介绍数据集
fuelType：燃料类型
rating：评级
renterTripsTaken：租房者出行
reviewCount：审阅计数
location.city：位置.城市
location.country：地点.国家/地区
location.latitude：位置.纬度
location.longitude：位置.长度
location.state：位置状态
owner.id：所有者id
rate.daily：每日费率
vehicle.make：车辆制造
vehicle.model：车辆型号
vehicle.type：车辆类型
vehicle.year：车辆.年
实战演练
数据集展示
import numpy as np
import pandas as pd
import matplotlib.pyplot as plt
import seaborn as sns
import plotly as py
import plotly_express as px
%matplotlib inline
import warnings
warnings.filterwarnings('ignore')
#导入数据
df = pd.read_csv('C:/CarRentalData.csv',engine='python')
df.head()
#数据集大小
df.shape
#(5851, 15)
数据的可视化
fuelType
df_fuelType = pd.DataFrame(df.fuelType.value_counts()).reset_index()
df_fuelType.rename(columns = {
'index':'fuelType', 'fuelType':'count'}, inplace=True)
fig = px.pie(df_fuelType, values = 'count', names='fuelType', title = 'Fuel Type',hole=.3,
color_discrete_map={
'ELECTRIC':'lightcyan',
'HYBRID':'cyan',
'GASOLINE':'royalblue',
'DIESEL':'darkblue'})
fig.show()
rating
print("Rating Statistics:")
print(df['rating'].describe())
'''
Rating Statistics:
count 5350.000000
mean 4.920325
std 0.182425
min 1.000000
25% 4.900000
50% 5.000000
75% 5.000000
max 5.000000
Name: rating, dtype: float64
'''
fig = px.histogram(df, x = 'rating', title = 'Histogram of Rental Car Rating')
fig.show()
renterTripsTaken
print("Renter Trips Taken Statistics:")
print(df['renterTripsTaken'].describe())
'''
Renter Trips Taken Statistics:
count 5851.000000
mean 33.477354
std 41.898954
min 0.000000
25% 5.000000
50% 18.000000
75% 46.000000
max 395.000000
Name: renterTripsTaken, dtype: float64
'''
fig = px.histogram(df, x = 'renterTripsTaken', title = 'Histogram of Renter Trips Taken')
fig.show()
reviewCount
print("Review Count Statistics:")
print(df['reviewCount'].describe())
'''
Review Count Statistics:
count 5851.000000
mean 28.454794
std 35.136113
min 0.000000
25% 4.000000
50% 16.000000
75% 39.000000
max 321.000000
Name: reviewCount, dtype: float64
'''
fig = px.histogram(df, x = 'reviewCount', title = 'Histogram of Review Count')
fig.show()
Car Rentals by City
import plotly.graph_objects as go
def get_average_lat_long(city, ltype):
choices = df[df['location.city'] == city]
lat = choices['location.latitude'].mean()
long = choices['location.longitude'].mean()
if ltype == 0:
return lat
else:
return long
df_location = pd.DataFrame(df['location.city'].value_counts()).reset_index()
df_location.rename(columns = {
'index':'city', 'location.city':'count'}, inplace=True)
df_location['latitude'] = df_location['city'].apply(lambda x: get_average_lat_long(x, 0))
df_location['longitude'] = df_location['city'].apply(lambda x: get_average_lat_long(x, 1))
df_location['text'] = df_location['city'] + '<br>Car Rentals ' + (df_location['count']).astype(str)
limits = [(0,20),(21,50),(51,150),(151,200),(201,1000),(1001,1500)]
colors = ["royalblue","orange","lightgrey","lightseagreen","red","crimson"]
scale = 0.5
fig = go.Figure()
for i in range(len(limits)):
lim = limits[i]
df_sub = df_location[lim[0]:lim[1]]
fig.add_trace(go.Scattergeo(
locationmode = 'USA-states',
lon = df_sub['longitude'],
lat = df_sub['latitude'],
text = df_sub['text'],
marker = dict(
size = df_sub['count']/scale,
color = colors[i],
line_color='rgb(40,40,40)',
line_width=0.5,
sizemode = 'area'
),
name = '{0} - {1}'.format(lim[0],lim[1])))
fig.update_layout(
title_text = 'Car Rentals by City',
showlegend = True,
geo = dict(
scope = 'usa',
landcolor = 'rgb(217, 217, 217)',
)
)
fig.show()
Car Rentals by State
df_state = pd.DataFrame(df['location.state'].value_counts()).reset_index()
df_state.rename(columns = {
'index':'state', 'location.state':'count'}, inplace=True)
fig = go.Figure(data=go.Choropleth(
locations=df_state['state'], # 空间坐标
z = df_state['count'].astype(float), # 要进行颜色编码的数据
