
import pandas as pd
import numpy as np 
import seaborn as sns
import matplotlib as mpl
import matplotlib.pyplot as plt
import plotly.express as px
from sklearn.linear_model import LogisticRegression
from sklearn.preprocessing import StandardScaler
from sklearn.model_selection import train_test_split
from sklearn.metrics import accuracy_score, precision_score, confusion_matrix, f1_score, recall_score
from sklearn.neighbors import KNeighborsClassifier
from sklearn.naive_bayes import MultinomialNB
import statsmodels.api as sm
from sklearn.metrics import roc_curve
%matplotlib inline


data_location = "dataset_infartos.csv"


data_inf = pd.read_csv(data_location)


data_inf.head()


data_inf.info()

<class 'pandas.core.frame.DataFrame'>
RangeIndex: 4238 entries, 0 to 4237
Data columns (total 16 columns):
 #   Column           Non-Null Count  Dtype  
---  ------           --------------  -----  
 0   male             4238 non-null   int64  
 1   age              4238 non-null   int64  
 2   education        4133 non-null   float64
 3   currentSmoker    4238 non-null   int64  
 4   cigsPerDay       4209 non-null   float64
 5   BPMeds           4185 non-null   float64
 6   prevalentStroke  4238 non-null   int64  
 7   prevalentHyp     4238 non-null   int64  
 8   diabetes         4238 non-null   int64  
 9   totChol          4188 non-null   float64
 10  sysBP            4238 non-null   float64
 11  diaBP            4238 non-null   float64
 12  BMI              4219 non-null   float64
 13  heartRate        4237 non-null   float64
 14  glucose          3850 non-null   float64
 15  TenYearCHD       4238 non-null   int64  
dtypes: float64(9), int64(7)
memory usage: 529.9 KB


#vemos la cantidad de nulos
cantidad_nulos = data_inf.isnull().sum()
cantidad_nulos

male                 0
age                  0
education          105
currentSmoker        0
cigsPerDay          29
BPMeds              53
prevalentStroke      0
prevalentHyp         0
diabetes             0
totChol             50
sysBP                0
diaBP                0
BMI                 19
heartRate            1
glucose            388
TenYearCHD           0
dtype: int64


total = data_inf.shape[0] 
nulos = data_inf.isnull().sum()  
porcentaje_nulos = 100 * nulos / total
porcentaje_nulos

male               0.000000
age                0.000000
education          2.477584
currentSmoker      0.000000
cigsPerDay         0.684285
BPMeds             1.250590
prevalentStroke    0.000000
prevalentHyp       0.000000
diabetes           0.000000
totChol            1.179802
sysBP              0.000000
diaBP              0.000000
BMI                0.448325
heartRate          0.023596
glucose            9.155262
TenYearCHD         0.000000
dtype: float64


#vemos los valores para la variable target, riesgo de cardioatia en los proximos 10 años. El daatset está desbalanceado 
data_inf["TenYearCHD"].value_counts()

0    3594
1     644
Name: TenYearCHD, dtype: int64


#Borramos los datos que son nulos del dataset, es un porcentajo muy bajo  
data_inf_sin_nulos = data_inf.dropna()


data_inf_sin_nulos.isnull().sum()

male               0
age                0
education          0
currentSmoker      0
cigsPerDay         0
BPMeds             0
prevalentStroke    0
prevalentHyp       0
diabetes           0
totChol            0
sysBP              0
diaBP              0
BMI                0
heartRate          0
glucose            0
TenYearCHD         0
dtype: int64


#vemos la cantidad de observaciones que tenemos dentro de cada categoría de la variabla target   
data_inf_sin_nulos["TenYearCHD"].value_counts()

0    3099
1     557
Name: TenYearCHD, dtype: int64


# Vemos cuántos "Si" tenemos
riesgo_si_mask = (data_inf_sin_nulos["TenYearCHD"] > 0)
data_riesgo_si = data_inf_sin_nulos[riesgo_si_mask]
data_riesgo_si.shape

(557, 16)


# Vemos cuántos "no" tenemos
riesgo_no_mask = (data_inf_sin_nulos["TenYearCHD"] < 1)
data_riesgo_no = data_inf_sin_nulos[riesgo_no_mask]
data_riesgo_no.shape

(3099, 16)


# Dataset con 44% de datos de personas que no tienen riesgo de cardiopatía en 10 años  
data_riesgo_no_557 = data_riesgo_no.sample(n=round(557*1.1) , random_state=123)
data_riesgo_no_557.head(3)


# concatenamos los datasets 
data_para_regresion = pd.concat([data_riesgo_si,data_riesgo_no_557],axis=0)


data_para_regresion.info()

