空难数据分析例子

数据集：Kaggle上的1908年收集的公开数据集

项目任务：

• 每年空难数分析
• 机上乘客数量
• 生还数、遇难数
  
  • 哪些航空公司空难数最多?
  • 哪些机型空难数最多?

# -*-coding: utf-8 -*-
import pandas as pd
import seaborn as sns
import matplotlib.pyplot as plt
from bokeh.io import output_notebook, output_file, show
from bokeh.charts import Bar,TimeSeries
from bokeh.layouts import column
from math import pi

- 查看数据信息

data_path = './dataset/Airplane_Crashes_and_Fatalities_Since_1908.csv'
df_data = pd.read_csv(data_path)

print u'数据集基本信息：'
print df_data.info()

数据集基本信息： RangeIndex: 5268 entries, 0 to 5267 Data columns (total 13 columns): Date 5268 non-null object Time 3049 non-null object Location 5248 non-null object Operator 5250 non-null object Flight # 1069 non-null object Route 3562 non-null object Type 5241 non-null object Registration 4933 non-null object cn/In 4040 non-null object Aboard 5246 non-null float64 Fatalities 5256 non-null float64 Ground 5246 non-null float64 Summary 4878 non-null object dtypes: float64(3), object(10) memory usage: 535.1+ KB None

print u'数据集有%i行，%i列' %(df_data.shape[0], df_data.shape[1])

数据集有5268行，13列

print u'数据预览：'
df_data.head()

数据预览：

	Date	Time	Location	Operator	Flight #	Route	Type	Registration	cn/In	Aboard	Fatalities	Ground	Summary
0	09/17/1908	17:18	Fort Myer, Virginia	Military - U.S. Army	NaN	Demonstration	Wright Flyer III	NaN	1	2.0	1.0	0.0	During a demonstration flight, a U.S. Army fly…
1	07/12/1912	06:30	AtlantiCity, New Jersey	Military - U.S. Navy	NaN	Test flight	Dirigible	NaN	NaN	5.0	5.0	0.0	First U.S. dirigible Akron exploded just offsh…
2	08/06/1913	NaN	Victoria, British Columbia, Canada	Private	-	NaN	Curtiss seaplane	NaN	NaN	1.0	1.0	0.0	The first fatal airplane accident in Canada oc…
3	09/09/1913	18:30	Over the North Sea	Military - German Navy	NaN	NaN	Zeppelin L-1 (airship)	NaN	NaN	20.0	14.0	0.0	The airship flew into a thunderstorm and encou…
4	10/17/1913	10:30	Near Johannisthal, Germany	Military - German Navy	NaN	NaN	Zeppelin L-2 (airship)	NaN	NaN	30.0	30.0	0.0	Hydrogen gas which was being vented was sucked…

• 处理缺失数据

# def process_missing_data(df_data):
#     """
#             处理缺失数据
#     """
#     if df_data.isnull().values.any():
#         # 存在缺失数据
#         print '存在缺失数据！'
#         df_data = df_data.fillna(0.)    # 填充nan
#         # df_data = df_data.dropna()    # 过滤nan
#     return df_data.reset_index()

- 数据转换

df_data['Date'] = pd.to_datetime(df_data['Date'])
# df_data['Date']

df_data['Year'] = df_data['Date'].map(lambda x: x.year)
df_data.head()

