PORTFOLIO OSWALDO L. ZÁRATE

IR AL PORTFOLIO DE PROYECTOS

---

PREDICCION DE GASTOS POR SEGUROS

Importamos todas las librerias necesarias

import numpy as np
import matplotlib.pyplot as plt
import pandas as pd
import seaborn as sns
from sklearn.preprocessing import StandardScaler
from sklearn.linear_model import LinearRegression
from xgboost import XGBRegressor
from sklearn.linear_model import LogisticRegression
from sklearn.ensemble._forest import RandomForestClassifier
#from sklearn.ensemble import RandomForestClassifier
from sklearn import preprocessing
from sklearn.model_selection import GridSearchCV
from sklearn.model_selection import train_test_split
from sklearn.metrics import precision_score, accuracy_score
from sklearn.metrics import confusion_matrix
from sklearn.metrics import classification_report 
from sklearn.metrics import auc
from sklearn.metrics import roc_curve

#get model duration
import time
from datetime import date

#use el siguiente script para ver la salida de todo el código.

from IPython.core.interactiveshell import InteractiveShell
InteractiveShell.ast_node_Interactivity = 'all'

Lectura del Dataset

insurance = pd.read_csv("insurance.csv")

---

Descrición de los datos

• Age: edad del beneficiario principal

• sex: sexo del contratista de seguros, mujer, hombre

• BMI: índice de masa corporal, que proporciona una comprensión del cuerpo, pesos relativamente altos o bajos en relación con la altura, índice objetivo de peso corporal (kg / m ^ 2) utilizando la relación entre la altura y el peso, idealmente 18,5 a 24,9

• Children: Número de niños cubiertos por seguro médico / Número de dependientes.

• Smoker: Fumar

• Region: Área residencial del beneficiario en los EE. UU., Noreste, sureste, suroeste, noroeste.

• Charges: Costos médicos individuales facturados por el seguro médico.

insurance.shape

(1338, 7)

insurance.describe()

	age	bmi	children	charges
count	1338.000000	1338.000000	1338.000000	1338.000000
mean	39.207025	30.663397	1.094918	13270.422265
std	14.049960	6.098187	1.205493	12110.011237
min	18.000000	15.960000	0.000000	1121.873900
25%	27.000000	26.296250	0.000000	4740.287150
50%	39.000000	30.400000	1.000000	9382.033000
75%	51.000000	34.693750	2.000000	16639.912515
max	64.000000	53.130000	5.000000	63770.428010

insurance.dtypes

age           int64
sex          object
bmi         float64
children      int64
smoker       object
region       object
charges     float64
dtype: object

insurance.head()

	age	sex	bmi	children	smoker	region	charges
0	19	female	27.900	0	yes	southwest	16884.92400
1	18	male	33.770	1	no	southeast	1725.55230
2	28	male	33.000	3	no	southeast	4449.46200
3	33	male	22.705	0	no	northwest	21984.47061
4	32	male	28.880	0	no	northwest	3866.85520

Identificación de valores faltantes

insurance.isnull().sum()

age         0
sex         0
bmi         0
children    0
smoker      0
region      0
charges     0
dtype: int64

total_miss = insurance.isnull().any()   #Parece que no faltan valores. 
total_miss

age         False
sex         False
bmi         False
children    False
smoker      False
region      False
charges     False
dtype: bool

Ingeniería de Funciones y Análisis Exploratorio de Datos (E.D.A.)

• Verificando correlaciones entre las características

insurance.corr() 

	age	bmi	children	charges
age	1.000000	0.109272	0.042469	0.299008
bmi	0.109272	1.000000	0.012759	0.198341
children	0.042469	0.012759	1.000000	0.067998
charges	0.299008	0.198341	0.067998	1.000000

• Mapa de calor, tabaquismo, edad y bmi tienen una relación más fuerte con los cargos

sns.heatmap(insurance.corr(), annot=True)

<AxesSubplot:>

• De la gráfica de abajo, podemos ver que los cargos del seguro aumentan con el aumento de la edad.

from scipy import stats
from warnings import filterwarnings
filterwarnings('ignore')

g = sns.JointGrid(insurance['age'], insurance['charges'])
g = g.plot(sns.regplot, sns.histplot).annotate(stats.pearsonr)
plt.show()