<class 'pandas.core.frame.DataFrame'>
Int64Index: 1170 entries, 3 to 163
Data columns (total 16 columns):
 #   Column           Non-Null Count  Dtype  
---  ------           --------------  -----  
 0   male             1170 non-null   int64  
 1   age              1170 non-null   int64  
 2   education        1170 non-null   float64
 3   currentSmoker    1170 non-null   int64  
 4   cigsPerDay       1170 non-null   float64
 5   BPMeds           1170 non-null   float64
 6   prevalentStroke  1170 non-null   int64  
 7   prevalentHyp     1170 non-null   int64  
 8   diabetes         1170 non-null   int64  
 9   totChol          1170 non-null   float64
 10  sysBP            1170 non-null   float64
 11  diaBP            1170 non-null   float64
 12  BMI              1170 non-null   float64
 13  heartRate        1170 non-null   float64
 14  glucose          1170 non-null   float64
 15  TenYearCHD       1170 non-null   int64  
dtypes: float64(9), int64(7)
memory usage: 155.4 KB


total = data_para_regresion.shape[0] 
nulos = data_para_regresion.isnull().sum()  
porcentaje_nulos = 100 * nulos / total
porcentaje_nulos

male               0.0
age                0.0
education          0.0
currentSmoker      0.0
cigsPerDay         0.0
BPMeds             0.0
prevalentStroke    0.0
prevalentHyp       0.0
diabetes           0.0
totChol            0.0
sysBP              0.0
diaBP              0.0
BMI                0.0
heartRate          0.0
glucose            0.0
TenYearCHD         0.0
dtype: float64


# chequeamos si el dataset para la regresión quedó más balanceado 
data_para_regresion.TenYearCHD.value_counts()

0    613
1    557
Name: TenYearCHD, dtype: int64


data_para_regresion.cigsPerDay.value_counts()

0.0     607
20.0    215
30.0     61
15.0     61
40.0     31
10.0     30
5.0      28
3.0      24
9.0      21
1.0      17
43.0     15
25.0     13
6.0       6
35.0      6
2.0       5
7.0       4
17.0      3
50.0      3
60.0      3
8.0       3
45.0      2
18.0      2
12.0      2
4.0       2
16.0      1
19.0      1
11.0      1
23.0      1
14.0      1
13.0      1
Name: cigsPerDay, dtype: int64


# Medicamento para Presión

data_raw_bpmed = pd.DataFrame(
    {'Posible CHD': ['No', 'No', 'Si', 'Si'],
     'Cantidad de Personas': [data_riesgo_no_557.BPMeds.value_counts()[0],data_riesgo_no_557.BPMeds.value_counts()[1], data_riesgo_si.BPMeds.value_counts()[0], data_riesgo_si.BPMeds.value_counts()[1]],
     'Medicamentos para control de presión arterial': ['No', 'Si', 'No', 'Si']
    })

fig = px.bar(data_raw_bpmed, x="Medicamentos para control de presión arterial", y="Cantidad de Personas", color='Posible CHD', barmode='group' , color_discrete_sequence=["#00CC96", "#EF553B"]
)
fig.show()


# ACV

data_raw_acv = pd.DataFrame(
    {'Posible CHD': ['No', 'No', 'Si', 'Si'],
     'Cantidad de Personas': [data_riesgo_no_557.prevalentStroke.value_counts()[0],data_riesgo_no_557.prevalentStroke.value_counts()[1], data_riesgo_si.prevalentStroke.value_counts()[0], data_riesgo_si.prevalentStroke.value_counts()[1]],
     'Antecedente ACV': ['No', 'Si', 'No', 'Si']
    })

fig = px.bar(data_raw_acv, x="Antecedente ACV", y="Cantidad de Personas", color='Posible CHD', barmode='group' , color_discrete_sequence=["#00CC96", "#EF553B"]
)
fig.show()


# Hipertenso

data_raw_ht = pd.DataFrame(
    {'Posible CHD': ['No', 'No', 'Si', 'Si'],
     'Cantidad de Personas': [data_riesgo_no_557.prevalentHyp.value_counts()[0],data_riesgo_no_557.prevalentHyp.value_counts()[1], data_riesgo_si.prevalentHyp.value_counts()[0], data_riesgo_si.prevalentHyp.value_counts()[1]],
     'Hipertenso': ['No', 'Si', 'No', 'Si']
    })

fig = px.bar(data_raw_ht, x="Hipertenso", y="Cantidad de Personas", color='Posible CHD', barmode='group' , color_discrete_sequence=["#00CC96", "#EF553B"]
)
fig.show()


