NYC出租车数据分析（Python）

1.1 导入需要使用的包：

  import pandas as pd
  import numpy as np
  from matplotlib import pyplot as plt
  %matplotlib inline

1.2 利用pd.read_csv导入数据， 命名为cab

 cab = pd.read_csv('/Users/tangyu/Desktop/test.csv')

1.3 查看cab的数据组成

cab.info

1.4 结果如下，注意1） 数据类型为DataFrame 2）列名和类型 3）内存使用情况

<class 'pandas.core.frame.DataFrame'>

RangeIndex: 625134 entries, 0 to 625133

Data columns (total 9 columns):

id                    625134 non-null object

vendor_id             625134 non-null int64

pickup_datetime       625134 non-null object

passenger_count       625134 non-null int64

pickup_longitude      625134 non-null float64

pickup_latitude       625134 non-null float64

dropoff_longitude     625134 non-null float64

dropoff_latitude      625134 non-null float64

store_and_fwd_flag    625134 non-null object

dtypes: float64(4), int64(2), object(3)

memory usage: 42.9+ MB

1.5 查看数据前/后5项

cab.head() #  前5项
cab.tail() # 后5项
len(cab.id) # 查看数据总数
>>> 625134

1.6 自定义函数Jan返回一月份日期，利用filter函数代替for循环

 dates = cab['pickup_datetime']
 pas = cab['passenger_count']
 def jan(date):
     return date <= '2016-01-31'
 jan_list = filter(jan, dates)

1.7 np.unique函数计算每辆车运营的乘客人数
sum()计算总人数
利用for循环计算每一种人数客车的比例

pas_per_car = np.unique(cab['passenger_count']) #       total_pas = sum(cab['passenger_count']) #     pas_per_car_ratio = [] #     
pas_per_car_counts = []
  for i in pas_per_car:
      pas_per_car_count = sum(cab['passenger_count'] == i)
      pas_per_car_counts.append(pas_per_car_count)
      ratio = pas_per_car_count/float(total_pas)
      pas_per_car_ratio.append(ratio)

1.8 计算每个月的数据，注意 & 和 | 表示并和或

# 获取每个月的数据     
cab_Jan = cab[cab['pickup_datetime'] <= '2016-02-01'] cab_Feb = cab[(cab['pickup_datetime'] <= '2016-03-01') & (cab['pickup_datetime'] >='2016-02-01')]
cab_May = cab[(cab['pickup_datetime'] <= '2016-04-01') & (cab['pickup_datetime'] >='2016-03-01')]
cab_Apr = cab[(cab['pickup_datetime'] <= '2016-05-01') & (cab['pickup_datetime'] >='2016-04-01')]
cab_Mar = cab[(cab['pickup_datetime'] <= '2016-06-01') & (cab['pickup_datetime'] >='2016-05-01')]
cab_Jun = cab[(cab['pickup_datetime'] <= '2016-07-01') & (cab['pickup_datetime'] >='2016-06-01')]

1.9 每个月乘客人数分别求；每个月乘客人数的最大值，可以看出每个月都一致

pas_list=[] # 每个月的passengers人数求和
pas_max_list = [] # 每个月人数最大值
pas_per_car_monthly_counts = [] #每个月的人数比较

for cab_mons in [cab_Jan, cab_Feb, cab_May, cab_Apr, cab_Mar, cab_Jun]: pas_count = sum(cab_mons['passenger_count']) #总人数
pas_list.append(pas_count) # list.append()
pas_max = max(cab_mons['passenger_count']) #
      cab_max_monthly = cab_mons[cab_mons['passenger_count'] == pas_max]
      pas_max_list.append(pas_max)
      for i in pas_per_car:
          pas_per_car_count_month = sum(cab_mons['passenger_count'] == i)
          pas_per_car_monthly_counts.append(pas_per_car_count_month)
  print pas_list
  print pas_max_list
  print pas_per_car_monthly_counts

