ex3TP1

November 5, 2024

[63]: #importation des bibliotheques

import numpy as np
import pandas as pd
import matplotlib.pyplot as plt
import seaborn as sns
import sklearn
from sklearn.linear_model import LinearRegression
from sklearn.model_selection import train_test_split
from sklearn.metrics import mean_squared_error, r2_score
from sklearn.preprocessing import StandardScaler
from sklearn.model_selection import train_test_split
from sklearn.preprocessing import StandardScaler
from tensorflow.keras.models import Sequential
from tensorflow.keras.layers import Dense,Input, Dropout
from tensorflow.keras.optimizers import Adam

[5]: #
df=pd.read_csv("Healthcare-Diabetes.csv")

[6]: df.head()

[6]: Id Pregnancies Glucose BloodPressure SkinThickness Insulin BMI \

0 1 6 148 72 35 0 33.6
1 2 1 85 66 29 0 26.6
2 3 8 183 64 0 0 23.3
3 4 1 89 66 23 94 28.1
4 5 0 137 40 35 168 43.1

DiabetesPedigreeFunction Age Outcome

0 0.627 50 1
1 0.351 31 0
2 0.672 32 1
3 0.167 21 0
4 2.288 33 1

[7]: df.tail()

1
[7]: Id Pregnancies Glucose BloodPressure SkinThickness Insulin BMI \
2763 2764 2 75 64 24 55 29.7
2764 2765 8 179 72 42 130 32.7
2765 2766 6 85 78 0 0 31.2
2766 2767 0 129 110 46 130 67.1
2767 2768 2 81 72 15 76 30.1

DiabetesPedigreeFunction Age Outcome

2763 0.370 33 0
2764 0.719 36 1
2765 0.382 42 0
2766 0.319 26 1
2767 0.547 25 0

[8]: df.info()

<class 'pandas.core.frame.DataFrame'>
RangeIndex: 2768 entries, 0 to 2767
Data columns (total 10 columns):
# Column Non-Null Count Dtype
--- ------ -------------- -----
0 Id 2768 non-null int64
1 Pregnancies 2768 non-null int64
2 Glucose 2768 non-null int64
3 BloodPressure 2768 non-null int64
4 SkinThickness 2768 non-null int64
5 Insulin 2768 non-null int64
6 BMI 2768 non-null float64
7 DiabetesPedigreeFunction 2768 non-null float64
8 Age 2768 non-null int64
9 Outcome 2768 non-null int64
dtypes: float64(2), int64(8)
memory usage: 216.4 KB

[9]: df.describe()

[9]: Id Pregnancies Glucose BloodPressure SkinThickness \

count 2768.000000 2768.000000 2768.000000 2768.000000 2768.000000
mean 1384.500000 3.742775 121.102601 69.134393 20.824422
std 799.197097 3.323801 32.036508 19.231438 16.059596
min 1.000000 0.000000 0.000000 0.000000 0.000000
25% 692.750000 1.000000 99.000000 62.000000 0.000000
50% 1384.500000 3.000000 117.000000 72.000000 23.000000
75% 2076.250000 6.000000 141.000000 80.000000 32.000000
max 2768.000000 17.000000 199.000000 122.000000 110.000000

Insulin BMI DiabetesPedigreeFunction Age \

count 2768.000000 2768.000000 2768.000000 2768.000000

2
 mean 80.127890 32.137392 0.471193 33.132225
std 112.301933 8.076127 0.325669 11.777230
min 0.000000 0.000000 0.078000 21.000000
25% 0.000000 27.300000 0.244000 24.000000
50% 37.000000 32.200000 0.375000 29.000000
75% 130.000000 36.625000 0.624000 40.000000
max 846.000000 80.600000 2.420000 81.000000

Outcome
count 2768.000000
mean 0.343931
std 0.475104
min 0.000000
25% 0.000000
50% 0.000000
75% 1.000000
max 1.000000