# Diabetes

data_raw_diab = pd.DataFrame(
    {'Posible CHD': ['No', 'No', 'Si', 'Si'],
     'Cantidad de Personas': [data_riesgo_no_557.diabetes.value_counts()[0],data_riesgo_no_557.diabetes.value_counts()[1], data_riesgo_si.diabetes.value_counts()[0], data_riesgo_si.diabetes.value_counts()[1]],
     'Diabetico': ['No', 'Si', 'No', 'Si']
    })

fig = px.bar(data_raw_diab, x="Diabetico", y="Cantidad de Personas", color='Posible CHD', barmode='group' , color_discrete_sequence=["#00CC96", "#EF553B"]
)
fig.show()


# SMOKER

data_raw_smoker = pd.DataFrame(
    {'Posible CHD': ['No', 'No', 'Si', 'Si'],
     'Cantidad de Personas': [data_riesgo_no_557.currentSmoker.value_counts()[0],data_riesgo_no_557.currentSmoker.value_counts()[1], data_riesgo_si.currentSmoker.value_counts()[0], data_riesgo_si.currentSmoker.value_counts()[1]],
     'Fumador': ['No', 'Si', 'No', 'Si']
    })

fig = px.bar(data_raw_smoker, x="Fumador", y="Cantidad de Personas", color='Posible CHD', barmode='relative' , color_discrete_sequence=["#00CC96", "#EF553B"]
)
fig.show()


data_para_regresion


# comparacion de edades entre quienes tienen riesgo y no tienen riesgo 
fig = px.box(data_para_regresion, x="TenYearCHD", y="age")
fig.show()


# comparacion de educacion entre quienes tienen riesgo y no tienen riesgo 
fig = px.box(data_para_regresion, x="TenYearCHD", y="education")
fig.show()


# comparacion colesterol
fig = px.box(data_para_regresion, x="TenYearCHD", y="totChol")
fig.show()


fig = px.box(data_para_regresion, x="TenYearCHD", y="sysBP")
fig.show()


fig = px.box(data_para_regresion, x="TenYearCHD", y="diaBP")
fig.show()


# BMI
fig = px.box(data_para_regresion, x="TenYearCHD", y="BMI")
fig.show()


# niveles de glucosa 
fig = px.box(data_para_regresion, x="TenYearCHD", y="glucose")
fig.show()


# comparacion educacion
fig = px.box(data_para_regresion, x="TenYearCHD", y="heartRate")
fig.show()


cigarrillo_CHD_grouped =data_para_regresion.groupby( ["TenYearCHD", "cigsPerDay"] )
cigarrillo_CHD_count = pd.DataFrame(cigarrillo_CHD_grouped.size().reset_index(name = "Group_Count"))


cigarrillo_CHD_count.shape

(50, 3)


cigarrillo_CHD_count.head(30)


#grafico tipo de propiedad por provincia/region#
fig = px.scatter(data_frame = cigarrillo_CHD_count, x = "TenYearCHD", y = "cigsPerDay",
                 size = "Group_Count", color = "cigsPerDay", opacity = 0.6,
                width=700, height=800)

fig.show()


EDAD_CHD_grouped =data_para_regresion.groupby( ["TenYearCHD", "age"] )
EDAD_CHD_count = pd.DataFrame(EDAD_CHD_grouped.size().reset_index(name = "Group_Count"))
EDAD_CHD_count.shape

(72, 3)


EDAD_CHD_count.head(50)


#grafico tipo de propiedad por provincia/region#
fig = px.scatter(data_frame = EDAD_CHD_count, x = "TenYearCHD", y = "age",
                 size = "Group_Count", color = "age", opacity = 0.6,
                width=700, height=800)

fig.show()


fig = px.density_heatmap(data_para_regresion, x="age", y="TenYearCHD", facet_row="male" , color_continuous_scale=px.colors.diverging.PRGn)
fig.show()


# vemos cuántos mujeres y hombres hay en el dataset 
data_para_regresion.male.value_counts()

0    608
1    562
Name: male, dtype: int64


data_inf_sin_nulos.head(4) #### NUESTRO DF LIMPIO PERO DESBALANCEADO:


X = data_inf_sin_nulos[["male", "age", "education", "currentSmoker", "cigsPerDay", "BPMeds", "prevalentStroke", "prevalentHyp", "diabetes", "totChol", "sysBP", "diaBP", "BMI", "heartRate", "glucose"]]
y = data_inf_sin_nulos.TenYearCHD