[162880, 169805, 182752, 178058, 179140, 166191] [9, 6, 6, 6, 6, 9] [2, 69319, 13984, 3848, 1848, 5396, 3278, 1, 2, 72881, 14592, 4012, 1897, 5464, 3466, 0, 8, 77849, 15545, 4576, 2177, 5875, 3667, 0, 3, 76139, 15647, 4480, 2055, 5683, 3425, 0, 6, 75934, 15793, 4484, 2068, 5814, 3471, 0, 2, 71325, 14466, 4286, 1972, 5179, 3214, 1]

pas_per_car #每辆车的人数
>>> array([0, 1, 2, 3, 4, 5, 6, 9])

1.10 最后一列为N的共3430行数

sum(cab.store_and_fwd_flag !='N')
>>> 3430

作图部分

2.1 配色卡

import seaborn as sns sns.palplot(sns.cubehelix_palette(8, start=.5, rot=-.75))

本文使用的配色卡

2.2 不同人数的乘客比例分布

plt.figure(figsize=(7,7))
  labels = pas_per_car
  explode = (0, 0.1, 0.1, 0, 0, 0, 0, 0)  # explode 1st slice
  # Plot
  plt.pie(pas_per_car_counts, explode=explode, labels=labels,
          autopct='%1.1f%%', shadow=True, startangle=140)
  plt.axis('equal')
  plt.legend(loc='upper left', bbox_to_anchor=(-0.1, 1))

2.3 不同月份中不同人数的占比分布：基本一致

 plt.figure(figsize=(20,20))
labels = pas_per_car
explode = (0, 0.1, 0.1, 0, 0, 0, 0, 0) # explode 1st slice
# Plot
y = 0;
xlabel = ['Jan', 'Feb', 'May', 'Apr', 'Mar', 'Jun']
for i in range(1,7):
    plt.subplot(3,2,i)
    plt.pie(pas_per_car_monthly_counts[y:y+8], explode=explode,
labels=labels,
            autopct='%1.1f%%', shadow=True, startangle=140)
    plt.axis('equal')
plt.xlabel(xlabel[i-1])
plt.legend(loc='upper left', bbox_to_anchor=(-0.1, 1))
y=y+8

2.4 发现有乘客为0的情况，可能的解释 1）这种情况可能是司机师傅没有打表计费； 2）乘车人取消了订单；3）数据错误

sum(cab['passenger_count'] == 0)
>>> 23

2.5 2016上半年总运载的乘客人数为1038826人次

total_pas
>>> 1038826

2.6 一月中乘客为9人的数据

cab_mons[cab_mons['passenger_count'] == pas_max]

2.7 一月到六月的乘客人数对比

# Plot
x = range(1,7) # 生成1～6的数组
plt.ylim((150000, 200000)) # 固定y的坐标轴值
plt.xlabel('Month') # 添加横坐标文字
plt.ylabel('Passenger count') # 添加纵坐标文字
sns.barplot(x, pas_list) 

一月到六月的乘客人数条形图对比

仔细看第一天的情况

3.1 获取第一天的数据

cab_1st_day = cab[cab['pickup_datetime'] <= '2016-01-02'] # the first day
   cab_hour = []
  for i in range(24):
      cab['pickup_datatime'] <=
      sum_by_hour = sum(cab[])
>>> (array([ 2057., 0., 605., 0., 155., 0., 103., 0., 168., 93.]), array([ 1. , 1.5, 2. , 2.5, 3. , 3.5, 4. , 4.5, 5. , 5.5, 6. ]), )

3.2 一天的人流变化

# 一天的人流变化
import datetime
date = [] 
time = []
pas_hour = 0
pas_hours = []
date_point = datetime.datetime.strptime("2016-01-01 22:59:59", "%Y-%m-%d %H:%M:%S") 
for i in range(0,len(cab_1st_day['pickup_datetime'])):
# for i in range(0,1100):
    date_time = cab_1st_day['pickup_datetime'].tolist()[i]
    date_obj = datetime.datetime.strptime(date_time,"%Y-%m-%d %H:%M:%S")
    if date_obj > date_point:
        pas_hour += sum(cab_1st_day[cab_1st_day['pickup_datetime'] == date_time].passenger_count) # 加该时间短的人数
    else:
        pas_hours.append(pas_hour)
        pas_hour = 0
        date_point -= datetime.timedelta(hours=1)# 1个小时作为一个间隔
        pas_hour += sum(cab_1st_day[cab_1st_day['pickup_datetime'] == date_time].passenger_count) # 加该时间短的人数

