用python对汽车油耗进行数据分析（anaconda python3.6完全跑通）

小莹莹

发布于 2018-04-23 16:14:48

9900

发布于 2018-04-23 16:14:48

编者按：前两天我们微信发了一篇文章《用python对汽车油耗进行数据分析》，有一网友学习后用python3.6重新跑了一下数据，请大家比较阅读。PPV课致力于为大家提供一个开放、分享、进步的数据科学社区，欢迎小伙伴们给我们分享你的学习心得和博客文章，有你的鼓励和支持，我们会做的更好。投稿请联系QQ149104196。

正文：

1.下载汽车油耗数据集并解压

下载地址：https://www.fueleconomy.gov/feg/download.shtml

vehiclesData.py:

#encoding = utf-8import pandas as pdimport numpy as npfrom ggplot import *import matplotlib.pyplot as plt
vehicles = pd.read_csv("../data/vehicles.csv")print(vehicles.head())print(len(vehicles))#- 查看有多少观测点（行）和多少变量（列）

运行结果：

barrels08 barrelsA08 charge120 charge240 city08 city08U cityA08 \ 0 15.695714 0.0 0.0 0.0 19 0.0 0 1 29.964545 0.0 0.0 0.0 9 0.0 0 2 12.207778 0.0 0.0 0.0 23 0.0 0 3 29.964545 0.0 0.0 0.0 10 0.0 0 4 17.347895 0.0 0.0 0.0 17 0.0 0 cityA08U cityCD cityE ... mfrCode c240Dscr charge240b \ 0 0.0 0.0 0.0 ... NaN NaN 0.0 1 0.0 0.0 0.0 ... NaN NaN 0.0 2 0.0 0.0 0.0 ... NaN NaN 0.0 3 0.0 0.0 0.0 ... NaN NaN 0.0 4 0.0 0.0 0.0 ... NaN NaN 0.0 c240bDscr createdOn modifiedOn \ 0 NaN Tue Jan 01 00:00:00 EST 2013 Tue Jan 01 00:00:00 EST 2013 1 NaN Tue Jan 01 00:00:00 EST 2013 Tue Jan 01 00:00:00 EST 2013 2 NaN Tue Jan 01 00:00:00 EST 2013 Tue Jan 01 00:00:00 EST 2013 3 NaN Tue Jan 01 00:00:00 EST 2013 Tue Jan 01 00:00:00 EST 2013 4 NaN Tue Jan 01 00:00:00 EST 2013 Tue Jan 01 00:00:00 EST 2013 startStop phevCity phevHwy phevComb 0 NaN 0 0 0 1 NaN 0 0 0 2 NaN 0 0 0 3 NaN 0 0 0 4 NaN 0 0 0 [5 rows x 83 columns] 39101

其中 pandas中Data Frame类的边界方法head，查看一个很有用的数据框data frame的中，包括每列的非空值数量和各列不同的数据类型的数量。

描述汽车油耗等数据

print(len(vehicles.columns))print(vehicles.columns)

83 Index(['barrels08', 'barrelsA08', 'charge120', 'charge240', 'city08', 'city08U', 'cityA08', 'cityA08U', 'cityCD', 'cityE', 'cityUF', 'co2', 'co2A', 'co2TailpipeAGpm', 'co2TailpipeGpm', 'comb08', 'comb08U', 'combA08', 'combA08U', 'combE', 'combinedCD', 'combinedUF', 'cylinders', 'displ', 'drive', 'engId', 'eng_dscr', 'feScore', 'fuelCost08', 'fuelCostA08', 'fuelType', 'fuelType1', 'ghgScore', 'ghgScoreA', 'highway08', 'highway08U', 'highwayA08', 'highwayA08U', 'highwayCD', 'highwayE', 'highwayUF', 'hlv', 'hpv', 'id', 'lv2', 'lv4', 'make', 'model', 'mpgData', 'phevBlended', 'pv2', 'pv4', 'range', 'rangeCity', 'rangeCityA', 'rangeHwy', 'rangeHwyA', 'trany', 'UCity', 'UCityA', 'UHighway', 'UHighwayA', 'VClass', 'year', 'youSaveSpend', 'guzzler', 'trans_dscr', 'tCharger', 'sCharger', 'atvType', 'fuelType2', 'rangeA', 'evMotor', 'mfrCode', 'c240Dscr', 'charge240b', 'c240bDscr', 'createdOn', 'modifiedOn', 'startStop', 'phevCity', 'phevHwy', 'phevComb'], dtype='object')