X_train, X_test, y_train, y_test = train_test_split(X, y, stratify=y, random_state = 12)


scaler = StandardScaler()
X_train = scaler.fit_transform(X_train)
X_test = scaler.transform(X_test)


logistic_regression = LogisticRegression(penalty='none') # sin regularización
logistic_regression.fit(X_train, y_train);


y_test_pred = logistic_regression.predict(X_test)
y_test_pred_proba = logistic_regression.predict_proba(X_test)


tn, fp, fn, tp =  confusion_matrix(y_test, y_test_pred).ravel()
(tn, fp, fn, tp)

(767, 8, 127, 12)


print("habia" , tn+fp , "que NO iban a tener problemas cardiacos ,  de los cuales :" , tn , "fueron acertados y los que fueron incorrectos son " , fp)
print("habia" , fn+tp , "que SI iban a tener problemas cardiacos ,  de los cuales :" , tp , "fueron acertados y los que fueron incorrectos son " , fn)

habia 775 que NO iban a tener problemas cardiacos ,  de los cuales : 767 fueron acertados y los que fueron incorrectos son  8
habia 139 que SI iban a tener problemas cardiacos ,  de los cuales : 12 fueron acertados y los que fueron incorrectos son  127


sns.heatmap(confusion_matrix(y_test, y_test_pred), annot=True, fmt='g', xticklabels=['Negativo Predicho','Positivo Predicho'], yticklabels=['Negativo Real','Positivo Real'],cmap='Blues')

<AxesSubplot:>


acc=accuracy_score(y_test, y_test_pred)
f1=f1_score(y_test, y_test_pred)
recall=recall_score(y_test, y_test_pred)
preci=precision_score(y_test, y_test_pred)
my_scores=[acc,f1,recall,preci]
fig = px.bar(x=["accuracy_score", "f1_score", "recall_score", "precision_score"],y=my_scores, title="Metricas del modelo de Regresíon Logística con Datos Desbalanceados")
fig.show()


#Separamos las variables predictoras de la variable target
X = data_para_regresion[["male", "age", "education", "currentSmoker", "cigsPerDay", "BPMeds", "prevalentStroke", "prevalentHyp", "diabetes", "totChol", "sysBP", "diaBP", "BMI", "heartRate", "glucose"]]
y = data_para_regresion.TenYearCHD


#Separamos en train y test de forma proporcionada con el stratify
X_train, X_test, y_train, y_test = train_test_split(X, y, stratify=y, random_state = 12)


scaler = StandardScaler()
X_train = scaler.fit_transform(X_train)
X_test = scaler.transform(X_test)


logistic_regression = LogisticRegression(penalty='none') # sin regularización
logistic_regression.fit(X_train, y_train);


y_train_pred = logistic_regression.predict(X_train)
#y_train_pred


accuracy_score(y_train, y_train_pred)

0.6716077537058153


confusion_matrix(y_train, y_train_pred)

array([[322, 137],
       [151, 267]], dtype=int64)


tn, fp, fn, tp =  confusion_matrix(y_train, y_train_pred).ravel()
(tn, fp, fn, tp)

(322, 137, 151, 267)


print("habia" , tn+fp , "que NO iban a tener problemas cardiacos ,  de los cuales :" , tn , "fueron acertados y los que fueron incorrectos son " , fp)
print("habia" , fn+tp , "que SI iban a tener problemas cardiacos ,  de los cuales :" , tp , "fueron acertados y los que fueron incorrectos son " , fn)

habia 459 que NO iban a tener problemas cardiacos ,  de los cuales : 322 fueron acertados y los que fueron incorrectos son  137
habia 418 que SI iban a tener problemas cardiacos ,  de los cuales : 267 fueron acertados y los que fueron incorrectos son  151


y_test_pred = logistic_regression.predict(X_test)
#y_test_pred


y_test_pred_proba = logistic_regression.predict_proba(X_test)
#y_test_pred_proba


accuracy_score(y_test, y_test_pred)

0.6962457337883959


f1_score(y_test, y_test_pred)

0.6691449814126393


recall_score(y_test, y_test_pred)

0.6474820143884892


precision_score(y_test, y_test_pred)

0.6923076923076923


confusion_matrix(y_test, y_test_pred)

array([[114,  40],
       [ 49,  90]], dtype=int64)


#Matriz de confusion
sns.heatmap(confusion_matrix(y_test, y_test_pred), annot=True, fmt='g', xticklabels=['Negativo Predicho','Positivo Predicho'], yticklabels=['Negativo Real','Positivo Real'],cmap='Blues')

<AxesSubplot:>


tn, fp, fn, tp =  confusion_matrix(y_test, y_test_pred).ravel()
(tn, fp, fn, tp)