[10]: df.isnull().sum()

[10]: Id 0
Pregnancies 0
Glucose 0
BloodPressure 0
SkinThickness 0
Insulin 0
BMI 0
DiabetesPedigreeFunction 0
Age 0
Outcome 0
dtype: int64

[11]: df.hist(figsize=(15,15))

[11]: array([[<Axes: title={'center': 'Id'}>,

<Axes: title={'center': 'Pregnancies'}>,
<Axes: title={'center': 'Glucose'}>],
[<Axes: title={'center': 'BloodPressure'}>,
<Axes: title={'center': 'SkinThickness'}>,
<Axes: title={'center': 'Insulin'}>],
[<Axes: title={'center': 'BMI'}>,
<Axes: title={'center': 'DiabetesPedigreeFunction'}>,
<Axes: title={'center': 'Age'}>],
[<Axes: title={'center': 'Outcome'}>, <Axes: >, <Axes: >]],
dtype=object)

3
[12]: plt.figure()
sns.scatterplot(x='Pregnancies',y='Glucose',data=df,␣
↪hue="Outcome",palette="coolwarm")

[12]: <Axes: xlabel='Pregnancies', ylabel='Glucose'>

4
[13]: df.corr()

[13]: Id Pregnancies Glucose BloodPressure \

Id 1.000000 -0.024222 0.015010 0.009717
Pregnancies -0.024222 1.000000 0.122839 0.147253
Glucose 0.015010 0.122839 1.000000 0.142095
BloodPressure 0.009717 0.147253 0.142095 1.000000
SkinThickness 0.017702 -0.068673 0.061023 0.201167
Insulin 0.007359 -0.075734 0.323445 0.087823
BMI 0.024007 0.018761 0.225308 0.281560
DiabetesPedigreeFunction -0.009695 -0.027731 0.127195 0.048471
Age -0.007404 0.540805 0.256958 0.238684
Outcome -0.006298 0.223796 0.460644 0.072900

SkinThickness Insulin BMI \

Id 0.017702 0.007359 0.024007
Pregnancies -0.068673 -0.075734 0.018761
Glucose 0.061023 0.323445 0.225308
BloodPressure 0.201167 0.087823 0.281560
SkinThickness 1.000000 0.445345 0.393494
Insulin 0.445345 1.000000 0.215926

5
 BMI 0.393494 0.215926 1.000000
DiabetesPedigreeFunction 0.179830 0.190500 0.129766
Age -0.111895 -0.073458 0.038175
Outcome 0.075603 0.123646 0.280928

DiabetesPedigreeFunction Age Outcome

Id -0.009695 -0.007404 -0.006298
Pregnancies -0.027731 0.540805 0.223796
Glucose 0.127195 0.256958 0.460644
BloodPressure 0.048471 0.238684 0.072900
SkinThickness 0.179830 -0.111895 0.075603
Insulin 0.190500 -0.073458 0.123646
BMI 0.129766 0.038175 0.280928
DiabetesPedigreeFunction 1.000000 0.028544 0.160664
Age 0.028544 1.000000 0.237050
Outcome 0.160664 0.237050 1.000000

[14]: plt.figure(figsize=(10,10))
sns.heatmap(df.corr(),annot=True,cmap="YlGnBu")
plt.show()

6
[15]: #Normalisation des donnees
X = df.drop(['Outcome'], axis=1)
X_norm=X.apply(lambda x:(x-x.min())/(x.max()-x.min()))
X_norm.head()

[15]: Id Pregnancies Glucose BloodPressure SkinThickness Insulin \

0 0.000000 0.352941 0.743719 0.590164 0.318182 0.000000
1 0.000361 0.058824 0.427136 0.540984 0.263636 0.000000
2 0.000723 0.470588 0.919598 0.524590 0.000000 0.000000
3 0.001084 0.058824 0.447236 0.540984 0.209091 0.111111
4 0.001446 0.000000 0.688442 0.327869 0.318182 0.198582