---------------------------------------------------------------------------
AttributeError                            Traceback (most recent call last)
~\AppData\Local\Temp/ipykernel_5072/2057381814.py in <module>
      4 
      5 g = sns.JointGrid(insurance['age'], insurance['charges'])
----> 6 g = g.plot(sns.regplot, sns.histplot).annotate(stats.pearsonr)
      7 plt.show()

AttributeError: 'JointGrid' object has no attribute 'annotate'

• Separando y agrupando a niños en tamaños de familia.

bins = [0,1,3,5]
labels = ['Small_Family','Medium_Family','Large_Family']
insurance['children_binned'] = pd.cut(insurance['children'], bins=bins, labels=labels)
sns.boxplot(x = 'children_binned',y='charges', data = insurance)

insurance.info()

<class 'pandas.core.frame.DataFrame'>
RangeIndex: 1338 entries, 0 to 1337
Data columns (total 8 columns):
 #   Column           Non-Null Count  Dtype   
---  ------           --------------  -----   
 0   age              1338 non-null   int64   
 1   sex              1338 non-null   object  
 2   bmi              1338 non-null   float64 
 3   children         1338 non-null   int64   
 4   smoker           1338 non-null   object  
 5   region           1338 non-null   object  
 6   charges          1338 non-null   float64 
 7   children_binned  764 non-null    category
dtypes: category(1), float64(2), int64(2), object(3)
memory usage: 74.7+ KB

Codificación de las características ('sex', 'smoker', 'region')

for col in ['sex', 'smoker', 'region']:
    print( col,':')
    print(insurance[col].value_counts())

sex :
male      676
female    662
Name: sex, dtype: int64
smoker :
no     1064
yes     274
Name: smoker, dtype: int64
region :
southeast    364
southwest    325
northwest    325
northeast    324
Name: region, dtype: int64

• Etiqueta de codificación final para los tipos de datos del objeto

for col in ['sex', 'smoker', 'region']:
    if (insurance[col].dtype == 'object'):
        le = preprocessing.LabelEncoder()
        le = le.fit(insurance[col])
        insurance[col] = le.transform(insurance[col])
        print('Completed Label encoding on',col)

Completed Label encoding on sex
Completed Label encoding on smoker
Completed Label encoding on region

insurance.head()

	age	sex	bmi	children	smoker	region	charges	children_binned
0	19	0	27.900	0	1	3	16884.92400	NaN
1	18	1	33.770	1	0	2	1725.55230	Small_Family
2	28	1	33.000	3	0	2	4449.46200	Medium_Family
3	33	1	22.705	0	0	1	21984.47061	NaN
4	32	1	28.880	0	0	1	3866.85520	NaN

La salida de arriba muestra el cambio de etiquetas de los objetos 'sex', 'smoker', 'region' que tenian valores no estructurados, y se lo cambio a valores categoricos como 0,1,2,3

insurance.corr()

	age	sex	bmi	children	smoker	region	charges
age	1.000000	-0.020856	0.109272	0.042469	-0.025019	0.002127	0.299008
sex	-0.020856	1.000000	0.046371	0.017163	0.076185	0.004588	0.057292
bmi	0.109272	0.046371	1.000000	0.012759	0.003750	0.157566	0.198341
children	0.042469	0.017163	0.012759	1.000000	0.007673	0.016569	0.067998
smoker	-0.025019	0.076185	0.003750	0.007673	1.000000	-0.002181	0.787251
region	0.002127	0.004588	0.157566	0.016569	-0.002181	1.000000	-0.006208
charges	0.299008	0.057292	0.198341	0.067998	0.787251	-0.006208	1.000000

• Mapa de calor, smoker, age y bmi tienen una relación más fuerte con los cargos

sns.heatmap(insurance.corr(), annot=True)

<AxesSubplot:>

insurance.dtypes

age                   int64
sex                   int32
bmi                 float64
children              int64
smoker                int32
region                int32
charges             float64
children_binned    category
dtype: object

sns.plot = sns.distplot(insurance['charges'])

for col in ['sex', 'smoker', 'region']:
    print( col,':')
    print(insurance[col].value_counts())

sex :
1    676
0    662
Name: sex, dtype: int64
smoker :
0    1064
1     274
Name: smoker, dtype: int64
region :
2    364
3    325
1    325
0    324
Name: region, dtype: int64

Por lo tanto hemos codificado la caracteristica 'sex' como: '0' para mujer & '1' para hombre.

