SCORING BANCAIRE (Prédiction bon et mauvais clients pour des emprunts)¶
Lorsqu'une banque prête de l'argent à une personne, elle prend le risque que cette dernière ne rembourse pas cet argent dans le délai convenu. Ce risque est appelé Risque de Crédit. Alors avant d'octroyer un crédit, les banques vérifient si le client (ou la cliente) qui demandent un prêt sera capable ou pas de le rembourser.¶
Cette vérification se fait grâce à l'analyse de plusieurs paramètres tels que les revenus, les biens, les dépenses actuelles du client, etc. Cette analyse est encore effectuée manuellement par plusieurs banques. Ainsi, elle est très consommatrice en temps et en ressources financières.¶
Grâce au Machine Learning, il est possible d'automatiser cette tâche et de pouvoir prédire avec plus de précision les clients qui seront en défaut de paiement.¶
Technologies Utilisées : Python, Machine Learning, Traitement de Données.¶
Dans ce projet, j’ai :¶
- Nettoyer et structurer des données brutes pour faciliter leur exploitation.
- Appliquer des algorithmes d’apprentissage automatique afin de tirer des insights précis et pertinents des données analysées.
- Visualiser les résultats sous forme de graphiques et rapports exploitables pour aider à la prise de décision.
- Développer un modèle de Machine Learning pour la prédiction.
CODING TIME ...¶
1- CHARGEMENT DES PACKAGES NECESSAIRES¶
In [72]:
import numpy as np
import pandas as pd
import matplotlib.pyplot as plt
import seaborn as sns
import sklearn.metrics as metrics
from sklearn.metrics import f1_score, confusion_matrix, classification_report
from sklearn.model_selection import learning_curve
from sklearn.model_selection import train_test_split
from sklearn.linear_model import LogisticRegression
from sklearn.preprocessing import LabelEncoder,RobustScaler
from sklearn.pipeline import make_pipeline
2- NETTOYAGE DE LA BASE DE DONNEES¶
ANALYSE DE LA FORME DES DONNEES¶
In [74]:
url='banking.csv'
data= pd.read_csv(url)
In [76]:
data.head()
Out[76]:
| age | job | marital | education | default | housing | loan | contact | month | day_of_week | ... | campaign | pdays | previous | poutcome | emp_var_rate | cons_price_idx | cons_conf_idx | euribor3m | nr_employed | y | |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| 0 | 44 | blue-collar | married | basic.4y | unknown | yes | no | cellular | aug | thu | ... | 1 | 999 | 0 | nonexistent | 1.4 | 93.444 | -36.1 | 4.963 | 5228.1 | 0 |
| 1 | 53 | technician | married | unknown | no | no | no | cellular | nov | fri | ... | 1 | 999 | 0 | nonexistent | -0.1 | 93.200 | -42.0 | 4.021 | 5195.8 | 0 |
| 2 | 28 | management | single | university.degree | no | yes | no | cellular | jun | thu | ... | 3 | 6 | 2 | success | -1.7 | 94.055 | -39.8 | 0.729 | 4991.6 | 1 |
| 3 | 39 | services | married | high.school | no | no | no | cellular | apr | fri | ... | 2 | 999 | 0 | nonexistent | -1.8 | 93.075 | -47.1 | 1.405 | 5099.1 | 0 |
| 4 | 55 | retired | married | basic.4y | no | yes | no | cellular | aug | fri | ... | 1 | 3 | 1 | success | -2.9 | 92.201 | -31.4 | 0.869 | 5076.2 | 1 |
5 rows × 21 columns
In [78]:
data.info()
<class 'pandas.core.frame.DataFrame'> RangeIndex: 41188 entries, 0 to 41187 Data columns (total 21 columns): # Column Non-Null Count Dtype --- ------ -------------- ----- 0 age 41188 non-null int64 1 job 41188 non-null object 2 marital 41188 non-null object 3 education 41188 non-null object 4 default 41188 non-null object 5 housing 41188 non-null object 6 loan 41188 non-null object 7 contact 41188 non-null object 8 month 41188 non-null object 9 day_of_week 41188 non-null object 10 duration 41188 non-null int64 11 campaign 41188 non-null int64 12 pdays 41188 non-null int64 13 previous 41188 non-null int64 14 poutcome 41188 non-null object 15 emp_var_rate 41188 non-null float64 16 cons_price_idx 41188 non-null float64 17 cons_conf_idx 41188 non-null float64 18 euribor3m 41188 non-null float64 19 nr_employed 41188 non-null float64 20 y 41188 non-null int64 dtypes: float64(5), int64(6), object(10) memory usage: 6.6+ MB
In [80]:
for col in data.select_dtypes('object'):
status = data[col]
print(status.value_counts())