#查看年份信息print(len(pd.unique(vehicles.year)))#总共的年份print(min(vehicles.year))#最小年份print(max(vehicles.year))#最大的年份

35 1984 2018

#查看变速箱类型print(pd.value_counts(vehicles.trany))#trany变量自动挡是以A开头，手动挡是以M开头；故创建一个新变量trany2：vehicles["trany2"] = vehicles.trany.str[0]print(pd.value_counts(vehicles.trany2))

Automatic 4-spd 11045 Manual 5-spd 8337 Automatic 3-spd 3151 Automatic (S6) 2857 Manual 6-spd 2563 Automatic 5-spd 2191 Automatic 6-spd 1521 Manual 4-spd 1483 Automatic (S8) 1181 Automatic (S5) 830 Automatic (variable gear ratios) 730 Automatic 7-spd 695 Automatic 8-spd 323 Automatic (AM-S7) 322 Automatic (S7) 288 Automatic (S4) 233 Automatic (AM7) 194 Automatic (AV-S6) 172 Automatic 9-spd 170 Automatic (A1) 134 Automatic (AM6) 122 Automatic (AM-S6) 106 Automatic (AV-S7) 95 Manual 7-spd 93 Manual 3-spd 77 Automatic (S9) 40 Automatic (AV-S8) 31 Automatic (S10) 30 Automatic (AM-S8) 26 Manual 4-spd Doubled 17 Automatic (AM5) 14 Automatic 10-spd 8 Automatic (AM8) 5 Automatic (L4) 2 Automatic (L3) 2 Automatic (AV-S10) 1 Automatic (AM-S9) 1 Name: trany, dtype: int64 A 26520 M 12570 Name: trany2, dtype: int64

同理可以查看其它特征数据

#分析汽车油耗随时间变化的趋势#- 先按照年份分组grouped = vehicles.groupby('year')#- 再计算其中三列的均值averaged= grouped['comb08', 'highway08', 'city08'].agg([np.mean])#- 为方便分析，对其进行重命名，然后创建一个‘year’的列，包含该数据框data frame的索引averaged.columns = ['comb08_mean', 'highwayo8_mean', 'city08_mean']
averaged['year'] = averaged.indexprint(averaged )

comb08_mean highwayo8_mean city08_mean year year 1984 19.881874 23.075356 17.982688 1984 1985 19.808348 23.042328 17.878307 1985 1986 19.550413 22.699174 17.665289 1986 1987 19.228549 22.445068 17.310345 1987 1988 19.328319 22.702655 17.333628 1988 1989 19.125759 22.465742 17.143972 1989 1990 19.000928 22.337662 17.033395 1990 1991 18.825972 22.253534 16.848940 1991 1992 18.862623 22.439786 16.805531 1992 1993 19.104300 22.780421 16.998170 1993 1994 19.012220 22.725051 16.918534 1994 1995 18.797311 22.671148 16.569804 1995 1996 19.584735 23.569211 17.289780 1996 1997 19.429134 23.451444 17.135171 1997 1998 19.518473 23.546798 17.113300 1998 1999 19.611502 23.552817 17.272300 1999 2000 19.526190 23.414286 17.221429 2000 2001 19.479693 23.328211 17.275521 2001 2002 19.168205 23.030769 16.893333 2002 2003 19.000958 22.836207 16.780651 2003 2004 19.067736 23.064171 16.740642 2004 2005 19.193825 23.297599 16.851630 2005 2006 18.959239 23.048913 16.626812 2006 2007 18.978686 23.083481 16.605684 2007 2008 19.276327 23.455771 16.900590 2008 2009 19.735195 24.017766 17.335025 2009 2010 20.589883 24.949413 18.106594 2010 2011 21.011525 25.170213 18.670213 2011 2012 21.820870 26.106957 19.365217 2012 2013 23.126164 27.502117 20.663844 2013 2014 23.518946 27.953871 21.029654 2014 2015 24.031471 28.568057 21.446105 2015 2016 25.151878 29.604317 22.597122 2016 2017 25.089634 29.418550 22.583788 2017 2018 23.396825 28.056689 20.692744 2018

