泰坦尼克号幸存者状况分析

星星在线

发布于 2020-05-22 15:06:06

8740

发布于 2020-05-22 15:06:06

以下练习数据来源均为seaborn库中提供，通过网络获取，如果出现网络获取慢或者失败的情况，可以到GitHub上搜索seaborn-data，下载后传入读取路径即可

from matplotlib import pyplot as plt
import seaborn as sns
import numpy as np

泰坦尼克号海难幸存状况分析

data = sns.load_dataset("titanic", data_home='/Volumes/Code/notebooks/seaborn-data')
data.head()

不同仓位等级中幸存和遇难的乘客比例

pclasses = []
    surviveds = [[], []]
    # 按等级分组，然后计算不同幸存者的数量
    for pclass, items in data.groupby(by=['class']):
        pclasses.append(pclass)
        count0 = items[items['survived'] == 0]['survived'].count()
        count1 = items[items['survived'] == 1]['survived'].count()
        surviveds[0].append(count0)
        surviveds[1].append(count1)

    # 分别绘制不同幸存者图
    plt.bar(pclasses, surviveds[0], color='r', width=0.3)
    plt.bar(pclasses, surviveds[1], bottom=surviveds[0], color='g', width=0.3)

    # 增加文字说明
    for i, pclass in enumerate(pclasses):
        total = surviveds[0][i] + surviveds[1][i]
        plt.text(pclass, surviveds[0][i] // 2, '%.2f%%' % ((surviveds[0][i] / total) * 100), ha='center')
        plt.text(pclass, surviveds[0][i] + surviveds[1][i] // 2, '%.2f%%' % ((surviveds[1][i] / total) * 100), ha='center')
    plt.xticks(pclasses, pclasses)
    plt.ylim([0, 600])
    plt.legend(['die', 'survive'], loc='upper right')
    plt.grid(axis='y', color='gray', linestyle=':', linewidth=2)
    plt.show()

从图中可以看出，低等舱的人死亡最多

不同性别的幸存比例

换一种统计方式

    temp = data.groupby(by=['survived', 'sex']).count()

    # 非幸存者女性
    t1 = temp.loc[0, :].loc['female', :].max()
    # 非幸存者男性
    t2 = temp.loc[0, :].loc['male', :].max()
    # 幸存者女性
    t3 = temp.loc[1, :].loc['female', :].max()
    # 幸存者男性
    t4 = temp.loc[1, :].loc['male', :].max()

    sexs = ['female', 'male']
    plt.bar(sexs, [t1, t2], color='r', width=0.3)
    plt.bar(sexs, [t3, t4], bottom=[t1, t2], color='g', width=0.3)

    survived = {'female': [t1, t3], 'male': [t2, t4]}
    for i, pclass in enumerate(sexs):
        total = sum(survived[pclass])
        plt.text(pclass, survived[pclass][0] // 2, '%.2f%%' % ((survived[pclass][0] / total) * 100), ha='center')
        plt.text(pclass, survived[pclass][0] + survived[pclass][1] // 2, '%.2f%%' % ((survived[pclass][1] / total) * 100),
                 ha='center')
    plt.xticks(sexs, sexs)
    plt.ylim([0, 600])
    plt.legend(['die', 'survive'], loc='upper left')
    plt.grid(axis='y', color='gray', linestyle=':', linewidth=2)
    plt.show()

从图中可以看出，女性幸存者比例还是很大的，果然妇女儿童先行

幸存和遇难乘客的票价分布

    surviveds = []
    fares = []
    for survived, items in data.groupby(by=['survived']):
        surviveds.append(survived)
        fares.append(items['fare'])
    str_sur = ['survived' if i else 'die' for i in surviveds]
    plt.boxplot(x=fares, patch_artist=True, labels=str_sur, showmeans=True,
                medianprops={'linestyle': '--', 'color': 'orange'})
    plt.show()

整体来讲，存活的人票价相对都高一些，这也解释了上面低等舱的人死亡过多的情况

幸存和遇难乘客的年龄分布

    surviveds = []
    fares = []

    # 删除年龄为NaN的行数据
    temp = data.dropna(subset=['age'], how='any')
    for survived, items in temp.groupby(by=['survived']):
        surviveds.append(survived)
        fares.append(items['age'])

    str_sur = ['survived' if i else 'die' for i in surviveds]
    plt.boxplot(x=fares, patch_artist=True, labels=str_sur, showmeans=True,
                medianprops={'linestyle': '--', 'color': 'orange'})

    plt.show()

