用Python处理"大"XLS文件

权当学习Python练手用的.

数据来data.gov.uk,大小有58.4MB

• 文件都是些什么内容?

  • ’Accident_Index’,
  • ‘Location_Easting_OSGR’,
  • ‘Location_Northing_OSGR’,
  • ‘Longitude’,
  • ‘Latitude’,
  • ‘Police_Force’,
  • ‘Accident_Severity’,
  • ‘Number_of_Vehicles’,
  • ‘Number_of_Casualties’,
  • ‘Date’,
  • ‘Day_of_Week’,
  • ‘Time’,
  • ‘Local_Authority_(District)’,
  • ‘Local_Authority_(Highway)’,
  • ‘1st_Road_Class’, ‘1st_Road_Number’,
  • ‘Road_Type’,
  • ‘Speed_limit’,
  • ‘Junction_Detail’,
  • ‘Junction_Control’,
  • ‘2nd_Road_Class’,
  • ‘2nd_Road_Number’,
  • ‘Pedestrian_Crossing-Human_Control’,
  • ‘Pedestrian_Crossing_Physical_Facilities’,
    
    • ’Light_Conditions’,
    • ‘Weather_Conditions’,
    • ‘Road_Surface_Conditions’,
    • ‘Special_Conditions_at_Site’,
    • ‘Carriageway_Hazards’,
    • ‘Urban_or_Rural_Area’,
    • ‘Did_Police_Officer_Attend_Scene_of_Accident’,
    • ‘LSOA_of_Accident_Location’

  

LowMemory 方式读取文件

#read the filefiledir='/home/derek/Desktop/python-data-analyis/large-excel-files/Accidents_2013.csv'data = pd.read_csv(filedir,low_memory=False)print data.ix[:10]['Day_of_Week']

• SQL likes 提取数据信息

print 'Accidents'print '----------'#选择星期日发生的事故accidents_sunday = data[data.Day_of_Week==1]print 'Accidents which happended on a Sunday: ',len(accidents_sunday)#选择星期日发生的且涉事人数在十人以上的事故accidents_sunday_twenty_cars = data[(data.Day_of_Week==1) & (data.Number_of_Vehicles>10)]print'Accidents which happened on a Sunday involving > 10 cars: ' , len(accidents_sunday_twenty_cars)#选择星期日发生的且涉事人数在十人以上且天气情况是下雨的事故(2对应的是无风下雨)accidents_sunday_twenty_cars_rain = data[(data.Day_of_Week==1) & (data.Number_of_Vehicles>10) & (data.Weather_Conditions==2)]print'Accidents which happened on a Sunday involving > 10 cars with rainning: ' , len(accidents_sunday_twenty_cars_rain)#选择在伦敦的星期日发生的事故london_data = data[(data['Police_Force'] == 1) & (data.Day_of_Week==1)]print 'Accidents in London on a Sunday',len(london_data)#选择在2000年的伦敦的星期日发生的事故london_data_2000 = london_data[((pd.to_datetime('2000-1-1', errors='coerce')) > (pd.to_datetime(london_data['Date'],errors='coerce'))) & (pd.to_datetime(london_data['Date'],errors='coerce') < (pd.to_datetime('2000-12-31', errors='coerce')))]print 'Accidents in London on a Sunday in 2000:',len(london_data_2000)

给人的感觉是特别像SQL语句,DataFrame的这种切片,方式特别好用,对不对?

pd.to_datetime(london_data['Date'],errors='coerce')

这里是日期转换函数.

输出:

Accidents----------Accidents which happended on a Sunday:  14854Accidents which happened on a Sunday involving > 10 cars:  1Accidents which happened on a Sunday involving > 10 cars with rainning:  1Accidents in London on a Sunday 2374Accidents in London on a Sunday in 2000: 0

• 将部分DataFrame数据以XLSX文件存储下来
  确保你安装了XlsxWriter
  

sudo pip install XlsxWriter

writer = pd.ExcelWriter('london_data.xlsx', engine='xlsxwriter')london_data.to_excel(writer, 'sheet1')writer.save()writer.close()

• 块读取,分析一个星期中那一天最有出事故的概率最大
  代码.2013,2014,2015三年的事故记录,在’Accidents_2013.csv’,’Accidents_2014.csv’, ‘Accidents_2015.csv’这三个文件中

import pandas as pdfrom pandas import Seriesimport matplotlib.pyplot as plt#read the filedir='/home/derek/Desktop/python-data-analyis/large-excel-files/'filedir=['Accidents_2013.csv','Accidents_2014.csv', 'Accidents_2015.csv']tot = Series([])for i in range(3):    #块读取文件, 每次读1000条记录    data = pd.read_csv(dir + filedir[i],chunksize=1000)    for piece in data:        tot = tot.add(piece['Day_of_Week'].value_counts(), fill_value=0)day_index = ['Sun', 'Mon', 'Tues', 'Wed', 'Thur', 'Fri', 'Sat']print 'data like:'#tot = tot.sort_values(ascending=False)print tot#重新构造一个Series,是为了给索引命名new_Series = Series(tot.values, index=day_index)new_Series.plot()plt.show()plt.close()

控制台输出:

data like:1    460522    609563    650064    640395    644456    693787    55162dtype: float64

图:

三年记录在案的有425038条记录.

结论: 看来,英国人在工作日出行要比在休息日造成更多的事故.星期五的出行造成的事故最多,或许,星期五急着回家,哈哈.相比起来,星期五不适合外出.

参考文章来源

文件没有提供,是因为:读者可以自己去下载,可能找到更想更好用Python分析的数据.