#- 使用ggplot包将结果绘成散点图allCarPlt = ggplot(averaged, aes('year', 'comb08_mean')) + geom_point(colour='steelblue') + xlab("Year") + ylab("Average MPG") + ggtitle("All cars")print(allCarPlt)

#- 去除混合动力汽车criteria1 = vehicles.fuelType1.isin(['Regular Gasoline', 'Premium Gasoline', 'Midgrade Gasoline'])
criteria2 = vehicles.fuelType2.isnull()
criteria3 = vehicles.atvType != 'Hybrid'vehicles_non_hybrid = vehicles[criteria1 & criteria2 & criteria3]#print(vehicles_non_hybrid)#- 将得到的数据框data frame按年份分组，并计算平均油耗grouped = vehicles_non_hybrid.groupby(['year'])
averaged = grouped['comb08'].agg([np.mean])
averaged['hahhahah'] = averaged.indexprint(averaged)

mean hahhahah year 1984 19.121622 1984 1985 19.394686 1985 1986 19.320457 1986 1987 19.164568 1987 1988 19.367607 1988 1989 19.141964 1989 1990 19.031459 1990 1991 18.838060 1991 1992 18.861566 1992 1993 19.137383 1993 1994 19.092632 1994 1995 18.872591 1995 1996 19.530962 1996 1997 19.368000 1997 1998 19.329545 1998 1999 19.239759 1999 2000 19.169345 2000 2001 19.075058 2001 2002 18.950270 2002 2003 18.761711 2003 2004 18.967339 2004 2005 19.005510 2005 2006 18.786398 2006 2007 18.987512 2007 2008 19.191781 2008 2009 19.738095 2009 2010 20.466736 2010 2011 20.838219 2011 2012 21.407328 2012 2013 22.228877 2013 2014 22.279835 2014 2015 22.418539 2015 2016 22.742509 2016 2017 22.817854 2017 2018 22.911504 2018

#- 查看是否大引擎的汽车越来越少print(pd.unique(vehicles_non_hybrid.displ))

[ 2. 4.9 2.2 5.2 1.8 1.6 2.3 2.8 4. 5. 3.3 3.1 3.8 4.6 3.4 3. 5.9 2.5 4.5 6.8 2.4 2.9 5.7 4.3 3.5 5.8 3.2 4.2 1.9 2.6 7.4 3.9 1.5 1.3 4.1 8. 6. 3.6 5.4 5.6 1. 2.1 1.2 6.5 2.7 4.7 5.5 1.1 5.3 4.4 3.7 6.7 4.8 1.7 6.2 8.3 1.4 6.1 7. 8.4 6.3 nan 6.6 6.4 0.9]

#- 去掉nan值，并用astype方法保证各个值都是float型的criteria = vehicles_non_hybrid.displ.notnull()
vehicles_non_hybrid = vehicles_non_hybrid[criteria]
vehicles_non_hybrid.loc[:,'displ'] = vehicles_non_hybrid.displ.astype('float')
criteria = vehicles_non_hybrid.comb08.notnull()
vehicles_non_hybrid = vehicles_non_hybrid[criteria]
vehicles_non_hybrid.loc[:,'comb08'] = vehicles_non_hybrid.comb08.astype('float')#- 最后用ggplot包来绘图gasOnlineCarsPlt =  ggplot(vehicles_non_hybrid, aes('displ', 'comb08')) + geom_point(color='steelblue') +xlab('Engine Displacement') + ylab('Average MPG') + ggtitle('Gasoline cars')print(gasOnlineCarsPlt)