locationmode = 'USA-states', # 位置集与locations中的条目匹配
colorscale = 'Reds',
colorbar_title = "Number of Cars Rented",
))
fig.update_layout(
title_text = 'Car Rentals by State',
geo_scope='usa',
)
fig.show()
own.id
df_owner = pd.DataFrame(df['owner.id'].value_counts()).reset_index()
df_owner.rename(columns = {
'index':'owner_id', 'owner.id':'number of rental cars'}, inplace=True)
print('Total Number of Unique Rental Cars per Owner Statistics:')
print(df_owner['number of rental cars'].describe())
'''
Total Number of Unique Rental Cars per Owner Statistics:
count 3093.000000
mean 1.891691
std 2.789205
min 1.000000
25% 1.000000
50% 1.000000
75% 2.000000
max 49.000000
Name: number of rental cars, dtype: float64
'''
fig = px.histogram(df_owner, x = 'number of rental cars', title='Total Number of Unique Rental Cars per Owner')
fig.show()
rate.daily
print('Daily Rate of Car Rental Statistics:')
print(df['rate.daily'].describe())
'''
Daily Rate of Car Rental Statistics:
count 5851.000000
mean 93.691506
std 96.080920
min 20.000000
25% 45.000000
50% 69.000000
75% 110.000000
max 1500.000000
Name: rate.daily, dtype: float64
'''
fig = px.histogram(df, x = 'rate.daily', title='Daily Rate of Car Rental')
fig.show()
Make and Model of Top 25 Most Rented Cars
df_make_model = df.groupby(['vehicle.make', 'vehicle.model']).size().reset_index()
df_make_model.rename(columns = {
0:'count'}, inplace=True)
df_make_model.replace('Mercedes-benz', 'Mercedes-Benz', inplace=True)
df_make_model['make_count'] = df_make_model['vehicle.make'].apply(lambda x : df_make_model[df_make_model['vehicle.make'] == x]['count'].sum())
df_make_model.sort_values(by = 'make_count', ascending=False, inplace=True)
fig = px.bar(df_make_model[df_make_model['make_count'] >45], x = 'vehicle.make', y='count', color = 'vehicle.model', title='Make and Model of Top 25 Most Rented Cars')
fig.update_layout(showlegend = False)
fig.show()
Vehicle Type of Rented Cars
df_vehicleType = pd.DataFrame(df['vehicle.type'].value_counts()).reset_index()
df_vehicleType.rename(columns = {
'index':'vehicle.type', 'vehicle.type':'count'}, inplace=True)
fig = px.pie(df_vehicleType, values = 'count', names='vehicle.type', title = 'Vehicle Type of Rented Cars')
fig.show()
Year of Vehicle
print('Vehicle Year Statistics:')
print(df['vehicle.year'].describe())
'''
Vehicle Year Statistics:
count 5851.000000
mean 2015.340113
std 4.050813
min 1955.000000
25% 2014.000000
50% 2016.000000
75% 2018.000000
max 2020.000000
Name: vehicle.year, dtype: float64
'''
fig = px.histogram(df, x = 'vehicle.year', title='Year of Vehicle')
fig.show()
热力图
plt.figure(figsize=(14,7))
sns.heatmap(df.corr(), annot=True)
数据预处理
#统计缺失值
df.isnull().sum()
'''
fuelType 75
rating 501
renterTripsTaken 0
reviewCount 0
location.city 0
location.country 0
location.latitude 0
location.longitude 0
location.state 0
owner.id 0
rate.daily 0
vehicle.make 0
vehicle.model 0
vehicle.type 0
vehicle.year 0
dtype: int64
'''
#数据集各列的数据类型，是否为空值，内存占用情况
df.info()
'''
<class 'pandas.core.frame.DataFrame'>
RangeIndex: 5851 entries, 0 to 5850
Data columns (total 15 columns):
# Column Non-Null Count Dtype
--- ------ -------------- -----
0 fuelType 5776 non-null object
1 rating 5350 non-null float64
2 renterTripsTaken 5851 non-null int64
3 reviewCount 5851 non-null int64
4 location.city 5851 non-null object
5 location.country 5851 non-null object
6 location.latitude 5851 non-null float64
7 location.longitude 5851 non-null float64
8 location.state 5851 non-null object
9 owner.id 5851 non-null int64
10 rate.daily 5851 non-null int64
11 vehicle.make 5851 non-null object
12 vehicle.model 5851 non-null object
13 vehicle.type 5851 non-null object
14 vehicle.year 5851 non-null int64
dtypes: float64(3), int64(5), object(7)