7
 BMI DiabetesPedigreeFunction Age
0 0.416873 0.234415 0.483333
1 0.330025 0.116567 0.166667
2 0.289082 0.253629 0.183333
3 0.348635 0.038002 0.000000
4 0.534739 0.943638 0.200000

[16]: Y=df.Outcome
Y

[16]: 0 1
1 0
2 1
3 0
4 1
..
2763 0
2764 1
2765 0
2766 1
2767 0
Name: Outcome, Length: 2768, dtype: int64

[17]: x_train,x_test,y_train,y_test=train_test_split(X_norm,Y,test_size=0.
↪2,random_state=42)

[77]: # Création du modèle

model = Sequential([
Input(shape=(x_train.shape[1],)),# Couche d'entrée pour définir la forme␣
↪des données

#Dense(64, activation='relu'),
#Dropout(0.5),
Dense(1, activation='sigmoid') # Couche de sortie linéaire pour la␣
↪classification binaire

])#reseau de neurone lineaire pour la classification binaire

# Compilation du modèle
model.compile(optimizer=Adam(learning_rate=0.001), loss='binary_crossentropy',␣
↪metrics=['accuracy'])

[53]: #entrainement du modele

#pour 5epochs
train_model = model.fit(x_train, y_train, epochs=5, batch_size=10,␣
↪validation_data=(x_test, y_test))

Epoch 1/5
222/222 �������������������� 6s 12ms/step -
accuracy: 0.6454 - loss: 0.6578 - val_accuracy: 0.6643 - val_loss: 0.6200

8
 Epoch 2/5
222/222 �������������������� 2s 8ms/step -
accuracy: 0.6622 - loss: 0.6111 - val_accuracy: 0.6931 - val_loss: 0.5793
Epoch 3/5
222/222 �������������������� 1s 3ms/step -
accuracy: 0.6993 - loss: 0.5776 - val_accuracy: 0.7274 - val_loss: 0.5511
Epoch 4/5
222/222 �������������������� 1s 5ms/step -
accuracy: 0.7178 - loss: 0.5439 - val_accuracy: 0.7401 - val_loss: 0.5290
Epoch 5/5
222/222 �������������������� 2s 7ms/step -
accuracy: 0.7327 - loss: 0.5335 - val_accuracy: 0.7437 - val_loss: 0.5167

[55]: #pour 10epochs

train_model = model.fit(x_train, y_train, epochs=10, batch_size=10,␣
↪validation_data=(x_test, y_test))

Epoch 1/10
222/222 �������������������� 2s 8ms/step -
accuracy: 0.7596 - loss: 0.5119 - val_accuracy: 0.7653 - val_loss: 0.5010
Epoch 2/10
222/222 �������������������� 2s 10ms/step -
accuracy: 0.7655 - loss: 0.4975 - val_accuracy: 0.7690 - val_loss: 0.4920
Epoch 3/10
222/222 �������������������� 2s 9ms/step -
accuracy: 0.7988 - loss: 0.4638 - val_accuracy: 0.7635 - val_loss: 0.4890
Epoch 4/10
222/222 �������������������� 2s 8ms/step -
accuracy: 0.7766 - loss: 0.4814 - val_accuracy: 0.7653 - val_loss: 0.4807
Epoch 5/10
222/222 �������������������� 2s 8ms/step -
accuracy: 0.8003 - loss: 0.4552 - val_accuracy: 0.7780 - val_loss: 0.4795
Epoch 6/10
222/222 �������������������� 1s 5ms/step -
accuracy: 0.7894 - loss: 0.4643 - val_accuracy: 0.7708 - val_loss: 0.4728
Epoch 7/10
222/222 �������������������� 1s 5ms/step -
accuracy: 0.7857 - loss: 0.4590 - val_accuracy: 0.7690 - val_loss: 0.4733
Epoch 8/10
222/222 �������������������� 1s 3ms/step -
accuracy: 0.7830 - loss: 0.4629 - val_accuracy: 0.7671 - val_loss: 0.4672
Epoch 9/10
222/222 �������������������� 1s 4ms/step -
accuracy: 0.7843 - loss: 0.4461 - val_accuracy: 0.7726 - val_loss: 0.4654
Epoch 10/10
222/222 �������������������� 1s 4ms/step -
accuracy: 0.7980 - loss: 0.4375 - val_accuracy: 0.7726 - val_loss: 0.4642