(114, 40, 49, 90)


print("habia" , tn+fp , "que NO iban a tener problemas cardiacos ,  de los cuales :" , tn , "fueron acertados y los que fueron incorrectos son " , fp)
print("habia" , fn+tp , "que SI iban a tener problemas cardiacos ,  de los cuales :" , tp , "fueron acertados y los que fueron incorrectos son " , fn)

habia 154 que NO iban a tener problemas cardiacos ,  de los cuales : 114 fueron acertados y los que fueron incorrectos son  40
habia 139 que SI iban a tener problemas cardiacos ,  de los cuales : 90 fueron acertados y los que fueron incorrectos son  49


logistic_regression.intercept_

array([-0.10166629])


logistic_regression.coef_

array([[ 0.2851362 ,  0.50653973,  0.0458867 ,  0.13492714,  0.1003944 ,
        -0.06420957,  0.10135185,  0.09317174,  0.10326965,  0.03427066,
         0.48445815, -0.11120784,  0.06289528, -0.01043178,  0.19476592]])


y_test.value_counts()

0    154
1    139
Name: TenYearCHD, dtype: int64


acc=accuracy_score(y_test, y_test_pred)
f1=f1_score(y_test, y_test_pred)
recall=recall_score(y_test, y_test_pred)
preci=precision_score(y_test, y_test_pred)
my_scores=[acc,f1,recall,preci]
fig = px.bar(x=["accuracy_score", "f1_score", "recall_score", "precision_score"],y=my_scores 
             , title="Metricas del modelo de Regresíon Logística")
fig.show()


##X_train, X_test, y_train, y_test 

X_train_stats = sm.add_constant(X_train)


# Instanciamos la clase
logit = sm.Logit(y_train, X_train_stats)
# Fiteamos el modelo
result = logit.fit()
# Imprimimos el resumen
print(result.summary2())

Optimization terminated successfully.
         Current function value: 0.605124
         Iterations 6
                         Results: Logit
=================================================================
Model:              Logit            Pseudo R-squared: 0.126     
Dependent Variable: TenYearCHD       AIC:              1093.3868 
Date:               2021-06-22 23:43 BIC:              1169.8109 
No. Observations:   877              Log-Likelihood:   -530.69   
Df Model:           15               LL-Null:          -606.93   
Df Residuals:       861              LLR p-value:      7.7697e-25
Converged:          1.0000           Scale:            1.0000    
No. Iterations:     6.0000                                       
-------------------------------------------------------------------
            Coef.    Std.Err.      z      P>|z|     [0.025   0.975]
-------------------------------------------------------------------
const      -0.1017     0.0743   -1.3676   0.1714   -0.2474   0.0440
x1          0.2851     0.0807    3.5336   0.0004    0.1270   0.4433
x2          0.5065     0.0882    5.7461   0.0000    0.3338   0.6793
x3          0.0459     0.0763    0.6010   0.5478   -0.1038   0.1955
x4          0.1349     0.1214    1.1115   0.2664   -0.1030   0.3729
x5          0.1004     0.1230    0.8163   0.4144   -0.1407   0.3415
x6         -0.0642     0.0803   -0.7999   0.4238   -0.2215   0.0931
x7          0.1014     0.0870    1.1648   0.2441   -0.0692   0.2719
x8          0.0931     0.1044    0.8922   0.3723   -0.1115   0.2977
x9          0.1033     0.1117    0.9243   0.3553   -0.1157   0.3223
x10         0.0343     0.0789    0.4345   0.6639   -0.1203   0.1888
x11         0.4845     0.1546    3.1338   0.0017    0.1815   0.7875
x12        -0.1112     0.1293   -0.8596   0.3900   -0.3647   0.1423
x13         0.0629     0.0810    0.7763   0.4376   -0.0959   0.2217
x14        -0.0104     0.0767   -0.1360   0.8919   -0.1608   0.1399
x15         0.1948     0.1246    1.5636   0.1179   -0.0494   0.4389
=================================================================


# x1 male ,	x2 age
# x3 education,  x4 currentSmoker
# x5 cigsPerDay, x6	BPMeds, x7 prevalentStroke
#x8 prevalentHyp, x9 diabetes, x10totChol, x11 sysBP	
# x12 diaBP, x13BMI, x14heartRate, x 15	glucose


Edad= 0.0001
Hombre= 0.0004
ACV = 0.2441
Fuma = 0.2664
Diabetes = 0.3553
Hipertension = 0.3723
p_value=[Edad, Hombre, ACV, Fuma, Diabetes, Hipertension ]
fig = px.bar(x=["1_Edad", "2_Hombre", "3_Fuma", "4_ACV", "5_Diabetes", "6_Hipertension"],y=p_value 
             , title="Variables predictoras más relevantes")
fig.show()