# plot
x = range(1,len(pas_hours)+1)
#plt.ylim((150000, 200000))
plt.xlabel('hour')
plt.ylabel('Passenger count in the first day')
# plt.bar(x , pas_hours) 
sns.barplot(x, pas_hours)

3.3 六月的情况

import datetime
date = [] 
time = []
pas_hour = 0
pas_hours = []
date_point = datetime.datetime.strptime("2016-06-30 23:59:59", "%Y-%m-%d %H:%M:%S")
for i in range(0,len(cab_Jun['id'])):
# for i in range(0,1100):
    date_time = cab_Jun['pickup_datetime'].tolist()[i]
    date_obj = datetime.datetime.strptime(date_time,"%Y-%m-%d %H:%M:%S")
    if date_obj > date_point:
        pas_hour += sum(cab_Jun[cab_Jun['pickup_datetime'] == date_time].passenger_count) # 加该时间短的人数
    else:
        pas_hours.append(pas_hour)
        pas_hour = 0
        date_point -= datetime.timedelta(days=1)# 1个小时作为一个间隔
        pas_hour += sum(cab_Jun[cab_Jun['pickup_datetime'] == date_time].passenger_count) # 加该时间短的人数

len(cab_Jun['id']) # 六月的记录条数
>>> 100445

 sum(cab_Jun['passenger_count']) # 六月的乘客数

画了最后三天的柱状图：

plt.figure(figsize=(10,10))
  x = range(1,len(pas_hours)+1)
plt.xlabel('hour')
  plt.ylabel('Passenger count in Jun')
sns.barplot(x, pas_hours)

三天的流量分布

根据经纬度画出地图

4.1 上车和下车的经纬度最大/最小值

   lon_max = max(cab['pickup_longitude'])
  lon_min = min(cab['pickup_longitude'])
  lat_max = max(cab['pickup_latitude'])
  lat_min = min(cab['pickup_latitude'])
  print(lon_max)
  print(lon_min)
  print(lat_max)
  print(lat_min)
>>> -69.248916626 -121.933128357 42.8149375916 37.3895874023

4.2 统计相关

 import math
  longitudes =list['pickup_longitude','dropoff_longitude']
  for longitude in longitudes:
df['longitude']

4.3 首先利用matplotlib 和 basemap 画出地图

from mpl_toolkits.basemap import Basemap, cm
  import numpy as np
  import matplotlib.pyplot as plt
  from matplotlib import cm

fig = plt.figure(figsize=(8,8))
ax = fig.add_axes([0.1,0.1,0.8,0.8])
# create polar stereographic Basemap instance.
# NYC地图
#m = Basemap(projection='stere',
#           lon_0=-73.93, # 地图中心
#           lat_0=40.65,
#           llcrnrlat=40.2,
#           urcrnrlat=40.9,
#           llcrnrlon=-74.10,
#           urcrnrlon=-73.75,
#           resolution='l')
# 美国整体地图
m = Basemap(projection='stere',lat_0=90,lon_0=-105,
            llcrnrlat=23.41 ,urcrnrlat=45.44,
           llcrnrlon=-118.67,urcrnrlon=-64.52,
           rsphere=6371200.,resolution='l',area_thresh=10000)

# draw coastlines, state and country boundaries, edge of map.
#m.drawcoastlines() # drew coastlines
m.drawstates() # draw states 
m.drawcountries() # draw countries
#m.drawcounties() # draw conties
# draw parallels.
parallels = np.arange(0.,90,10.)
m.drawparallels(parallels,labels=[1,0,0,0],fontsize=10)
# draw meridians
#meridians = np.arange(180.,360.,10.)
#m.drawmeridians(meridians,labels=[0,0,0,1],fontsize=10)
meridians = np.arange(-110.,-60.,10.)
m.drawmeridians(meridians,labels=[0,0,0,1],fontsize=10)
m.fillcontinents(color = 'coral')
# ny = data.shape[0]; nx = data.shape[1]
#lons, lats = m.makegrid(cab_1st_day['pickup_longitude'], cab_1st_day['pickup_latitude']) # get lat/lons of ny by nx evenly space grid.