Para la característica 'smoker' : '0' para no & '1' para yes.

Para la característica 'region': '0' para northeast, '1' para southwest, '2' para southeast & '3' para northwest.

El resumen puede verse a continuación:

sex :

• male 676 sex = 1 is male.
• female 662 sex = 0 is female.

smoker :

• no 1064 smoker = 0 means no (not a smoker)
• yes 274 smoker = 1 means yes (a smoker)

region :

• southeast 364 2 - southeast
• northwest 325 3 - northwest
• southwest 325 1 - southwest
• northeast 324 0 - northeast

Guardar el conjunto de datos limpio y codificado

insurance.to_csv('insurance_encoded.csv', index = False)

---

Dividiendo el conjunto de datos en variables independientes y dependientes

insurance_input = insurance.drop(['charges', 'children_binned'],axis=1)
insurance_target = insurance['charges']

Escala de características

Estandarizar datos

x_scaled = StandardScaler().fit_transform(insurance_input)

Train/Test split

x_train, x_test, y_train, y_test = train_test_split(x_scaled,
                                                    insurance_target,
                                                    test_size = 0.25,
                                                    random_state=1211)

---

---

Countrucción de Modelos

- 1.Linear Regression

from sklearn import linear_model
from sklearn.linear_model import LinearRegression

#Inicialización
linReg = LinearRegression()

start_time = time.time()

# Ajustar el modelo lineal a el conjunto de datos de entrenamiento
linReg_model = linReg.fit(x_train, y_train)

today = date.today()
print("--- %s seconds ---" % (time.time() - start_time))

--- 0.348020076751709 seconds ---

linReg.coef_

array([3690.94980639, -154.24448937, 1926.27073861,  520.55396204,
       9697.07408404, -410.85583285])

---

---

A lo largo del módulo, designamos el vector w = (w_1, ..., wp) como `coefy w_0 comointercept_`.

$$\hat{y}(w,x)=w_0+w_1{x}_1+...+w_p{x}_p$$

---

---

• Obtenemos los valores de los coeficientes

input_columns = insurance.columns[:6]  #Obtener nombres de las caracteristicas

coeff_df = pd.DataFrame(linReg.coef_, input_columns, columns=['Coefficient'])
coeff_df

	Coefficient
age	3690.949806
sex	-154.244489
bmi	1926.270739
children	520.553962
smoker	9697.074084
region	-410.855833

De los Coeficientes de encima podemos observar que las caracteristicas : smoker, age y bmi tienen el mayor efecto en los cargos del seguro,

lo cual es correcto porque incluso en el mundo real: la edad, la obesidad y los hábitos de fumar tienen una influencia significativa sobre los riesgos para la salud. y, por tanto, mayores gastos de seguro.

• ### Validar el modelo

y_pred_train = linReg.predict(x_train)    # Predict on train data.
y_pred_train[y_pred_train < 0] = y_pred_train.mean()
y_pred = linReg.predict(x_test)   # Predict on test data.
y_pred[y_pred < 0] = y_pred.mean()
diff = pd.DataFrame({'Actual': y_test, 'Predicted': y_pred})
diff.head(5)

	Actual	Predicted
926	2913.56900	675.576998
490	1748.77400	2813.561034
1245	5615.36900	4312.605591
854	24106.91255	34277.985644
1002	1972.95000	1575.497282

Trazando 'Actual' y 'Predicted'

diff.plot(kind='bar',figsize=(8,8))
plt.grid(which='major', linestyle='-', linewidth='0.5', color='green')
plt.grid(which='minor', linestyle=':', linewidth='0.5', color='black')
plt.show()

• Porsupuesto el grafico de encima no es muy claro, Trazaremos una grafica con los 10 primeros valores.