#- 查看是否平均起来汽车越来越少了grouped_by_year = vehicles_non_hybrid.groupby(['year'])
avg_grouped_by_year = grouped_by_year['displ', 'comb08'].agg([np.mean])#- 计算displ和conm08的均值，并改造数据框data frameavg_grouped_by_year['year'] = avg_grouped_by_year.index
melted_avg_grouped_by_year = pd.melt(avg_grouped_by_year, id_vars='year')#- 创建分屏绘图p = ggplot(aes(x='year', y='value', color = 'variable_0'), data=melted_avg_grouped_by_year)
p + geom_point() + facet_grid("variable_0",scales="free") #scales参数fixed表示固定坐标轴刻度，free表示反馈坐标轴刻度print(p)

调查汽车的制造商和型号

#- 首先查看cylinders变量有哪些可能的值print(pd.unique(vehicles_non_hybrid.cylinders))

[ 4. 12. 8. 6. 5. 10. 2. 3. 16. nan]

#- 再将cylinders变量转换为float类型，这样可以轻松方便地找到data frame的子集vehicles_non_hybrid.cylinders = vehicles_non_hybrid.cylinders.astype('float')
pd.unique(vehicles_non_hybrid.cylinders)#- 现在，我们可以查看各个时间段有四缸引擎汽车的品牌数量vehicles_non_hybrid_4 = vehicles_non_hybrid[(vehicles_non_hybrid.cylinders==4.0)]

grouped_by_year_4_cylinder =vehicles_non_hybrid_4.groupby(['year']).make.nunique()#fig = grouped_by_year_4_cylinder.plot()plt.plot(grouped_by_year_4_cylinder)
plt.xlabel("Year")
plt.ylabel("Number of 4-Cylinder Maker")
plt.show()

分析：

我们可以从上图中看到，从1985年以来四缸引擎汽车的品牌数量呈下降趋势。然而，需要注意的是，这张图可能会造成误导，因为我们并不知道汽车品牌总数是否在同期也发生了变化。为了一探究竟，我们继续一下操作。

# -  查看各年有四缸引擎汽车的品牌的列表，找出每年的品牌列表grouped_by_year_4_cylinder = vehicles_non_hybrid_4.groupby(['year'])
unique_makes = []from functools import reducefor name, group in grouped_by_year_4_cylinder:    # list中存入set(),set里包含每年中的不同品牌    unique_makes.append(set(pd.unique(group['make'])))
unique_makes = reduce(set.intersection, unique_makes)print(unique_makes)

{'Nissan', 'Toyota', 'Chevrolet', 'Honda', 'Mazda', 'Volkswagen', 'Dodge', 'Subaru', 'Ford', 'Jeep'}

由上可知：在此期间只有10家制造商每年都制造四缸引擎汽车。

接下来，我们去发现这些汽车生产商的型号随时间的油耗表现。

#创建一个空列表，最终用来产生布尔值Booleansboolean_mask = []#用iterrows生成器generator遍历data frame中的各行来产生每行及索引for index, row in vehicles_non_hybrid_4.iterrows():    #判断每行的品牌是否在此前计算的unique_makes集合中,在将此布尔值Blooeans添加在Booleans_mask集合后面    make = row['make']
    boolean_mask.append(make in unique_makes)
df_common_makes = vehicles_non_hybrid_4[boolean_mask]#-  先将数据框data frame按year和make分组，然后计算各组的均值df_common_makes_grouped = df_common_makes.groupby(['year', 'make']).agg(np.mean).reset_index()#-  最后利用ggplot提供的分屏图来显示结果oilWithTime = ggplot(aes(x='year', y='comb08'), data = df_common_makes_grouped) + geom_line() + facet_wrap('make')print(oilWithTime)