Ritmo_cardiaco= 0.8919
Colesterol= 0.6639
Nivel_educativo = 0.5478
p_value=[Ritmo_cardiaco, Colesterol, Nivel_educativo]
fig = px.bar(x=["Ritmo_cardiaco", "Colesterol", "Nivel_educativo"],y=p_value 
             , title="Variables predictoras menos relevantes")
fig.show()


from sklearn.naive_bayes import GaussianNB

gnb = GaussianNB()

gnb.fit(X_train, y_train)

GaussianNB()


y_pred = gnb.predict(X_test)


accuracy_score(y_test, y_pred)

0.6313993174061433


f1_score(y_test, y_pred)

0.443298969072165


recall_score(y_test, y_pred)

0.30935251798561153


precision_score(y_test, y_pred)

0.7818181818181819


tn, fp, fn, tp =  confusion_matrix(y_test, y_pred).ravel()
(tn, fp, fn, tp)

(142, 12, 96, 43)


print("habia" , tn+fp , "que NO iban a tener problemas cardiacos ,  de los cuales :" , tn , "fueron acertados y los que fueron incorrectos son " , fp)
print("habia" , fn+tp , "que SI iban a tener problemas cardiacos ,  de los cuales :" , tp , "fueron acertados y los que fueron incorrectos son " , fn)

habia 154 que NO iban a tener problemas cardiacos ,  de los cuales : 142 fueron acertados y los que fueron incorrectos son  12
habia 139 que SI iban a tener problemas cardiacos ,  de los cuales : 43 fueron acertados y los que fueron incorrectos son  96


acc=accuracy_score(y_test, y_pred)
f1=f1_score(y_test, y_pred)
recall=recall_score(y_test, y_pred)
preci=precision_score(y_test, y_pred)
my_scores=[acc,f1,recall,preci]
fig = px.bar(x=["accuracy_score", "f1_score", "recall_score", "precision_score"],y=my_scores 
             , title="Metricas del modelo de Naive Bayes")
fig.show()


scaler = StandardScaler()
X_train = scaler.fit_transform(X_train)
X_test = scaler.transform(X_test)


knn = KNeighborsClassifier(n_neighbors=59)


y_test_pred_proba = logistic_regression.predict_proba(X_test)
#y_test_pred_proba


# Ajustamos a los datos de entrenamiento.
knn.fit(X_train, y_train)

# Ajustamos a los datos de entrenamiento.
knn.fit(X_train, y_train);

# Predecimos etiquetas para los datos de test.
y_pred = knn.predict(X_test)


accuracy_score(y_test, y_pred)

0.6552901023890785


accuracy_score(y_test, y_pred)

0.6552901023890785


f1_score(y_test, y_pred)

0.5738396624472575


recall_score(y_test, y_pred)

0.4892086330935252


precision_score(y_test, y_pred)

0.6938775510204082


acc=accuracy_score(y_test, y_pred)
f1=f1_score(y_test, y_pred)
recall=recall_score(y_test, y_pred)
preci=precision_score(y_test, y_pred)
my_scores=[acc,f1,recall,preci]
fig = px.bar(x=["accuracy_score", "f1_score", "recall_score", "precision_score"],y=my_scores 
             , title="Metricas del modelo de KNN")
fig.show()


tn, fp, fn, tp =  confusion_matrix(y_test, y_pred).ravel()
(tn, fp, fn, tp)

(124, 30, 71, 68)


print("habia" , tn+fp , "que NO iban a tener problemas cardiacos ,  de los cuales :" , tn , "fueron acertados y los que fueron incorrectos son " , fp)
print("habia" , fn+tp , "que SI iban a tener problemas cardiacos ,  de los cuales :" , tp , "fueron acertados y los que fueron incorrectos son " , fn)

habia 154 que NO iban a tener problemas cardiacos ,  de los cuales : 124 fueron acertados y los que fueron incorrectos son  30
habia 139 que SI iban a tener problemas cardiacos ,  de los cuales : 68 fueron acertados y los que fueron incorrectos son  71


X_train_train, X_validation, y_train_train, y_validation = train_test_split(X_train, y_train, random_state=2)

k_range = list(range(1, 50))
scores = []
for k in k_range:
    knn = KNeighborsClassifier(n_neighbors=k)
    knn.fit(X_train_train, y_train_train)
    y_pred = knn.predict(X_validation)
    scores.append(f1_score(y_validation, y_pred , zero_division=1))


plt.figure(figsize=(4,3))
plt.plot(k_range, scores)
plt.xlabel('Valor de K'); plt.ylabel('Test F1')
plt.grid(); plt.show()


