共享单车项目

import numpy as np
import pandas as pd
from sklearn.decomposition import PCA,TruncatedSVD
from sklearn.linear_model import Lasso,Ridge,LinearRegression
from sklearn.pipeline import  Pipeline
from sklearn.preprocessing import StandardScaler,MinMaxScaler,LabelEncoder,Binarizer,OneHotEncoder
from sklearn.model_selection import train_test_split,GridSearchCV
import seaborn as sns
import matplotlib.pyplot as plt
import datetime
import calendar
from sklearn.metrics import r2_score
import warnings
warnings.filterwarnings('ignore')
def get_month(mdt):
    mobj=datetime.datetime.fromisoformat(mdt)
    mobj=mobj.month
    return calendar.month_name[mobj]

def get_day(ddt):
    dobj=datetime.datetime.fromisoformat(ddt)
    return dobj.day

def get_weekday(wdt):
    wobj=datetime.datetime.fromisoformat(wdt)
    wobjw=wobj.weekday()
    return calendar.day_name[wobjw]

def get_hour(hdt):
    hobj=datetime.datetime.fromisoformat(hdt)
    return hobj.hour

def change_to_int(x):
    return int(x)


def main():
    data=pd.read_csv('train.csv')
    print(data)
    data['month_section']=data['datetime'].apply(func=get_month)
    data['day_section']=data['datetime'].apply(func=get_day)
    data['weekday_section'] = data['datetime'].apply(func=get_weekday)
    data['hour_section'] = data['datetime'].apply(func=get_hour)

    data['temp_int']=data['temp'].apply(func=change_to_int)
    data['atemp_int'] = data['atemp'].apply(func=change_to_int)
    data['windspeed_int'] = data['windspeed'].apply(func=change_to_int)
    #绘制去除噪音之前的条形图
    fig,axes=plt.subplots(nrows=2,ncols=2)
    #1.绘制去除噪音之前的箱线图
    sns.boxplot(
        data=data,
        x='season',
        y='count',
        ax=axes[0,0]
    )
    #去除箱线图的噪音
    #1.取出骑行量数据
    dcount=data.loc[:,['count']]
    d_mean=dcount.mean()
    d_std=dcount.std()
    d_good=dcount-d_mean<=2*d_std
    d_good_val=d_good.values.flatten()
    # 将非噪声数据筛选出来
    d_good_no_noise=data.loc[d_good_val,:]
    #2.绘制去除噪音箱线图
    sns.boxplot(
        data=d_good_no_noise,
        x='season',
        y='count',
        ax=axes[0,1]
    )
    #3.绘制热力图
    ls= ["season", "month_section", "day_section","weekday_section", "hour_section","count",]
    d_heat=d_good_no_noise.loc[:,ls]
    d_corr=d_heat.corr()
    sns.heatmap(data=d_corr,annot=True,ax=axes[1,0])

    #4.绘制柱形图
    dd_count_reset_index=d_good_no_noise.groupby(by=['season']).agg(np.mean).reset_index()
    sns.barplot(
        data=dd_count_reset_index,
        x='season',
        y='count',
        ax=axes[1,1]
    )
    plt.show()
    #5.绘制点图
    fig,axes=plt.subplots(nrows=2,ncols=2)
    dd_point_reset_index=d_good_no_noise.groupby(by=['hour_section','weekday_section']).agg(np.mean).reset_index()
    sns.pointplot(
        data=dd_point_reset_index,
        x='hour_section',
        y='count',
        hue='weekday_section',
        ax=axes[0,0]
    )
    dd_point_reset_index = d_good_no_noise.groupby(by=['hour_section', 'month_section']).agg(np.mean).reset_index()
    sns.pointplot(
        data=dd_point_reset_index,
        x='hour_section',
        y='count',
        hue='month_section',
        ax=axes[0, 1]
    )

    plt.show()
    d_good_no_noise['weekday_hour']=d_good_no_noise.apply(
        lambda line:line['weekday_section']+'_'+str(line['hour_section']),axis=1)
    d_good_no_noise['month_hour'] = d_good_no_noise.apply(
        lambda line: str(line['month_section']) + '_' + str(line['hour_section']), axis=1)
    d_good_no_noise['season_hour'] = d_good_no_noise.apply(
        lambda line: str(line['season']) + '_' + str(line['hour_section']), axis=1)
    print(d_good_no_noise.columns)

    """ ------ 删除冗余字段 开始 ------ """
    la=['datetime','temp','atemp','windspeed']
    for i in la:
        d_good_no_noise.drop(i,inplace=True,axis=1)
    print(d_good_no_noise.columns)
    print(len(d_good_no_noise.columns))
    #独热(对里面是字符型数据)
    lb=['season_hour','month_hour','weekday_hour','season','weather']
    pd.get_dummies(data=d_good_no_noise,columns=lb)

    print(d_good_no_noise.columns)
    print(len(d_good_no_noise.columns))
    y=d_good_no_noise.pop('count')
    x=d_good_no_noise
    print(x.shape)
    print(y.shape)
    #缩放
    sobj=StandardScaler()
    #降维
    x=sobj.fit_transform(x)
    pobj=PCA(n_components=4)
    x=pobj.fit_transform(x)
    trainx,testx,trainy,testy=train_test_split(x,y,train_size=0.7,shuffle=True)
    #网格最优化参数
    robj=Ridge()
    params={
        'alpha':[i for i in range(5,10)]
    }
    gobj=GridSearchCV(estimator=robj,param_grid=params,cv=5)
    gobj.fit(x,y)
    print(gobj.best_params_)
    #调用线性回归包，传入最好的alpha
    rgobj=Ridge(alpha=gobj.best_params_['alpha'])
    rgobj.fit(trainx,trainy)
    print(rgobj.score(testx, testy))
    testz=rgobj.predict(testx)
    #r2精度
    print(r2_score(testy, testz))

if __name__ == '__main__':
    main()