机器学习——共享单车数据集单项分析

加载需要的库文件

import matplotlib.pyplot as plt 
import pandas as pd
import numpy as np

import numpy as np # 用来存储和处理大型矩阵；NumPy和稀疏矩阵运算包SciPy配合使用更加方便
import pandas as pd # 数据处理，CSV文件输入输出 

import matplotlib.pyplot as plt #数据可视化工具
import seaborn as sns   #基于Matplotlib的Python可视化工具包，提供更高层次的用户接口，可以给
                        #出漂亮的数据统计图
color = sns.color_palette()#返回一个颜色定义颜色调色板

总租车人数cnt 的直方图／分布

# 读入数据
data = pd.read_csv("day.csv")
# 目标y（总租车人数cnt）的直方图／分布

#  解决中文乱码;
plt.rc('font', family='SimHei', size=13)
# 创建一个新图形
fig = plt.figure()
# 灵活的单变量分布的曲线图
# data.cnt.values：观察数据。如果这一系列目标的属性的名称，该名称将用于标签的数据；
# bins:直方图中箱子个数
# kda：是否为高斯核密度估计
sns.distplot(data.cnt.values, bins=50, kde=True)
# 设置当前轴的x轴标签
plt.xlabel('sum of person use bicycle', fontsize=12)
# 展示生成的图形
plt.show()
# 观测一下数据还算符合正态分布

总租车数的散点图分析

# 单个特征散点图
# 散点图中的x与y不同大小和/或颜色标记
# 1.range(data.shape[0])：数据位置
# 2.data["cnt"].values:
fig = plt.figure()
plt.scatter(range(data.shape[0]), data["cnt"].values,color='purple')
# 解决中文乱码;
plt.rc('font', family='SimHei', size=13)

plt.title("sum of bicycle");

工作日出现的次数

1为工作日，0为周末或节假日

#直方图 不连续
fig = plt.figure()  
#工作日
sns.countplot(data.workingday.values, order=[0, 1]);
#解决中文乱码;
plt.rc('font', family='SimHei', size=13)
plt.xlabel('weekday');
plt.ylabel('times');

风速分析

#直方图  连续
fig = plt.figure()
#风速
sns.distplot(data.windspeed.values, bins=30, kde=False)
#解决中文乱码;
plt.rc('font', family='SimHei', size=13)
plt.xlabel('风速', fontsize=12)
plt.show()

独热图表示两两特征之间的相关性

#获得所有需要的列值
cols=data.columns 
#cols=data[["holiday","workingday","weathersit","temp","atemp","hum","windspeed","cnt"]]
#data=data[["holiday","workingday","weathersit","temp","atemp","hum","windspeed","cnt"]]
# Calculates pearson co-efficient for all combinations，通常认为相关系数大于0.5的为强相关
data_corr = data.corr().abs()  

#独热图
plt.subplots(figsize=(11, 9))
sns.heatmap(data_corr,annot=True)

# Mask unimportant features
sns.heatmap(data_corr, mask=data_corr < 2, cbar=False)

plt.savefig('day_coor.png' )
plt.show()

只选择高度相关的两两属性

#设置阈值只选择高度相关的属性
threshold = 0.5
# 成对与以上阈值相关的列表 
corr_list = []
#size = data.shape[1]
size = data_corr.shape[0]

#搜索高相关对 
for i in range(0, size): #特性的数量
    for j in range(i+1,size): #避免重复
        if (data_corr.iloc[i,j] >= threshold and data_corr.iloc[i,j] < 1) or (data_corr.iloc[i,j] < 0 and data_corr.iloc[i,j] <= -threshold):
            corr_list.append([data_corr.iloc[i,j],i,j]) #存储相关性和列索引 
#首先显示高级的     
s_corr_list = sorted(corr_list,key=lambda x: -abs(x[0]))
#打印相关性和列名 
for v,i,j in s_corr_list:
    print ("%s and %s = %.2f" % (cols[i],cols[j],v))

weathersit and temp = 0.99
casual and registered = 0.95
instant and season = 0.87
dteday and yr = 0.83
windspeed and registered = 0.67
instant and casual = 0.66
temp and registered = 0.63
instant and registered = 0.63
weathersit and registered = 0.63
season and casual = 0.59
workingday and atemp = 0.59
season and registered = 0.57
temp and casual = 0.54
temp and windspeed = 0.54
weathersit and windspeed = 0.54
weathersit and casual = 0.54
weekday and windspeed = 0.52