9
[71]: #pour 15epochs
train_model = model.fit(x_train, y_train, epochs=15, batch_size=10,␣
↪validation_data=(x_test, y_test))

Epoch 1/15
222/222 �������������������� 4s 10ms/step -
accuracy: 0.5684 - loss: 0.6904 - val_accuracy: 0.6606 - val_loss: 0.6560
Epoch 2/15
222/222 �������������������� 1s 6ms/step -
accuracy: 0.5905 - loss: 0.6804 - val_accuracy: 0.6606 - val_loss: 0.6468
Epoch 3/15
222/222 �������������������� 2s 9ms/step -
accuracy: 0.6189 - loss: 0.6707 - val_accuracy: 0.6606 - val_loss: 0.6433
Epoch 4/15
222/222 �������������������� 1s 4ms/step -
accuracy: 0.6450 - loss: 0.6660 - val_accuracy: 0.6625 - val_loss: 0.6406
Epoch 5/15
222/222 �������������������� 1s 4ms/step -
accuracy: 0.6578 - loss: 0.6553 - val_accuracy: 0.6625 - val_loss: 0.6381
Epoch 6/15
222/222 �������������������� 1s 4ms/step -
accuracy: 0.6745 - loss: 0.6425 - val_accuracy: 0.6625 - val_loss: 0.6357
Epoch 7/15
222/222 �������������������� 1s 3ms/step -
accuracy: 0.6579 - loss: 0.6410 - val_accuracy: 0.6625 - val_loss: 0.6327
Epoch 8/15
222/222 �������������������� 1s 3ms/step -
accuracy: 0.6611 - loss: 0.6469 - val_accuracy: 0.6625 - val_loss: 0.6306
Epoch 9/15
222/222 �������������������� 1s 4ms/step -
accuracy: 0.6507 - loss: 0.6564 - val_accuracy: 0.6625 - val_loss: 0.6284
Epoch 10/15
222/222 �������������������� 1s 4ms/step -
accuracy: 0.6521 - loss: 0.6503 - val_accuracy: 0.6625 - val_loss: 0.6265
Epoch 11/15
222/222 �������������������� 1s 4ms/step -
accuracy: 0.6504 - loss: 0.6472 - val_accuracy: 0.6625 - val_loss: 0.6248
Epoch 12/15
222/222 �������������������� 2s 9ms/step -
accuracy: 0.6493 - loss: 0.6382 - val_accuracy: 0.6625 - val_loss: 0.6233
Epoch 13/15
222/222 �������������������� 1s 4ms/step -
accuracy: 0.6600 - loss: 0.6368 - val_accuracy: 0.6625 - val_loss: 0.6219
Epoch 14/15
222/222 �������������������� 2s 9ms/step -
accuracy: 0.6734 - loss: 0.6334 - val_accuracy: 0.6625 - val_loss: 0.6206
Epoch 15/15
222/222 �������������������� 1s 4ms/step -

10
 accuracy: 0.6664 - loss: 0.6299 - val_accuracy: 0.6625 - val_loss: 0.6192

[79]: #pour 25epochs

train_model = model.fit(x_train, y_train, epochs=25, batch_size=10,␣
↪validation_data=(x_test, y_test))

