python-K近邻算法（附代码）

Knowledge

1. 估计器：用于分类、聚类和回归分析

1.1 fit()：训练算法，设置内部参数。该函数接收训练集及其类别两个参数。

1.2 predict()：参数为测试集。预测测试集类别，并返回一个包含测试集各条数据类别的数组。

3. 转换器：用于数据预处理和数据转换

4. 流水线：组合数据挖掘流程，便于再次使用

Example

1. 数据集：http://archive.ics.uci.edu/ml/datasets/Ionosphere

下载ionosphere.data 和 ionosphere.names

2. 代码

2.1 引入包

import numpy
import numpy as np
import  csv
from sklearn.model_selection import train_test_split, ShuffleSplit
from sklearn.neighbors import KNeighborsClassifier
from sklearn.model_selection import cross_val_score
import matplotlib.pyplot as plt

2.2 一折验证

data_filename = os.path.join(os.getcwd(),"ionosphere.data")
# print(data_filename)

x = numpy.zeros((351,34),dtype='float')
y = np.zeros((351,),dtype='bool')

with open(data_filename,'r') as input_file:
    reader = csv.reader(input_file)

    for i,row in enumerate(reader):
        data = [float(datum) for datum in row[:-1]]
        x[i] = data
        y[i] = row[-1] == 'g'
x_train,x_test,y_train,y_test = train_test_split(x,y,random_state=14)
estimator = KNeighborsClassifier(n_neighbors=1)
estimator.fit(x_train,y_train)
y_predicted = estimator.predict(x_test)
accuracy = np.mean(y_test==y_predicted)*100
print("The accuracuy is {0:.1f}%".format(accuracy))

2.3 k折验证

data_filename = os.path.join(os.getcwd(),"ionosphere.data")
# print(data_filename)

x = numpy.zeros((351,34),dtype='float')
y = np.zeros((351,),dtype='bool')

with open(data_filename,'r') as input_file:
    reader = csv.reader(input_file)

    for i,row in enumerate(reader):
        data = [float(datum) for datum in row[:-1]]
        x[i] = data
        y[i] = row[-1] == 'g'
        
# cv : set x-fold cross validation
cv = ShuffleSplit(n_splits=10, test_size=0.3, random_state=0)
# cross validation
scores = cross_val_score(estimator,x,y,scoring='accuracy',cv=cv)
print(scores)
average_accuracy = np.mean(scores) *100
print("The accuracuy is {0:.1f}%".format(average_accuracy))

2.4 找到最优k并plot出来

data_filename = os.path.join(os.getcwd(),"ionosphere.data")
# print(data_filename)

x = numpy.zeros((351,34),dtype='float')
y = np.zeros((351,),dtype='bool')

with open(data_filename,'r') as input_file:
    reader = csv.reader(input_file)

    for i,row in enumerate(reader):
        data = [float(datum) for datum in row[:-1]]
        x[i] = data
        y[i] = row[-1] == 'g'
avg_scores = []
all_scores = []
parameter_values = list(range(1,21))
for n_neighbors in parameter_values:
    estimator = KNeighborsClassifier(n_neighbors=n_neighbors)
    scores = cross_val_score(estimator,x,y,scoring='accuracy')
    avg_scores.append(np.mean(scores))
    all_scores.append(scores)

print(all_scores)
plt.plot(parameter_values,avg_scores,'-o')
plt.show()

2.5 标准预处理

2.5.1 导入from sklearn.preprocessing import MinMaxScaler

2.5.2 对x进行预处理x_transformed = MinMaxScaler().fit_transform(x)

data_filename = os.path.join(os.getcwd(),"ionosphere.data")
x = numpy.zeros((351,34),dtype='float')
y = np.zeros((351,),dtype='bool')

with open(data_filename,'r') as input_file:
    reader = csv.reader(input_file)

    for i,row in enumerate(reader):
        data = [float(datum) for datum in row[:-1]]
        x[i] = data
        y[i] = row[-1] == 'g'
        
x_transformed = MinMaxScaler().fit_transform(x)
x_train,x_test,y_train,y_test = train_test_split(x_transformed,y,random_state=14)
estimator = KNeighborsClassifier(n_neighbors=2)
estimator.fit(x_train,y_train)
y_predicted = estimator.predict(x_test)
scores = cross_val_score(estimator,x,y,scoring='accuracy')
print("The accuracuy is {0:.1f}%".format(np.mean(scores)*100))

2.6 流水线

2.6.1 库：from sklearn.pipeline import Pipeline

scaling_pipeline = Pipeline([('scale',MinMaxScaler()),
                             ('predict',KNeighborsClassifier(n_neighbors=2))])

data_filename = os.path.join(os.getcwd(),"ionosphere.data")
x = numpy.zeros((351,34),dtype='float')
y = np.zeros((351,),dtype='bool')

with open(data_filename,'r') as input_file:
    reader = csv.reader(input_file)

    for i,row in enumerate(reader):
        data = [float(datum) for datum in row[:-1]]
        x[i] = data
        y[i] = row[-1] == 'g'

scaling_pipeline = Pipeline([('scale',MinMaxScaler()),
                             ('predict',KNeighborsClassifier(n_neighbors=2))])
scores = cross_val_score(scaling_pipeline,x,y,scoring='accuracy')

print("The accuracuy is {0:.1f}%".format(np.mean(scores)*100))

2.6.2 循环找最优k

data_filename = os.path.join(os.getcwd(),"ionosphere.data")
# print(data_filename)

x = numpy.zeros((351,34),dtype='float')
y = np.zeros((351,),dtype='bool')

with open(data_filename,'r') as input_file:
    reader = csv.reader(input_file)

    for i,row in enumerate(reader):
        data = [float(datum) for datum in row[:-1]]
        x[i] = data
        y[i] = row[-1] == 'g'
x_transformed = MinMaxScaler().fit_transform(x)

# cv : set x-fold cross validation
cv = ShuffleSplit(n_splits=10, test_size=0.3, random_state=0)

avg_scores = []
all_scores = []
parameter_values = list(range(1,21))
for n_neighbors in parameter_values:
    estimator = KNeighborsClassifier(n_neighbors=n_neighbors)
    # scores = cross_val_score(estimator,x,y,scoring='accuracy')
    # pipeline
    scaling_pipeline = Pipeline([('scale', MinMaxScaler()),
                                 ('predict', KNeighborsClassifier(n_neighbors=n_neighbors))])
    scores = cross_val_score(scaling_pipeline, x, y, scoring='accuracy')
    avg_scores.append(np.mean(scores))
    all_scores.append(scores)

plt.plot(parameter_values,avg_scores,'-o')
plt.show()