lons = np.array(cab_1st_day['pickup_longitude'])
lats = np.array(cab_1st_day['pickup_latitude'])
x, y = m(lons, lats)
#m.scatter(x,y,50,marker='.',color='b')
plt.scatter(x,y,50,cmap=cm.hsv, color ='b')

plt.scatter(lons,lats,50,cmap=cm.hsv, color = 'b') #

仔细看一天的地图

5.1 获取第一天的上下车经纬度和时间

lons = np.array(cab_1st_day['pickup_longitude'])
  lats = np.array(cab_1st_day['pickup_latitude'])
  time = np.array(cab_1st_day['pickup_datetime'])

5.2 第一天的上下车情况，以一个小时做为时间间隔，画出每个小时的散点图

# 一天的人流变化
import folium
import datetime
import time
import os
from selenium import webdriver
#date = [] 
#time = []
#pas_hour = 0
#pas_hours = []
j=1;
m = folium.Map(location=[40.8, -73.9],zoom_start=11)
#p_lons = np.array(cab_1st_day['pickup_longitude'])
#p_lats = np.array(cab_1st_day['pickup_latitude'])
d_lons = np.array(cab_1st_day['dropoff_longitude'])
d_lats = np.array(cab_1st_day['dropoff_latitude'])
date_point = datetime.datetime.strptime("2016-01-01 22:59:59", "%Y-%m-%d %H:%M:%S") 
#for p_lat, p_lon, i in zip(p_lats, p_lons,  range(0,len(p_lons))):
for d_lat, d_lon, i in zip(d_lats, d_lons,  range(0,len(d_lons))):
# for i in range(0,1000):
    date_time = cab_1st_day['pickup_datetime'].tolist()[i]
    date_obj = datetime.datetime.strptime(date_time,"%Y-%m-%d %H:%M:%S")
    if date_obj > date_point:
        #m.add_child(folium.Circle(location=[p_lat,p_lon]))
        m.add_child(folium.Circle(location=[d_lat,d_lon],color = '#FF0000'))
        #pas_hour += sum(cab_1st_day[cab_1st_day['pickup_datetime'] == date_time].passenger_count) # 加该时间短的人数
    else:
        #pas_hours.append(pas_hour)
        #pas_hour = 0
        display(m)
        #time.sleep(4)
        lj = str(j)
        fn = lj + '.html'
        # fn='testmap.html'
        tmpurl='file://{path}/{mapfile}'.format(path=os.getcwd(),mapfile=fn)
        m.save(fn)
        browser = webdriver.Safari()
        browser.get(tmpurl)
        #Give the map tiles some time to load
        time.sleep(4)
        browser.save_screenshot(lj + '.jpg')
        browser.quit()
        date_point -= datetime.timedelta(hours=1)# 1个小时作为一个间隔
        #m.add_child(folium.Circle(location=[p_lat,p_lon]))
        m.add_child(folium.Circle(location=[d_lat,d_lon],color = '#FF0000'))
        j += 1
        # pas_hour += sum(cab_1st_day[cab_1st_day['pickup_datetime'] == date_time].passenger_count) # 加该时间短的人数

好看多了

5.3 制作gif

from PIL import Image, ImageSequence
import sys, os
filenames=sorted(fn for fn in os.listdir('.') if fn.endswith('.jpg')) # 获取文件并排序        
filenames.sort(key = lambda f: int(filter(str.isdigit, f))) # sort by int
# 再次按照int排序：保证排序       
im=Image.open('1.jpg')
print(filenames)
im.save('traffic_1st_day_drop.gif', save_all=True, append_images= [Image.open(filename) for filename in filenames],loop=5,duration=500)