Epoch 1/25
222/222 �������������������� 4s 11ms/step -
accuracy: 0.5764 - loss: 0.6762 - val_accuracy: 0.6805 - val_loss: 0.6574
Epoch 2/25
222/222 �������������������� 2s 9ms/step -
accuracy: 0.6529 - loss: 0.6524 - val_accuracy: 0.6643 - val_loss: 0.6463
Epoch 3/25
222/222 �������������������� 2s 8ms/step -
accuracy: 0.6670 - loss: 0.6378 - val_accuracy: 0.6643 - val_loss: 0.6413
Epoch 4/25
222/222 �������������������� 2s 8ms/step -
accuracy: 0.6518 - loss: 0.6407 - val_accuracy: 0.6661 - val_loss: 0.6367
Epoch 5/25
222/222 �������������������� 2s 9ms/step -
accuracy: 0.6779 - loss: 0.6183 - val_accuracy: 0.6661 - val_loss: 0.6327
Epoch 6/25
222/222 �������������������� 1s 6ms/step -
accuracy: 0.6671 - loss: 0.6263 - val_accuracy: 0.6661 - val_loss: 0.6286
Epoch 7/25
222/222 �������������������� 1s 4ms/step -
accuracy: 0.6619 - loss: 0.6285 - val_accuracy: 0.6661 - val_loss: 0.6249
Epoch 8/25
222/222 �������������������� 1s 3ms/step -
accuracy: 0.6806 - loss: 0.6164 - val_accuracy: 0.6733 - val_loss: 0.6212
Epoch 9/25
222/222 �������������������� 1s 3ms/step -
accuracy: 0.6773 - loss: 0.6158 - val_accuracy: 0.6805 - val_loss: 0.6177
Epoch 10/25
222/222 �������������������� 1s 3ms/step -
accuracy: 0.6771 - loss: 0.6109 - val_accuracy: 0.6805 - val_loss: 0.6144
Epoch 11/25
222/222 �������������������� 1s 4ms/step -
accuracy: 0.6809 - loss: 0.6090 - val_accuracy: 0.6841 - val_loss: 0.6113
Epoch 12/25
222/222 �������������������� 1s 4ms/step -
accuracy: 0.6775 - loss: 0.6096 - val_accuracy: 0.6823 - val_loss: 0.6084
Epoch 13/25
222/222 �������������������� 1s 3ms/step -
accuracy: 0.6732 - loss: 0.6112 - val_accuracy: 0.6859 - val_loss: 0.6054
Epoch 14/25
222/222 �������������������� 1s 5ms/step -
accuracy: 0.6843 - loss: 0.5997 - val_accuracy: 0.6877 - val_loss: 0.6027

11
 Epoch 15/25
222/222 �������������������� 1s 4ms/step -
accuracy: 0.7025 - loss: 0.5918 - val_accuracy: 0.6931 - val_loss: 0.6001
Epoch 16/25
222/222 �������������������� 1s 4ms/step -
accuracy: 0.6991 - loss: 0.5897 - val_accuracy: 0.6913 - val_loss: 0.5977
Epoch 17/25
222/222 �������������������� 1s 6ms/step -
accuracy: 0.6812 - loss: 0.5976 - val_accuracy: 0.6877 - val_loss: 0.5954
Epoch 18/25
222/222 �������������������� 1s 3ms/step -
accuracy: 0.6963 - loss: 0.5865 - val_accuracy: 0.6931 - val_loss: 0.5933
Epoch 19/25
222/222 �������������������� 2s 10ms/step -
accuracy: 0.6807 - loss: 0.5947 - val_accuracy: 0.6931 - val_loss: 0.5912
Epoch 20/25
222/222 �������������������� 2s 7ms/step -
accuracy: 0.6990 - loss: 0.5890 - val_accuracy: 0.6968 - val_loss: 0.5892
Epoch 21/25
222/222 �������������������� 1s 3ms/step -
accuracy: 0.6920 - loss: 0.5914 - val_accuracy: 0.7022 - val_loss: 0.5873
Epoch 22/25
222/222 �������������������� 1s 4ms/step -
accuracy: 0.7050 - loss: 0.5788 - val_accuracy: 0.7022 - val_loss: 0.5854
Epoch 23/25
222/222 �������������������� 1s 3ms/step -
accuracy: 0.6870 - loss: 0.5928 - val_accuracy: 0.7022 - val_loss: 0.5836
Epoch 24/25
222/222 �������������������� 1s 3ms/step -
accuracy: 0.6877 - loss: 0.5791 - val_accuracy: 0.7022 - val_loss: 0.5819
Epoch 25/25
222/222 �������������������� 1s 3ms/step -
accuracy: 0.6906 - loss: 0.5762 - val_accuracy: 0.7040 - val_loss: 0.5803