diff1 = diff.head(10)
diff1

	Actual	Predicted
926	2913.56900	675.576998
490	1748.77400	2813.561034
1245	5615.36900	4312.605591
854	24106.91255	34277.985644
1002	1972.95000	1575.497282
448	5910.94400	7289.580969
475	28868.66390	37298.040845
1059	4462.72180	7388.135839
683	9863.47180	9473.012595
1278	22462.04375	32398.763860

diff1.plot(kind='bar',figsize=(8,8))
plt.grid(which='major', linestyle='-', linewidth='0.5', color='green')
plt.grid(which='minor', linestyle=':', linewidth='0.5', color='black')
plt.show()

def calculate_accuracy(actual, predicted):
    SST = 0
    SSR = 0
    SSE = 0
    RMSE = 0
    VIF = 0
    RSqr = 0
    MAE = 0
    MAPE = 0
    SST = sum((actual - np.mean(predicted))**2)    # Calculate the SST
    SSR = sum((predicted - np.mean(predicted))**2) # Calculate the SSR
    SSE = sum((actual - predicted)**2)             # Calculate the SSE
    RMSE = np.sqrt((sum((predicted - actual)**2))/len(predicted))  # Calculate the RMSE
    RSqr = 1 - (SSE/SST)                           # Calcualte the R_square
    if RSqr != 1:
        VIF = 1 / (1 - RSqr)                           # Calculate the VIF
    #MAPE_house_price = mape(dtc_predict_train, y_train)           # Calculate the MAPE
    #MAE_house_price = mae(y_train, dtc_predict_train)             # Calculate the MAE
    return RMSE, RSqr, VIF

# Finding MAE, MSE and other, metrics.
from sklearn import metrics
print('Mean Absolute Error:', metrics.mean_absolute_error(y_test, y_pred))  
print('Mean Squared Error:', metrics.mean_squared_error(y_test, y_pred))  
print('Root Mean Squared Error:', np.sqrt(metrics.mean_squared_error(y_test, y_pred)))
# print the intercept and coefficients
print('Intercept: ',linReg.intercept_)
#print('r2 score: ',linReg.score(x_train, y_train))
#print('r2 score: ',linReg.score(x_test, y_test))

Mean Absolute Error: 4068.9236952525494
Mean Squared Error: 36334934.764302835
Root Mean Squared Error: 6027.846610880442
Intercept:  13207.388159129547

# calculate the accuracies
RMSE, RSqr, VIF = calculate_accuracy(y_train,y_pred_train)
print('Linear RMSE train = ',RMSE, 'R-Square train = ',RSqr, 'VIF train = ',VIF)

RMSE, RSqr, VIF = calculate_accuracy(y_test,y_pred)
print('Linear RMSE test = ',RMSE, 'R-Square test = ',RSqr, 'VIF test = ',VIF)

Linear RMSE train =  6200.042303094933 R-Square train =  0.7437907235802795 VIF train =  3.9030593036054095
Linear RMSE test =  6027.846610880442 R-Square test =  0.7328599025559049 VIF test =  3.743354178454142

!pip install statsmodels

Collecting statsmodels
  Using cached statsmodels-0.13.1-cp38-none-win_amd64.whl (9.4 MB)
Requirement already satisfied: numpy>=1.17 in c:\users\tera\anaconda3\envs\py2m\lib\site-packages (from statsmodels) (1.21.4)
Requirement already satisfied: pandas>=0.25 in c:\users\tera\anaconda3\envs\py2m\lib\site-packages (from statsmodels) (1.3.4)
Requirement already satisfied: scipy>=1.3 in c:\users\tera\anaconda3\envs\py2m\lib\site-packages (from statsmodels) (1.7.2)
Collecting patsy>=0.5.2
  Using cached patsy-0.5.2-py2.py3-none-any.whl (233 kB)
Requirement already satisfied: pytz>=2017.3 in c:\users\tera\anaconda3\envs\py2m\lib\site-packages (from pandas>=0.25->statsmodels) (2021.3)
Requirement already satisfied: python-dateutil>=2.7.3 in c:\users\tera\anaconda3\envs\py2m\lib\site-packages (from pandas>=0.25->statsmodels) (2.8.2)
Requirement already satisfied: six in c:\users\tera\anaconda3\envs\py2m\lib\site-packages (from patsy>=0.5.2->statsmodels) (1.16.0)
Installing collected packages: patsy, statsmodels
Successfully installed patsy-0.5.2 statsmodels-0.13.1

