Chang-Xiao Li Machine Learning 2019 week1

 

读取csv文件 

import csv    #加载csv包便于读取csv文件
import numpy as np
import sys
from pandas import DataFrame
import pandas as pd



def getdata():
    # 读取文件
    filepath = "C:/Users/user/Downloads/"
    traindata = pd.read_csv(filepath + 'train.csv')
    testdata = pd.read_csv(filepath + 'test.csv')


    train = traindata[traindata['observation'] == 'PM2.5']
    test = testdata[testdata['AMB_TEMP'] == 'PM2.5']
    train = train.drop(['Date', 'stations', 'observation'], axis=1)
    test_x = test.iloc[:, 2:]


    train_x = []
    train_y = []
    for i in range(15):
        x = train.iloc[:, i:i + 9]
        x.columns = np.array(range(9))
        y = train.iloc[:, i + 9]
        y.columns = np.array(range(1))
        train_x.append(x)
        train_y.append(y)


    train_x = pd.concat(train_x)
    train_y = pd.concat(train_y)

    #
    train_y = np.array(train_y, float)
    test_x = np.array(test_x, float)
    print(train_x.shape, train_y.shape)
    return train_x,test_x,train_y

LR算法 

import numpy as np
import math
import pandas as pd
from sklearn.preprocessing import StandardScaler
'''损失函数'''
def lossfunc(x_train,y_train,theta):
    x_train = np.array(x_train,float)
    y_train = np.array(y_train, float)
    return np.sum((y_train - x_train.dot(theta)) ** 2) / len(y_train)
'''损失函数导数'''
def dlossfunc(x_train,y_train,theta):
    x_train = np.array(x_train,float)
    y_train = np.array(y_train, float)
    return x_train.T.dot(x_train.dot(theta) - y_train) * 2. / len(y_train)
'''归一化'''
def standardfunc(data):
    standard = StandardScaler()
    standard.fit(data)
    data =standard.transform(data)
    return data
'''预测'''
def predict( predict_x,theta):
     x = np.hstack([np.ones((len(predict_x), 1)), predict_x])
     result = x .dot(theta)
     return result
'''梯度下降'''
def gdmethod(x_train, y_train, ini_theta, eta, itbumber=1e4, epsilon=1e-8):
    theta = ini_theta
    inumber = 0

    while inumber < itbumber:
        gradient = dlossfunc(x_train, y_train,theta)
        pretheta = theta
        theta = theta - eta * gradient
        if (abs(lossfunc( x_train, y_train,pretheta) - lossfunc( x_train, y_train,pretheta)) < epsilon):
            break

        inumber += 1

    return theta

main：

import numpy as np
import DataHandler
import regressalgorithm

train_x,test_x,train_y = DataHandler.getdata()
print(train_x.shape)
print(train_y.shape)
print(test_x.shape)
train_x = regressalgorithm.standardfunc(train_x)

test_x = regressalgorithm.standardfunc(test_x)

train_x = np.hstack([np.ones((len(train_x), 1)), train_x])

initial_theta = np.zeros(train_x.shape[1])

theta =regressalgorithm.gdmethod(train_x,train_y,initial_theta,0.01,1e4)

predict_y = regressalgorithm.predict(test_x,theta)
print(predict_y)

结果文件：