第二周<导学/分类>

分类学习####

分类算法各有不同

knn

naivebyes

regression

dnn

sklearn.linear_modlel 线性函数

sklearn.preprocessing 非线性函数

分类##

基本分类模型###

knn####

sklearn.neighbors.KNeighborsClassifier

n_neighbors

weights 设置选择k个点对分类效果的影响权重

uniform,distance

algothrim 计算临近点的方法 ball_tree kd_tree brute auto

...

• 实例介绍
  
  neigh=KNeightborsClassifier(neigh)
  
  neigh.fit(X,y)
  
  neigh.predict([[1.1]])

决策树####

构造好的分类决策树

sklearn.tree.DecisionTreeClassifier

criterion : 用于选择属性的准侧，可以传入gini代表基尼系数,entropy代表信息增益

max_features: 表示节点分裂时，从所少个特征中选择最优特征，可以设定固定数木，百分比，或其他个数，默认是所有特征

人体姿态预测###

这个数据集好偏啊，运行效果并不好。

# -*- coding: utf-8 -*-
"""
Created on Wed May 24 17:41:04 2017

@author: sfzyk
"""
import matplotlib.pyplot as plt
import numpy as np
import pandas as pd
from sklearn.preprocessing import Imputer
from sklearn.model_selection import train_test_split
from sklearn.metrics import classification_report

from sklearn.neighbors import KNeighborsClassifier
from sklearn.tree import DecisionTreeClassifier
from sklearn.naive_bayes import GaussianNB
import os

def load_dataset(feature_paths,label_paths):
    feature=np.ndarray(shape=(0,41))
    label=np.ndarray(shape=(0,1))
    for file in feature_paths:
        # 使用逗号分隔符读取特征数据，将问号替换标记为缺失值，不包含表头
        df=pd.read_table(file,delimiter=',',na_values='?',header=None)
        imp=Imputer(missing_values='NaN',strategy='mean',axis=0)
        #imp 是一个Imputer实例 用于预处理数据
        imp.fit(df)
        df=imp.transform(df)
        #进行特征合并
        feature=np.concatenate((feature,df))
    for file in label_paths:
        df=pd.read_table(file,header=None)
        label=np.concatenate((label,df))
    label=np.ravel(label)
    return feature,label

if __name__=='__main__':

    os.chdir(r"D:\mechine_learning\mooc_data\bodydata")
    feature_paths=['A.feature','B.feature','C.feature','D.feature','E.feature',]

    #label_paths=['A.label','B.label','C.label']
    #x_train,y_train=load_dataset(feature_paths[:2],label_paths[:2])
    #x_test,y_test=load_dataset(feature_paths[2:],label_paths[2:])
    label_paths=['A.label','B.label','C.label','D.label','E.label',]
    x_train,y_train=load_dataset(feature_paths[:4],label_paths[:4])
    x_test,y_test=load_dataset(feature_paths[4:],label_paths[4:])
    x_train,x_,y_train,y_=train_test_split(x_train,y_train,test_size=0.0)

    '''
    mask=np.random.permutation(x_test.shape[0])[0:10000]
    x_test=x_test[mask,:]
    y_test=y_test[mask]
    x_train,y_train=x_train[0:100000,:],y_train[0:100000]
    '''

    #只是随机打乱
    '''
    print("Start training knn")
    knn=KNeighborsClassifier(algorithm='kd_tree')
    knn.fit(x_train,y_train,)
    print("begin predic")
    answer_knn=knn.predict(x_test)
    print("knn are done")

    print("DT are training")
    DT=DecisionTreeClassifier()
    DT.fit(x_train,y_train)
    print("begin predic")
    answer_DT=DT.predict(x_test)
    print("DT are done")

    print("naive bayes are training")
    nb=GaussianNB()
    nb.fit(x_train,y_train)
    print("begin predic")
    answer_nb=nb.predict(x_test)
    print("naive nb are done!")
    print("knn")
    print(classification_report(y_test,answer_knn))

    print("DT")
    print(classification_report(y_test,answer_DT))

    print("nb")
    print(classification_report(y_test,answer_nb))
    '''

股票涨跌预测###

也是一个效果微乎其微的预测

就算只是预测单天是涨是跌，也只能做到53%左右的准确度。

# -*- coding: utf-8 -*-
"""
Created on Fri May 26 18:59:17 2017

@author: sfzyk
"""
import pandas as pd
import numpy as np
from sklearn import svm
from sklearn import cross_validation
import os
os.chdir("D:\mechine_learning\mooc_data\stock")

data=pd.read_csv("000777.csv",encoding='gbk',parse_dates=[0],index_col=0)
data.sort_index(0,ascending=True,inplace=True)

dayfeature=150
featurenum=5*dayfeature
x=np.zeros((data.shape[0]-dayfeature,featurenum+1))
y=np.zeros((data.shape[0]-dayfeature))

for i in range(0,data.shape[0]-dayfeature):
    x[i,0:featurenum]=np.array(data[i:i+dayfeature][['收盘价','最低价','最高价','开盘价','成交量']]).reshape((1,featurenum))
    x[i,featurenum]=data.ix[i+dayfeature]['开盘价']

for i in range(0,data.shape[0]-dayfeature):
    if data.ix[i+dayfeature]['收盘价']>=data.ix[i+dayfeature]['开盘价']:
        y[i]=1
    else:
        y[i]=0

clf=svm.SVC(kernel='rbf')
result=[]
for i in range(5):
    x_train,x_test,y_train,y_test=cross_validation.train_test_split(x,y,test_size=0.2)

    clf.fit(x_train,y_train)

    result.append(np.mean(y_test==clf.predict(x_test)))