memory usage: 685.8+ KB
'''
df['fuelType'].unique()
'''
array(['ELECTRIC', 'HYBRID', 'GASOLINE', nan, 'DIESEL'], dtype=object)
'''
df['fuelType'].value_counts().reset_index()
'''
index fuelType
0 GASOLINE 4810
1 ELECTRIC 622
2 HYBRID 274
3 DIESEL 70
'''
#填补缺失值
df['fuelType'] = df['fuelType'].fillna('GASOLINE',axis=0)
#求众数
mode = df['rating'].mode()
mode = int(mode)
mode
'''
5
'''
df['rating'] = df['rating'].fillna(mode,axis=0)
pd.isnull(df).sum()
'''
fuelType 0
rating 0
renterTripsTaken 0
reviewCount 0
location.city 0
location.country 0
location.latitude 0
location.longitude 0
location.state 0
owner.id 0
rate.daily 0
vehicle.make 0
vehicle.model 0
vehicle.type 0
vehicle.year 0
dtype: int64
'''
可以发现缺失值已经填补完毕，没有缺失值了
df['fuelType'].unique()
'''
array(['ELECTRIC', 'HYBRID', 'GASOLINE', 'DIESEL'], dtype=object)
'''
#映射
fuelType_mapping= {
'ELECTRIC':0.0, 'HYBRID':1.0,'GASOLINE':2.0,'DIESEL':3.0}
df['fuelType'] = df['fuelType'].map(fuelType_mapping)
df['location.state'].unique()
'''
array(['WA', 'NM', 'GA', 'SC', 'FL', 'TX', 'NC', 'CT', 'MA', 'ME', 'AL',
'MT', 'TN', 'KY', 'ID', 'UT', 'MD', 'DC', 'IA', 'OH', 'CO', 'VA',
'MI', 'NJ', 'IN', 'WI', 'KS', 'MO', 'NV', 'CA', 'LA', 'AR', 'IL',
'MS', 'NH', 'MN', 'OK', 'NE', 'OR', 'PA', 'DE', 'AZ', 'WV', 'RI',
'AK', 'HI'], dtype=object)
'''
locationState_mapping= {
'WA':0.0, 'NM':1.0, 'GA':2.0, 'SC':3.0, 'FL':4.0, 'TX':5.0, 'NC':6.0, 'CT':7.0, 'MA':8.0, 'ME':9.0, 'AL':10.0,
'MT':11.0, 'TN':12.0, 'KY':13.0, 'ID':14.0, 'UT':15.0, 'MD':16.0, 'DC':17.0, 'IA':18.0, 'OH':19.0, 'CO':20.0, 'VA':21.0,
'MI':21.0, 'NJ':22.0, 'IN':23.0, 'WI':24.0, 'KS':25.0, 'MO':26.0, 'NV':27.0, 'CA':28.0, 'LA':29.0, 'AR':30.0, 'IL':31.0,
'MS':32.0, 'NH':33.0, 'MN':34.0, 'OK':35.0, 'NE':36.0, 'OR':37.0, 'PA':38.0, 'DE':39.0, 'AZ':40.0, 'WV':41.0, 'RI':42.0,
'AK':43.0, 'HI':44.0}
df['location.state'] = df['location.state'].map(locationState_mapping)
df['vehicle.type'].unique()
'''
array(['suv', 'car', 'truck', 'minivan', 'van'], dtype=object)
'''
vehicleType_mapping= {
'suv':0.0, 'car':1.0,'truck':2.0,'minivan':3.0,'van':4.0}
df['vehicle.type'] = df['vehicle.type'].map(vehicleType_mapping)
其他列的值比较繁多，所以不做考虑，删掉它们就可以了
df['vehicle.make'].unique()
df['vehicle.model'].unique()
df = df.drop(labels=['location.city','location.country','vehicle.make','vehicle.model'],axis=1)
检查数据
df.head()
归一化与切分数据集
from sklearn.model_selection import train_test_split
from sklearn.preprocessing import StandardScaler
y = df.pop('vehicle.type')
X = df
scaler = StandardScaler()
X_scaled = scaler.fit_transform(X)
X_train, X_valid, y_train, y_valid = train_test_split(X_scaled, y)
数据训练
import sklearn
from sklearn.preprocessing import StandardScaler,PolynomialFeatures
from sklearn.impute import SimpleImputer
from sklearn.preprocessing import OrdinalEncoder
from sklearn.ensemble import RandomForestRegressor
from sklearn.model_selection import train_test_split
from sklearn.metrics import mean_squared_error, r2_score
rfc = RandomForestRegressor()
rfc.fit(X_train, y_train)
# Feature Importances
features = X.columns
feature_imps = rfc.feature_importances_#特征重要性
# 重要特征的可视化
fig = px.bar(x=features, y=feature_imps)
fig.update_layout({
'title':{
'text':"Feature Importnace", 'x':0.5}})
fig.show()
#多分类线性模型
from sklearn.svm import LinearSVC
linear_svm = LinearSVC().fit(X_train,y_train)
print("模型train得分:{:.3f}".format(linear_svm.score(X_train,y_train)))
print("模型test得分:{:.3f}".format(linear_svm.score(X_valid,y_valid)))
'''
模型train得分:0.624
模型test得分:0.643
'''
# Prediction
pred = linear_svm.predict(X_valid)
linear_svm_mse = mean_squared_error(y_valid, pred)
linear_svm_rmse = np.sqrt(linear_svm_mse)
linear_svm_r2 = r2_score(y_valid, pred)
print(linear_svm_mse,linear_svm_rmse,linear_svm_r2)
'''
0.5399863294600137 0.7348376211517845 -0.11486343099011109
'''