model.py

# -*- coding: utf-8 -*-
"""model.ipynb

Automatically generated by Colaboratory.

Original file is located at
    https://colab.research.google.com/drive/1hA9Xz_VjzMVj66qS_j3A5dKcGkAfScKM
"""

pip install pycaret

from scipy import stats

import numpy as np

import pandas as pd

import pickle

import matplotlib.pyplot as plt
# create fictitious data set 
simple_life_dataset = pd.DataFrame({'Age':[0, 60], 'Life Expectancy':[90, 30]})
simple_life_dataset.head()

import numpy as np
from scipy import stats
slope, intercept, r_value, p_value, std_err = stats.linregress(simple_life_dataset['Age'],simple_life_dataset['Life Expectancy'])
print('intercept: ', intercept)
print('slope: ', slope)

np.ceil(slope * 20 + intercept)

fig, axes = plt.subplots(figsize=(5,5))
x = [0,20,60]
y = [90, 70, 30]
axes.plot(x,y, color='blue', linestyle='--', marker='o')
fig.suptitle('Life Expectancy')
axes.set_xlabel('age')
axes.set_xlim([-5,100])
axes.set_ylabel('life_expectancy')
axes.set_ylim([0,100])
plt.grid()
plt.show()

# load WHO longevity data
# http://apps.who.int/gho/data/node.main.688
who_list = pd.read_csv('/content/drive/MyDrive/WHOSIS_000001,WHOSIS_000015.csv')
# save a local copy of the data set for our Flask prototype later on
who_list.to_csv('WHOSIS_000001,WHOSIS_000015.csv')

# Keep only useful features fix case display of country text
who_list = who_list[['GHO (DISPLAY)', 'YEAR (CODE)' , 'COUNTRY (DISPLAY)', 'SEX (DISPLAY)', 'Numeric']]
who_list['COUNTRY (DISPLAY)'] = [ctry.title() for ctry in who_list['COUNTRY (DISPLAY)'].values]
# print a few rows
who_list[who_list['COUNTRY (DISPLAY)']=='France'].head(10)

country = 'United States Of America'
sex = 'Male'
# pull latest entries for birth and 60 years for a country and gender
sub_set = who_list[who_list['COUNTRY (DISPLAY)'].str.startswith(country, na=False)]
sub_set = sub_set[sub_set['SEX (DISPLAY)'] == sex]
# sort by year in descending order to work with the latest read
sub_set = sub_set.sort_values('YEAR (CODE)', ascending=False)
sub_set_birth = sub_set[sub_set['GHO (DISPLAY)'] == 'Life expectancy at birth (years)']
sub_set_60 = sub_set[sub_set['GHO (DISPLAY)'] == 'Life expectancy at age 60 (years)']
print('sub_set_birth:')
print(sub_set_birth.head(5))
print('sub_set_60:')
print(sub_set_60.head(5))

# create data set with both points as shown in first example
lf_at_birth = sub_set_birth['Numeric'].values[0]
lf_at_60 = sub_set_60['Numeric'].values[0]
# let's organize our data and plot
age = [0,60]
life_expectancy = [lf_at_birth, lf_at_60]
fig, axes = plt.subplots(figsize=(5,5))
x = age
y = life_expectancy
axes.plot(x,y, color='blue', linestyle='--', marker='o')
fig.suptitle('Life Expectancy')
axes.set_xlabel('age')
axes.set_xlim([-5,100])
axes.set_ylabel('life expectancy')
axes.set_ylim([0,100])
plt.grid()
plt.show()

# model 
slope, intercept, r_value, p_value, std_err = stats.linregress(age, life_expectancy)
print('intercept: ', intercept)
print('slope: ', slope)

# predict life expectancy for an 49-year-old male in the USA:
np.ceil(slope * 49 + intercept)

def get_life_expectancy(age, country, sex):
    # pull latest entries for birth and 60 years
    sub_set = who_list[who_list['COUNTRY (DISPLAY)'].str.startswith(country, na=False)]
    sub_set = sub_set[sub_set['SEX (DISPLAY)'] == sex]
    sub_set = sub_set.sort_values('YEAR (CODE)', ascending=False)
    sub_set_birth = sub_set[sub_set['GHO (DISPLAY)'] == 'Life expectancy at birth (years)']
    sub_set_60 = sub_set[sub_set['GHO (DISPLAY)'] == 'Life expectancy at age 60 (years)']

    # not all combinations exsits so check that we have data for both
    if len(sub_set_birth['Numeric']) > 0 and len(sub_set_60['Numeric']) > 0:
        # create data set with both points as shown in first example
        lf_at_birth = sub_set_birth['Numeric'].values[0]
        lf_at_60 = sub_set_60['Numeric'].values[0]

        # model
        slope, intercept, r_value, p_value, std_err = stats.linregress([0,60],[lf_at_birth, lf_at_60])

        # predict for the age variable
        return(np.ceil(slope * age + intercept))
    else:
        return None

list(set(who_list['COUNTRY (DISPLAY)']))[0:10]

# test the function out using a 22-year-old Japanese female:
get_life_expectancy(22, 'Japan', 'Female')

get_life_expectancy(22,'Pakistan','Female')

get_life_expectancy(21,'India','Male')

missing_values_count = who_list.isnull().sum()

print(missing_values_count)

get_life_expectancy(80,'Pakistan','Female')

pickle.dump(get_life_expectancy,open('model.pkl','wb'))

model=pickle.load(open('model.pkl','rb'))