以散射图显示高相关的属性

# 仅高相关对的散射图
for v,i,j in s_corr_list:
    sns.pairplot(data, size=6, x_vars=cols[i],y_vars=cols[j] )
    plt.show()
#    temp and atemp,天气温度和人体感温直接关系
#    atemp and cnt人体感温与租车数相关很大，温度合适租车会比较多，同理天气温度也是如此
#   weathersit and hum 天气情况会直接影响适度，因此关联度较高

分割数据集

import os
def load_data():#导入数据
    global x_data,y_data,name_data

    if not os.path.isfile("FE_day.csv"):#调用已经做好特征工程的文件，如果文件不存在，就调用函数生成该文件
        Data_preprocessing()
        
    data = pd.read_csv("FE_day.csv")

    data = data.drop(['instant','hum','windspeed'], axis = 1)#去掉编号、湿度、风速等不相关数据
##    print(data)
    
    y_data = data['cnt']
    x_data = data.drop('cnt', axis = 1)

    y_data=np.array(y_data)
    x_data=np.array(x_data)
    name_data =list(data.columns)#返回对象列索引
##### #将数据分割训练数据与测试数据
from sklearn.model_selection import train_test_split

load_data()
        # # 随机采样20%的数据构建测试样本，其余作为训练样本
X_train, X_test, y_train, y_test = train_test_split(x_data, y_data, random_state=0, test_size=0.20)
# X_train.shape

缺省参数的线性回归

# 线性回归
#class sklearn.linear_model.LinearRegression(fit_intercept=True, normalize=False, copy_X=True, n_jobs=1)
from sklearn.linear_model import LinearRegression

# 使用默认配置初始化
lr = LinearRegression()

# 训练模型参数
lr.fit(X_train, y_train)

# 预测
y_test_pred_lr = lr.predict(X_test)
y_train_pred_lr = lr.predict(X_train)


# 看看各特征的权重系数，系数的绝对值大小可视为该特征的重要性
fs = pd.DataFrame({"columns":list(data.columns), "coef":list((lr.coef_.T))})
fs.sort_values(by=['coef'],ascending=False)
# temp  windspeed  weathersit  相关度很高

正则化的线性回归

#岭回归／L2正则
#class sklearn.linear_model.RidgeCV(alphas=(0.1, 1.0, 10.0), fit_intercept=True, 
#                                  normalize=False, scoring=None, cv=None, gcv_mode=None, 
#                                  store_cv_values=False)
from sklearn.linear_model import  RidgeCV

from sklearn.metrics import r2_score
 


#设置超参数（正则参数）范围
alphas = [ 0.01, 0.1, 1, 10,100]
#n_alphas = 20
#alphas = np.logspace(-5,2,n_alphas)

#生成一个RidgeCV实例
ridge = RidgeCV(alphas=alphas, store_cv_values=True)  

#模型训练
ridge.fit(X_train, y_train)    

#预测
y_test_pred_ridge = ridge.predict(X_test)
y_train_pred_ridge = ridge.predict(X_train)


# 评估，使用r2_score评价模型在测试集和训练集上的性能
print ('对ridgecv测试R2-test评分', r2_score(y_test, y_test_pred_ridge))
print ('对ridgecv测试R2-test评分', r2_score(y_train, y_train_pred_ridge))
# 同样为负数

对ridgecv测试R2-test评分 0.8546366847253501
对ridgecv测试R2-test评分 0.827239578558045

from sklearn.linear_model import LinearRegression
from sklearn.linear_model import Ridge
from sklearn.linear_model import RidgeCV
from sklearn.linear_model import LassoCV
from sklearn.linear_model import ElasticNet

#lrg=LinearRegression()
#ridge=Ridge()
lasso = LassoCV(alphas= alphas)
lasso.fit(X_train, y_train)

mses = np.mean(lasso.mse_path_, axis = 1)
plt.plot(np.log10(lasso.alphas_), mses) 
#plt.plot(np.log10(lasso.alphas_)*np.ones(3), [0.3, 0.4, 1.0])
plt.xlabel('log(alpha)')
plt.ylabel('mse')
plt.show()    

print ('alpha is:', lasso.alpha_)

# 看看各特征的权重系数，系数的绝对值大小可视为该特征的重要性
fs = pd.DataFrame({"columns":list(data.columns), "coef_lr":list((lr.coef_.T)), "coef_ridge":list((ridge.coef_.T)), "coef_lasso":list((lasso.coef_.T))})
fs.sort_values(by=['coef_lr'],ascending=False)

alpha is: 1.0

mses = np.mean(lasso.mse_path_, axis = 1)
plt.plot(np.log10(lasso.alphas_), mses) 
#plt.plot(np.log10(lasso.alphas_)*np.ones(3), [0.3, 0.4, 1.0])
plt.xlabel('log(alpha)')
plt.ylabel('mse')
plt.show()    

print ('alpha =:', lasso.alpha_)