	Date	Time	Location	Operator	Flight #	Route	Type	Registration	cn/In	Aboard	Fatalities	Ground	Summary	Year
0	1908-09-17	17:18	Fort Myer, Virginia	Military - U.S. Army	NaN	Demonstration	Wright Flyer III	NaN	1	2.0	1.0	0.0	During a demonstration flight, a U.S. Army fly…	1908
1	1912-07-12	06:30	AtlantiCity, New Jersey	Military - U.S. Navy	NaN	Test flight	Dirigible	NaN	NaN	5.0	5.0	0.0	First U.S. dirigible Akron exploded just offsh…	1912
2	1913-08-06	NaN	Victoria, British Columbia, Canada	Private	-	NaN	Curtiss seaplane	NaN	NaN	1.0	1.0	0.0	The first fatal airplane accident in Canada oc…	1913
3	1913-09-09	18:30	Over the North Sea	Military - German Navy	NaN	NaN	Zeppelin L-1 (airship)	NaN	NaN	20.0	14.0	0.0	The airship flew into a thunderstorm and encou…	1913
4	1913-10-17	10:30	Near Johannisthal, Germany	Military - German Navy	NaN	NaN	Zeppelin L-2 (airship)	NaN	NaN	30.0	30.0	0.0	Hydrogen gas which was being vented was sucked…	1913

• 数据分析与可视化——空难数vs年份

a) seaborn

plt.figure(figsize=(15.0,10.0))
sns.countplot(x='Year', data=df_data)

plt.rcParams['font.sans-serif']=['SimHei']
plt.rcParams['axes.unicode_minus']=False

plt.title(u'空难次数VS年份')
plt.xlabel(u'年份')
plt.ylabel(u'空难次数')
plt.xticks(rotation=90)
plt.show()

b) bokeh

p = Bar(df_data,'Year',title=u'空难次数 VS 年份',plot_width=1000,legend=False,xlabel=u'年份',ylabel=u'空难次数')
p.xaxis.major_label_orientation = pi/2
output_notebook()
show(p)

• 数据分析与可视化——乘客数量vs遇难数vs年份

grouped_year_sum_data = df_data.groupby('Year',as_index=False).sum()
grouped_year_sum_data.head()

	Year	Aboard	Fatalities	Ground
0	1908	2.0	1.0	0.0
1	1912	5.0	5.0	0.0
2	1913	51.0	45.0	0.0
3	1915	60.0	40.0	0.0
4	1916	109.0	108.0	0.0

a) seaborn

grouped_year_sum_data = df_data.groupby('Year',as_index=False).sum()
grouped_year_sum_data.head()

plt.title(u'乘客数量vs遇难数vs年份')
plt.xlabel(u'年份')
plt.ylabel(u'乘客数量vs遇难数')
plt.xticks(rotation=90)
plt.show()

b) bokeh

tsline = TimeSeries(data=grouped_year_sum_data, x='Year',y=['Aboard','Fatalities'],color=['Aboard', 'Fatalities'],
                    dash=['Aboard', 'Fatalities'],title=u'乘客数量vs遇难数vs年份',xlabel=u'年份',ylabel=u'乘客数vs遇难数',
                    legend=True)
tspoint = TimeSeries(data=grouped_year_sum_data,x='Year',y=['Aboard','Fatalities'],color=['Aboard', 'Fatalities'],
                    dash=['Aboard', 'Fatalities'],builder_type='point',title=u'乘客数量vs遇难数vs年份',xlabel=u'年份',
                    ylabel=u'乘客数vs遇难数',legend=True)
output_notebook()
show(column(tsline,tspoint))

• top n 分析

grouped_data = df_data.groupby(by='Type',as_index=False)['Date'].count()
grouped_data.rename(columns={'Date':'Count'},inplace=True)
top_n = 10
top_n_grouped_data = grouped_data.sort_values('Count',ascending=False).iloc[:top_n, :]
top_n_grouped_data

	Type	Count
1178	Douglas DC-3	334
2388	de Havilland Canada DHC-6 Twin Otter 300	81
1097	Douglas C-47A	74
1089	Douglas C-47	62
1230	Douglas DC-4	40
2340	Yakovlev YAK-40	37
125	Antonov AN-26	36
1598	Junkers JU-52/3m	32
1119	Douglas C-47B	29
1045	De Havilland DH-4	28

- 可视化结果

plt.figure(figsize=(15.0,10.0))
sns.barplot(x='Count',y='Type',data=top_n_grouped_data)

plt.title('Count vs Type',fontsize=20)
plt.xlabel('Type')
plt.ylabel('Count')
plt.show()