#Backward elimination to check for high significance feature

import statsmodels.api as sm
a = 0
b = 0
a, b = insurance_input.shape
insurance_input = np.append(arr = np.ones((a, 1)).astype(int), values = insurance_input, axis = 1)
print (insurance_input.shape)

insurance_input_opt = insurance_input[:, [0, 1, 2, 4]]
##OrdinaryLeastSquares
regressorOLS = sm.OLS(endog = insurance_target, exog = insurance_input_opt).fit()
regressorOLS.summary()

(1338, 7)

OLS Regression Results

Dep. Variable:	charges	R-squared:	0.096
Model:	OLS	Adj. R-squared:	0.094
Method:	Least Squares	F-statistic:	47.43
Date:	Wed, 17 Nov 2021	Prob (F-statistic):	4.00e-29
Time:	03:41:02	Log-Likelihood:	-14410.
No. Observations:	1338	AIC:	2.883e+04
Df Residuals:	1334	BIC:	2.885e+04
Df Model:	3
Covariance Type:	nonrobust

	coef	std err	t	P>|t|	[0.025	0.975]
const	1837.2804	1022.425	1.797	0.073	-168.455	3843.016
x1	256.8610	22.458	11.437	0.000	212.804	300.918
x2	1515.1059	630.399	2.403	0.016	278.424	2751.788
x3	545.1609	261.732	2.083	0.037	31.709	1058.613

Omnibus:	397.122	Durbin-Watson:	2.046
Prob(Omnibus):	0.000	Jarque-Bera (JB):	856.804
Skew:	1.721	Prob(JB):	8.86e-187
Kurtosis:	4.876	Cond. No.	139.

Notes:
[1] Standard Errors assume that the covariance matrix of the errors is correctly specified.

#Backward elimination to check for high significance feature

import statsmodels.api as sm
a = 0
b = 0
a, b = x_scaled.shape
x_scaled = np.append(arr = np.ones((a, 1)).astype(int), values = x_scaled, axis = 1)
print (x_scaled.shape)

x_scaled_opt = x_scaled[:, [0, 1, 2, 4]]
##OrdinaryLeastSquares
regressorOLS = sm.OLS(endog = insurance_target, exog = x_scaled_opt).fit()
regressorOLS.summary()

(1338, 7)

OLS Regression Results

Dep. Variable:	charges	R-squared:	0.096
Model:	OLS	Adj. R-squared:	0.094
Method:	Least Squares	F-statistic:	47.43
Date:	Wed, 17 Nov 2021	Prob (F-statistic):	4.00e-29
Time:	03:42:02	Log-Likelihood:	-14410.
No. Observations:	1338	AIC:	2.883e+04
Df Residuals:	1334	BIC:	2.885e+04
Df Model:	3
Covariance Type:	nonrobust

	coef	std err	t	P>|t|	[0.025	0.975]
const	1.327e+04	315.062	42.120	0.000	1.27e+04	1.39e+04
x1	3607.5381	315.421	11.437	0.000	2988.764	4226.313
x2	757.5115	315.182	2.403	0.016	139.204	1375.819
x3	656.9418	315.398	2.083	0.037	38.211	1275.673

Omnibus:	397.122	Durbin-Watson:	2.046
Prob(Omnibus):	0.000	Jarque-Bera (JB):	856.804
Skew:	1.721	Prob(JB):	8.86e-187
Kurtosis:	4.876	Cond. No.	1.05

Notes:
[1] Standard Errors assume that the covariance matrix of the errors is correctly specified.

---

2. Decision Tree

from sklearn.tree import DecisionTreeRegressor
dtc = DecisionTreeRegressor(random_state=1)

# create the model
dtc.fit(x_train,y_train)

# prediction on train data
dtc_predict_train = dtc.predict(x_train)

# prediction on test data
dtc_predict_test = dtc.predict(x_test)

# calculate the accuracies
RMSE, RSqr, VIF = calculate_accuracy(y_train,dtc_predict_train)
print('Decision Tree RMSE train = ',RMSE, 'R-Square train = ',RSqr, 'VIF train = ',VIF)

RMSE, RSqr, VIF = calculate_accuracy(y_test,dtc_predict_test)
print('Decision Tree RMSE test = ',RMSE, 'R-Square test = ',RSqr, 'VIF test = ',VIF)