X_train_train, X_validation, y_train_train, y_validation = train_test_split(X_train, y_train, random_state=2)

k_range = list(range(1, 100))
scores = []
for k in k_range:
    knn = KNeighborsClassifier(n_neighbors=k)
    knn.fit(X_train_train, y_train_train)
    y_pred = knn.predict(X_validation)
    scores.append(recall_score(y_validation, y_pred))


plt.figure(figsize=(4,3))
plt.plot(k_range, scores)
plt.xlabel('Valor de K'); plt.ylabel('Test recall')
plt.grid(); plt.show()


from sklearn.metrics import roc_curve, auc
y_probs_nb = gnb.predict_proba(X_test) #calculo las probabilidades para Naive Bayes


fpr_nb,tpr_nb,thr_nb = roc_curve(y_test, y_probs_nb[:,1]) #curva ROC para Naive Bayes
fpr_log,tpr_log,thr_log = roc_curve(y_test, y_test_pred_proba[:,1]) #curva ROC para Logistic Regresion

plt.axis([0, 1.01, 0, 1.01])
plt.xlabel('1 - Specificty')
plt.ylabel('TPR / Sensitivity')
plt.title('ROC Curve')
plt.plot(fpr_nb,tpr_nb)
plt.plot(fpr_log,tpr_log)
plt.plot(np.arange(0,1, step =0.01), np.arange(0,1, step =0.01))
plt.legend(['NB','Logit'])
plt.show()


print('AUC-NB=', auc(fpr_nb, tpr_nb))
print('AUC-Logit=', auc(fpr_log, tpr_log))

AUC-NB= 0.7181164159581427
AUC-Logit= 0.738064094179202

	male	age	education	currentSmoker	cigsPerDay	BPMeds	prevalentStroke	prevalentHyp	diabetes	totChol	sysBP	diaBP	BMI	heartRate	glucose	TenYearCHD
3	0	61	3.0	1	30.0	0.0	0	1	0	225.0	150.0	95.0	28.58	65.0	103.0	1
6	0	63	1.0	0	0.0	0.0	0	0	0	205.0	138.0	71.0	33.11	60.0	85.0	1
15	0	38	2.0	1	20.0	0.0	0	1	0	221.0	140.0	90.0	21.35	95.0	70.0	1
17	0	46	2.0	1	20.0	0.0	0	0	0	291.0	112.0	78.0	23.38	80.0	89.0	1
25	1	47	4.0	1	20.0	0.0	0	0	0	294.0	102.0	68.0	24.18	62.0	66.0	1
...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...
1719	1	48	1.0	1	20.0	0.0	0	0	0	268.0	116.5	82.0	21.34	60.0	82.0	0
3964	0	39	3.0	0	0.0	0.0	0	0	0	213.0	125.0	87.0	16.73	110.0	75.0	0
3325	1	62	1.0	1	5.0	0.0	0	1	0	243.0	157.0	96.0	28.83	75.0	71.0	0
1933	1	41	2.0	1	30.0	0.0	0	0	0	224.0	114.0	68.0	21.42	66.0	107.0	0
163	1	57	2.0	1	1.0	0.0	0	0	0	242.0	102.0	61.0	23.50	70.0	82.0	0

VEAMOS PRIMERO QUE PASA SI USAMOS EL DATASET DESBALANCEADO :¶

REGRESION LOGÍSITCA - SKLEARN¶

COMPARACIÓN DE MÉTRICAS¶

REGRESION LOGÍSITCA - STATSMODEL¶

NAIVE BAYES¶

COMPARACIÓN DE MÉTRICAS¶

KNN¶

VALOR DE K¶

CURVA ROC Y AUC¶

CONCLUSION FINAL¶

La métrica más importante en este tipo de estudios médicos va a ser la Recall Sensitivity.¶

Ya que buscamos tener la mayor relación entre los Verdaderos Positivos ( Predicciones correctas a Gente con un posible CHD ) y Cantidad de casos predichos con CHD (Predicciones correctas a Gente con un posible CHD + Predicciones incorrectas a Gente con un posible CHD)¶

