Surrender_Analysis/__data1__createInitPortfolio.py at master · mtkier94/Surrender_Analysis · GitHub

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
import numpy as np
import seaborn as sns
import pandas as pd
import matplotlib.pyplot as plt
import scipy as sc
import os

from functions.sub_simulate_events import simulate_contracts
from global_vars import path_plots, path_portfolio


def run_simulation():

    N_contracts = 30000
    Portfolio = simulate_contracts( N=N_contracts, option_new_business=False)
    Portfolio.to_csv(os.path.join(path_portfolio, r'Portfolio.csv'), header=True)


def visualize_portfolio():

    try:
        Portfolio = pd.read_csv(os.path.join(path_portfolio,r'Portfolio.csv'), index_col=0)
    except:
        raise ValueError('Portfolio cannot be loaded!')


    ## marginal distributions of relevant contract features
    df = Portfolio.drop(['Death', 'Time', 'Lapsed', 'Age_init', 'Duration_remain', 'Premium'], axis = 1)
    fig, axes = plt.subplots(ncols=len(df.columns), figsize = (2*len(df.columns),3))
    for ax, col in zip(axes, df.columns):
        if col == 'Premium_freq':
            ls = sum(df[col]==0)/len(df)
            a = sum(df[col]==1)/len(df)
            m = sum(df[col]==12)/len(df)
            ax.bar(['up-front', 'annual', 'monthly'], [ls,a,m], width = 0.1, color='grey')
            ax.set_xticklabels(['up-front', 'annual', 'monthly'], Rotation=20)
            ax.set_xlabel('premium (freq.)')
        else:
            sns.distplot(df[col], ax=ax, color='grey')
            if col == 'Age':
                ax.set_xlabel('current age')
            elif col == 'Face_amount':
                ax.set_xlabel('face amount')
            elif col == 'Duration':
                ax.set_xlabel('duration')
            elif col == 'Duration_elapsed':
                ax.set_xlabel('duration (elapsed)')
            elif col == 'Premium_annual':
                ax.set_xlabel('premium (p.a.)')
            else:
                print('Unknown feature ', col)
                raise ValueError


    plt.tight_layout()
    plt.savefig(os.path.join(path_plots,r'portfolio_dist.png'), bbox_inches='tight')
    plt.savefig(os.path.join(path_plots,r'portfolio_dist.eps'), bbox_inches='tight')
    plt.close()


    ### additional visualizations
    ### note: experimental EDA on portfolio at time t=0
    pd.set_option('precision', 4)
    sns.set_style('ticks')
    sns.set_context('paper')

    ages = Portfolio['Age']
    premium_freq = Portfolio['Premium_freq']
    N_contracts = len(ages)

    # Visualize underwriting age
    # Compare to Milhaud, 'Lapse Tables [..]': 47.46% in [0,34], 34.4% in [35,54], 18.5% in [55,84]
    plt.hist(ages,bins= 100,  density= True)
    plt.vlines(x = 34.5, ymin = 0, ymax = 0.04, linestyle = '--', color = 'red')
    plt.vlines(x = 54.5, ymin = 0, ymax = 0.04, linestyle = '--', color = 'red')
    share_1 = sum(ages<34)/N_contracts
    share_2 = sum((ages>34)*(ages<54))/N_contracts
    share_3 = 1- share_1-share_2
    plt.text(x = 15, y = 0.03, s = str(round(share_1*100,ndigits=2))+'%', color = 'red')
    plt.text(x = 40, y = 0.03, s = str(round(share_2*100,ndigits=2))+'%', color = 'red')
    plt.text(x = 60, y = 0.03, s = str(round(share_3*100,ndigits=2))+'%', color = 'red')
    plt.show()

    # ## Durations (of Endowments)
    sns.axes_style('white')
    sns.distplot(Portfolio['Duration'], bins = 150, norm_hist= True)
    plt.title('Distribution: Duration of Endowments ')
    plt.show()

    # ## Elapsed duration
    sns.distplot(Portfolio['Duration_elapsed'], bins = 150)
    plt.show()

    # ## Face Amounts (S)
    # Choice arbitrary -> Backtest by looking and resulting premiums and compare range and variance to Milhaud's paper
    plt.hist(Portfolio['Face_amount'], bins = 100, density= True)
    plt.vlines(x= np.median(Portfolio['Face_amount']), ymin= 0, ymax= 0.00032, color = 'red', linestyles= '--')
    plt.show()
    sc.stats.describe(Portfolio['Face_amount'])

    # ## Annualize Premiums (P_ann)
    print('Median premium: \t ' +str(np.median(Portfolio.Premium_annual)))
    print('Mean premium: \t\t' + str(sc.stats.describe(Portfolio.Premium_annual).mean))
    print('Variance of premiums: \t' + str(np.sqrt(sc.stats.describe(Portfolio.Premium_annual).variance)))

    sns.distplot(Portfolio.Premium_annual[Portfolio.Premium_annual<Portfolio.Premium_annual.quantile(q=0.99)])
    plt.vlines(x=np.median(Portfolio.Premium_annual), ymin= 0, ymax= 0.0009, colors= 'red', linestyles= ':')
    plt.show()

    # Correlation of Features
    sns.heatmap(Portfolio.corr(),cmap= 'coolwarm', annot = True, annot_kws= {'size': 10})
    plt.show()


    sns.distplot(Portfolio['Age'], label='Age')
    sns.distplot(Portfolio['Age_init'], label = 'Age_init')
    plt.xlabel('Age [years]')
    plt.legend()

    fig, ax = plt.subplots(nrows=len(Portfolio.columns)//5+1, ncols= 5, figsize = (16,8) )
    ax=ax.flatten()

    for i, feature in enumerate(Portfolio.columns):
        sns.distplot(Portfolio[feature], ax = ax[i])
        ax[i].set_xlabel(feature)
    plt.tight_layout()
    plt.show()

if __name__ == "__main__":

    boolSimulate = False
    boolVisualize = True

    if boolSimulate:
        # create portfolio for time t=0
        run_simulation()
    if boolVisualize:
        # visualize distribution of data
        visualize_portfolio()