[45]: # Visualisation de la performance du modèle

# Courbe de précision
plt.plot(train_model.history['accuracy'], label='Train Accuracy')
plt.plot(train_model.history['val_accuracy'], label='Validation Accuracy')
plt.title('Accuracy over epochs')
plt.xlabel('Epochs')
plt.ylabel('Accuracy')
plt.legend()
plt.show()
# Courbe de perte
plt.plot(train_model.history['loss'], label='Train Loss')
plt.plot(train_model.history['val_loss'], label='Validation Loss')
plt.title('Loss over epochs')

12
plt.xlabel('Epochs')
plt.ylabel('Loss')
plt.legend()
plt.show()
#interpretation de accuracy
'''
on voit qu'il y a une difference entre train et validation accuracy au cours␣
↪des epochs la premiere vers 0,68et la deuxieme vers 0,7 c

ela signifie que le modele generalise bien aux nouvelles donnees que les␣
↪donnees d'entrainement

'''
#interpretation de loss
'''
on remarque que validation loss inferieur a training loss ce qui indique la␣
↪devision des donnees est difficile a modeliser et on a underfitting

et le modele n'est pas assez bon pour entrainer les donnees

'''
#performance de modele
'''
on va ajouter le nombre des epochs
'''

13
[81]: # Visualisation de la performance du modèle
# Courbe de précision
plt.plot(train_model.history['accuracy'], label='Train Accuracy')
plt.plot(train_model.history['val_accuracy'], label='Validation Accuracy')
plt.title('Accuracy over epochs')
plt.xlabel('Epochs')
plt.ylabel('Accuracy')
plt.legend()
plt.show()
# Courbe de perte
plt.plot(train_model.history['loss'], label='Train Loss')
plt.plot(train_model.history['val_loss'], label='Validation Loss')
plt.title('Loss over epochs')
plt.xlabel('Epochs')
plt.ylabel('Loss')
plt.legend()
plt.show()

14
15
[ ]: #interpretation d'accuracy
'''
apres l'ajout des epochs on voit que train accuracy et validation accuracy␣
↪augmentent au cours des epochs

ce qui montre que le modèle apprend efficacement sur les données d'entraînement␣
↪tout en restant généralisable aux nouvelles données

'''
#interpretation de loss
'''
apres l'ajout des epochs on voit que train loss et validation loss se diminuent␣
↪vers 0 cela indique que le modele

s’entraîne correctement et généralise bien aux données de validation et fait de␣

↪bonnes prédictions sur les données qu'il n’a pas vues.

'''

[83]: model.summary()
# Le modèle est rapide à entraîner et léger avec seulement 10 paramètres␣
↪entraînables

Model: "sequential_6"

16
 ����������������������������������������������������������������������������
� Layer (type) � Output Shape � Param # �
����������������������������������������������������������������������������
� dense_7 (Dense) � (None, 1) � 10 �
����������������������������������������������������������������������������

Total params: 32 (132.00 B)

Trainable params: 10 (40.00 B)

Non-trainable params: 0 (0.00 B)

Optimizer params: 22 (92.00 B)

[89]: model.save("C:\\Users\\king\\Desktop\\AI PROJECTS\\diabete_model.keras")

[ ]:

17