 1)
# 额外添加判断，避免极端情况（比如测试集样本数过少导致k为0或负数）
if k <= 0:
    k = 1

方式2：向下取整（语义更清晰） #

用math.floor()明确对数值向下取整，逻辑更直观：

#Determine k by using sqrt
import math
k = math.sqrt(len(Y_test))
print(k)
#Make k uneven and floor to integer
k = math.floor(k - 1)
# 确保k为正整数
if k <= 0:
    k = 1

修改后的完整代码 #

#Importing necessary libraries
import pandas as pd
import numpy as np
#Imports for KNN models
from sklearn.model_selection import train_test_split
from sklearn.neighbors import KNeighborsClassifier
#Imports for testing the model
from sklearn.metrics import confusion_matrix
from sklearn.metrics import f1_score
from sklearn.metrics import accuracy_score

#Import the data file
data = pd.read_csv("/content/drive/MyDrive/Python/Colab Notebooks/Onlyinttest.csv")

#Split data
X = data.loc[:,['X1','X2','X3']]
Y = data.loc[:,'Y']

X_train, X_test, Y_train, Y_test = train_test_split(X,Y, random_state=0, test_size=0.2)

#Determine k by using sqrt
import math
k = math.sqrt(len(Y_test))
print(k)
#Make k uneven and convert to integer
k = int(k - 1)
# 确保k是正整数，防止测试集样本数过少导致k无效
if k <= 0:
    k = 1

#KNN Model
classifer = KNeighborsClassifier(n_neighbors=k, p=2,metric='euclidean')
classifer.fit(X_train,Y_train)

内容的提问来源于stack exchange，提问作者Fabian