幸存者和死亡者平均年龄基本相同

不同上船港口的乘客仓位等级分布

    embarkeds = []
    pclasses = []
    for embarked, items in data.groupby(by=['embarked']):
        embarkeds.append(embarked)
        pclasss = []
        for pclass, values in items.groupby(by=['class']):
            pclasss.append(values['class'].count())
        pclasses.append(pclasss)

    width = 0.25
    x = np.array(range(len(embarkeds)))
    plt.bar(x-width, [i[0] for i in pclasses], width=0.2, color='r')
    plt.bar(x, [i[1] for i in pclasses], width=0.2, color='g')
    plt.bar(x+width, [i[2] for i in pclasses], width=0.2, color='b')

    for i, val in enumerate(x):
        item = pclasses[i]
        print(item)
        plt.text(val-width, item[0] + 5, '%s' % item[0], ha='center')
        plt.text(val, item[1] + 5, '%s' % item[1], ha='center')
        plt.text(val+width, item[2] + 5, '%s' % item[2], ha='center')
    plt.xticks(x, embarkeds)
    plt.legend(['First', 'Second', 'Third'])
    plt.grid(axis='y', color='gray', linestyle=':', linewidth=2)
    plt.xlabel('embarked')
    plt.ylabel('pclass')
    plt.show()

幸存和遇难乘客堂兄弟姐妹的数量分布

sibsps = []
    survived = []
    for survive, items in data.groupby(by=['survived']):
        survived.append(survive)
        sibsps.append(items['sibsp'])

    plt.boxplot(x=sibsps, patch_artist=True, labels=survived, showmeans=True,
                medianprops={'linestyle': '--', 'color': 'orange'})
    plt.xlabel('survived')
    plt.ylabel('sibsp')
    plt.show()

幸存和遇难乘客父母子女的数量分布

    sibsps = []
    survived = []
    for survive, items in data.groupby(by=['survived']):
        survived.append(survive)
        sibsps.append(items['parch'])

    plt.boxplot(x=sibsps, patch_artist=True, labels=survived, showmeans=True,
                medianprops={'linestyle': '--', 'color': 'orange'})
    plt.xlabel('alive')
    plt.ylabel('parch')
    plt.show()

单独乘船与否和幸存之间有没有联系

    survived = np.array([0, 1])
    temp = data.groupby(by=['alone', 'survived']).count()

    # 单独上传并且没有幸存
    t1 = temp.loc[False, :].loc[0, :].max()
    t2 = temp.loc[False, :].loc[1, :].max()
    t3 = temp.loc[True, :].loc[0, :].max()
    t4 = temp.loc[True, :].loc[1, :].max()

    width = 0.1
    plt.bar(survived - width, [t1, t3], color='r', width=width*2)
    plt.bar(survived + width, [t2, t4], color='g', width=width*2)

    plt.xlabel('Alone')
    plt.ylabel('Alive')
    plt.xticks(survived, ['No Alone', 'Alone'])
    plt.legend(['die', 'survive'])
    plt.show()

从图中可以看出，非单独乘船的死亡率和存活率相当，单独乘船的死亡率要高于存活率

是否成年男性和幸存之间有没有联系

    temp = data.groupby(by=['adult_male', 'alive']).count()

    # 非成年死亡者
    t1 = temp.loc[False, :].loc['no', :].max()
    # 非成年幸存者
    t2 = temp.loc[False, :].loc['yes', :].max()
    # 成年死亡者
    t3 = temp.loc[True, :].loc['no', :].max()
    # 成年幸存者
    t4 = temp.loc[True, :].loc['yes', :].max()

    sexs = ['Not Adult', 'Adult']
    plt.bar(sexs, [t1, t3], color='r', width=0.3)
    plt.bar(sexs, [t2, t4], bottom=[t1, t3], color='g', width=0.3)

    survived = {'Not Adult': [t1, t2], 'Adult': [t3, t4]}
    for i, pclass in enumerate(sexs):
        total = sum(survived[pclass])
        plt.text(pclass, survived[pclass][0] // 2, '%.2f%%' % ((survived[pclass][0] / total) * 100), ha='center')
        plt.text(pclass, survived[pclass][0] + survived[pclass][1] // 2,
                 '%.2f%%' % ((survived[pclass][1] / total) * 100),
                 ha='center')
    plt.xticks(sexs, sexs)
    plt.ylim([0, 600])
    plt.legend(['die', 'survive'], loc='upper left')
    plt.grid(axis='y', color='gray', linestyle=':', linewidth=2)
    plt.show()