Por lo que elegimos la Regresión Logística¶

	male	age	education	currentSmoker	cigsPerDay	prevalentHyp	totChol	sysBP	diaBP	BMI	heartRate	glucose	TenYearCHD
0	1	39	4.0	0	0.0	0	195.0	106.0	70.0	26.97	80.0	77.0	0
1	0	46	2.0	0	0.0	0	250.0	121.0	81.0	28.73	95.0	76.0	0
2	1	48	1.0	1	20.0	0	245.0	127.5	80.0	25.34	75.0	70.0	0
3	0	61	3.0	1	30.0	1	225.0	150.0	95.0	28.58	65.0	103.0	1
4	0	46	3.0	1	23.0	0	285.0	130.0	84.0	23.10	85.0	85.0	0

	male	age	education	BPMeds	prevalentHyp	diabetes	totChol	sysBP	diaBP	BMI	heartRate	glucose
3749	0	54	1.0	0.0	0	1	223.0	110.0	67.5	21.22	78.0	294.0
1995	0	50	2.0	1.0	1	0	241.0	132.0	85.0	23.81	55.0	84.0
3829	1	55	4.0	0.0	0	0	232.0	132.5	87.0	20.72	72.0	71.0

	TenYearCHD	cigsPerDay	Group_Count
0	0	0.0	335
1	0	1.0	10
2	0	2.0	3
3	0	3.0	15
4	0	4.0	2
5	0	5.0	15
6	0	6.0	3
7	0	7.0	1
8	0	8.0	3
9	0	9.0	12
10	0	10.0	14
11	0	12.0	1
12	0	13.0	1
13	0	14.0	1
14	0	15.0	30
15	0	16.0	1
16	0	17.0	1
17	0	19.0	1
18	0	20.0	112
19	0	25.0	4
20	0	30.0	24
21	0	35.0	2
22	0	40.0	12
23	0	43.0	4
24	0	45.0	2
25	0	50.0	2
26	0	60.0	2
27	1	0.0	272
28	1	1.0	7
29	1	2.0	2

	TenYearCHD	age	Group_Count
0	0	33	2
1	0	34	2
2	0	35	4
3	0	36	18
4	0	37	13
5	0	38	17
6	0	39	30
7	0	40	32
8	0	41	34
9	0	42	30
10	0	43	17
11	0	44	20
12	0	45	24
13	0	46	24
14	0	47	27
15	0	48	26
16	0	49	17
17	0	50	25
18	0	51	23
19	0	52	15
20	0	53	17
21	0	54	27
22	0	55	21
23	0	56	16
24	0	57	16
25	0	58	9
26	0	59	12
27	0	60	17
28	0	61	18
29	0	62	12
30	0	63	9
31	0	64	17
32	0	65	8
33	0	66	3
34	0	67	7
35	0	68	3
36	0	69	1
37	1	35	2
38	1	36	2
39	1	37	4
40	1	38	7
41	1	39	6
42	1	40	11
43	1	41	10
44	1	42	14
45	1	43	11
46	1	44	11
47	1	45	13
48	1	46	13
49	1	47	18

	TenYearCHD	age	Group_Count
0	0	33	2
1	0	34	2
2	0	35	4
3	0	36	18
4	0	37	13
5	0	38	17
6	0	39	30
7	0	40	32
8	0	41	34
9	0	42	30
10	0	43	17
11	0	44	20
12	0	45	24
13	0	46	24
14	0	47	27
15	0	48	26
16	0	49	17
17	0	50	25
18	0	51	23
19	0	52	15
20	0	53	17
21	0	54	27
22	0	55	21
23	0	56	16
24	0	57	16
25	0	58	9
26	0	59	12
27	0	60	17
28	0	61	18
29	0	62	12
30	0	63	9
31	0	64	17
32	0	65	8
33	0	66	3
34	0	67	7
35	0	68	3
36	0	69	1
37	1	35	2
38	1	36	2
39	1	37	4
40	1	38	7
41	1	39	6
42	1	40	11
43	1	41	10
44	1	42	14
45	1	43	11
46	1	44	11
47	1	45	13
48	1	46	13
49	1	47	18

	TenYearCHD	age	Group_Count
0	0	33	2
1	0	34	2
2	0	35	4
3	0	36	18
4	0	37	13
5	0	38	17
6	0	39	30
7	0	40	32
8	0	41	34
9	0	42	30
10	0	43	17
11	0	44	20
12	0	45	24
13	0	46	24
14	0	47	27
15	0	48	26
16	0	49	17
17	0	50	25
18	0	51	23
19	0	52	15
20	0	53	17
21	0	54	27
22	0	55	21
23	0	56	16
24	0	57	16
25	0	58	9
26	0	59	12
27	0	60	17
28	0	61	18
29	0	62	12
30	0	63	9
31	0	64	17
32	0	65	8
33	0	66	3
34	0	67	7
35	0	68	3
36	0	69	1
37	1	35	2
38	1	36	2
39	1	37	4
40	1	38	7
41	1	39	6
42	1	40	11
43	1	41	10
44	1	42	14
45	1	43	11
46	1	44	11
47	1	45	13
48	1	46	13
49	1	47	18