Decision Tree RMSE train =  279.1173376029252 R-Square train =  0.9994805465077318 VIF train =  1925.1001579244978
Decision Tree RMSE test =  6479.645615986229 R-Square test =  0.6912970796606894 VIF test =  3.239360349752606

from sklearn.ensemble import RandomForestRegressor
# Random forest model
rfc = RandomForestRegressor()

rfc.fit(x_train,y_train)

# prediction on train data
rfc_predict_train = rfc.predict(x_train)

# prediction on test data
rfc_predict_test = rfc.predict(x_test)

# calculate the accuracies
RMSE, RSqr, VIF = calculate_accuracy(y_train,rfc_predict_train)
print('Random Forest RMSE train = ',RMSE, 'R-Square train = ',RSqr, 'VIF train = ',VIF)

RMSE, RSqr, VIF = calculate_accuracy(y_test,rfc_predict_test)
print('Random Forest RMSE test = ',RMSE, 'R-Square test = ',RSqr, 'VIF test = ',VIF)

Random Forest RMSE train =  1843.180644813415 R-Square train =  0.9773495387681166 VIF train =  44.14921134552318
Random Forest RMSE test =  4805.639146875192 R-Square test =  0.8302233541387685 VIF test =  5.890091625542885

---

Resumeniendo las metricas de precision de todos los modelos.

print('Metrics of linear regression:')
RMSE, RSqr, VIF = calculate_accuracy(y_train,y_pred_train)
print('Linear RMSE train = ',RMSE, 'R-Square train = ',RSqr, 'VIF train = ',VIF)
RMSE, RSqr, VIF = calculate_accuracy(y_test,y_pred)
print('Linear RMSE test = ',RMSE, 'R-Square test = ',RSqr, 'VIF test = ',VIF)
print('                                  ')

print('Metrics of Decision Tree:')
RMSE, RSqr, VIF = calculate_accuracy(y_train,dtc_predict_train)
print('Decision Tree RMSE train = ',RMSE, 'R-Square train = ',RSqr, 'VIF train = ',VIF)
RMSE, RSqr, VIF = calculate_accuracy(y_test,dtc_predict_test)
print('Decision Tree RMSE test = ',RMSE, 'R-Square test = ',RSqr, 'VIF test = ',VIF)
print('                                  ')

print('Metrics of Random Forest:')
RMSE, RSqr, VIF = calculate_accuracy(y_train,rfc_predict_train)
print('Random Forest RMSE train = ',RMSE, 'R-Square train = ',RSqr, 'VIF train = ',VIF)
RMSE, RSqr, VIF = calculate_accuracy(y_test,rfc_predict_test)
print('Random Forest RMSE test = ',RMSE, 'R-Square test = ',RSqr, 'VIF test = ',VIF)

Metrics of linear regression:
Linear RMSE train =  6200.042303094933 R-Square train =  0.7437907235802795 VIF train =  3.9030593036054095
Linear RMSE test =  6027.846610880442 R-Square test =  0.7328599025559049 VIF test =  3.743354178454142
                                  
Metrics of Decision Tree:
Decision Tree RMSE train =  279.1173376029252 R-Square train =  0.9994805465077318 VIF train =  1925.1001579244978
Decision Tree RMSE test =  6479.645615986229 R-Square test =  0.6912970796606894 VIF test =  3.239360349752606
                                  
Metrics of Random Forest:
Random Forest RMSE train =  1843.180644813415 R-Square train =  0.9773495387681166 VIF train =  44.14921134552318
Random Forest RMSE test =  4805.639146875192 R-Square test =  0.8302233541387685 VIF test =  5.890091625542885

---

---

Dado que el modelo Random Forest funciona mejor en mi conjunto de datos, usaré Random Forest como modelo final para implementar y predecir los cargos del seguro en función de los atributos de una persona.

---

---

import pickle

• Guardar el modelo a el disco local

filename = 'streamlit_insurance_predictcharges.pkl'
pickle.dump(rfc, open(filename, 'wb'))

• Cargar el modelo desde el Disco local

loaded_model = pickle.load(open(filename, 'rb'))
result = loaded_model.score(x_test, y_test)
print(result) 

0.8301987341704054

Ahora usaremos otro notebook llamado 'predecir cargos de seguro'

• Para crear la aplicación para predecir los cargos del seguro usando streamlit y nuevamente guardar el portátil en el directorio local con el nombre 'streamlitpredictinsurancecharges'