使用skleran实现手写数字识别 模型：KNeighborsClassifier

1.加载数据集并分隔为特征和标签 导包

import numpy as np
import matplotlib.pyplot as plt
from sklearn import neighbors,datasets,cross_validation

def load_classification_data():
    digits=datasets.load_digits()
    print('数据集的大小',digits.data.shape)
    x_train=digits.data
    y_train=digits.target
    return cross_validation.train_test_split(x_train,y_train,test_size=0.25,random_state=0,stratify=y_train)

x_train,x_test,y_train,y_test = load_classification_data()

2.建立模型 测试评分

def test_KNN(*data):
    x_train,x_test,y_train,y_test = data
    clf=neighbors.KNeighborsClassifier()
    clf.fit(x_train,y_train)
    print('训练集上的评分:%f'%clf.score(x_train,y_train))
    print('测试集上的评分:%f'%clf.score(x_test,y_test))

输出结果：

训练集上的评分:0.991091
测试集上的评分:0.980000

3.调参  利用超参数搜索知道找最优参数

KNeighborsClassifier模型中几个比较重要的参数

k 选取几个近领点

weights=['uniform','distance'] weights为uniform表示统一权重 distance按距离加权

P 几种不同的距离计算公式

#超参数搜索
from sklearn.model_selection import GridSearchCV
ps=[1,2,10]
ks=[4,5,6,7,8,9,10]
ws=['uniform','distance']
param_grid={'p':ps,'n_neighbors':ks,'weights':ws}
cls=GridSearchCV(neighbors.KNeighborsClassifier(),param_grid,cv=5)
cls.fit(x_train,y_train)

print('最佳效果：%0.3f'%cls.best_score_)
print('最佳参数组合')
best_params=cls.best_estimator_.get_params()
for param_name in sorted(param_grid.keys()):
    print('\t%s:%r' %(param_name,best_params[param_name]))

print('训练集上的评分:%f'%cls.score(x_train,y_train))
print('测试集上的评分:%f'%cls.score(x_test,y_test))

#使用最佳参数重新预测
digits=datasets.load_digits()
clf=neighbors.KNeighborsClassifier(p=2,n_neighbors=4,weights='uniform')
clf.fit(x_train,y_train)
y_pre=cls.predict(x_test)
print("准确率",clf.score(x_test,y_test))
from sklearn.metrics import classification_report
print(classification_report(y_test,y_pre))