job admin. 10422 blue-collar 9254 technician 6743 services 3969 management 2924 retired 1720 entrepreneur 1456 self-employed 1421 housemaid 1060 unemployed 1014 student 875 unknown 330 Name: count, dtype: int64 marital married 24928 single 11568 divorced 4612 unknown 80 Name: count, dtype: int64 education university.degree 12168 high.school 9515 basic.9y 6045 professional.course 5243 basic.4y 4176 basic.6y 2292 unknown 1731 illiterate 18 Name: count, dtype: int64 default no 32588 unknown 8597 yes 3 Name: count, dtype: int64 housing yes 21576 no 18622 unknown 990 Name: count, dtype: int64 loan no 33950 yes 6248 unknown 990 Name: count, dtype: int64 contact cellular 26144 telephone 15044 Name: count, dtype: int64 month may 13769 jul 7174 aug 6178 jun 5318 nov 4101 apr 2632 oct 718 sep 570 mar 546 dec 182 Name: count, dtype: int64 day_of_week thu 8623 mon 8514 wed 8134 tue 8090 fri 7827 Name: count, dtype: int64 poutcome nonexistent 35563 failure 4252 success 1373 Name: count, dtype: int64
In [82]:
df = data.copy()
In [84]:
df.dtypes.value_counts().plot.pie()
Out[84]:
<Axes: ylabel='count'>
In [86]:
taux_valeurs_manquantes =(df.isna().sum()/df.shape[0]).sort_values(ascending=True)
taux_valeurs_manquantes
Out[86]:
age 0.0 euribor3m 0.0 cons_conf_idx 0.0 cons_price_idx 0.0 emp_var_rate 0.0 poutcome 0.0 previous 0.0 pdays 0.0 campaign 0.0 nr_employed 0.0 duration 0.0 month 0.0 contact 0.0 loan 0.0 housing 0.0 default 0.0 education 0.0 marital 0.0 job 0.0 day_of_week 0.0 y 0.0 dtype: float64
IMPUTATION¶
In [88]:
# b. Drop unrepresentative features
#df.drop(columns = ["month", "previous", "day_of_week", "pdays"],inplace = True)
#df.head()
#Numerical features: ["age", "balance", "duration", "campaign"]
#Categorical Features: ["job", "marital", "education", "default", "housing", "loan", "poutcome", "y"]
def imputation(df):
df = df.dropna(axis=0)
df.drop(columns = ["month", "previous", "day_of_week", "pdays","emp_var_rate","cons_price_idx","cons_conf_idx",
"euribor3m","nr_employed","contact"],inplace = True)
return df
SUPPRESSION DES INCOHERENCES¶
In [90]:
def deleteInconherenteVal(df):
for col in df.select_dtypes('object'):
df[col].replace(["unknown"],df[col].mode(),inplace = True)
return df
ENCODAGE¶
In [92]:
def encodage(df):
for col in df.select_dtypes('object'):
encoder = LabelEncoder()
encode = encoder.fit_transform(df[col].unique())
cl=df[col].unique()
code = {a:b for a,b in zip(cl,encode)}
df[col] = df[col].map(code)
#print(encode)
#print(cl)
#print(code)
return df
DEFINITION DE LA FONCTION DE NETTOYAGE DE LA BASE DE DONNEES¶
In [94]:
def apurement(df):
df=imputation(df)
df=deleteInconherenteVal(df)
df=encodage(df)
return df
EXECUTION DE LA FONCTION DE NETTOYAGE DE LA BASE DE DONNEES¶
In [96]:
df_apurer=apurement(df)
df_apurer.head()
C:\Users\lucre\AppData\Local\Temp\ipykernel_22932\4192318323.py:3: FutureWarning: A value is trying to be set on a copy of a DataFrame or Series through chained assignment using an inplace method.
The behavior will change in pandas 3.0. This inplace method will never work because the intermediate object on which we are setting values always behaves as a copy.
For example, when doing 'df[col].method(value, inplace=True)', try using 'df.method({col: value}, inplace=True)' or df[col] = df[col].method(value) instead, to perform the operation inplace on the original object.
df[col].replace(["unknown"],df[col].mode(),inplace = True)
Out[96]:
| age | job | marital | education | default | housing | loan | duration | campaign | poutcome | y | |
|---|---|---|---|---|---|---|---|---|---|---|---|
| 0 | 44 | 1 | 1 | 0 | 0 | 1 | 0 | 210 | 1 | 1 | 0 |
| 1 | 53 | 9 | 1 | 6 | 0 | 0 | 0 | 138 | 1 | 1 | 0 |
| 2 | 28 | 4 | 2 | 6 | 0 | 1 | 0 | 339 | 3 | 2 | 1 |
| 3 | 39 | 7 | 1 | 3 | 0 | 0 | 0 | 185 | 2 | 1 | 0 |
| 4 | 55 | 5 | 1 | 0 | 0 | 1 | 0 | 137 | 1 | 2 | 1 |
MATRICE DE CORRELATION DES VARIABLES¶
In [98]:
plt.figure(figsize=(20,10))
sns.heatmap(df_apurer.corr(),annot = True)
Out[98]:
<Axes: >
3- Analyses Descriptives (Duration,âge,job,default)¶
Duration Feature¶
In [115]:
f, (ax_box, ax_hist) = plt.subplots(2, sharex=True) # gridspec_kw= {"height_ratios": (0.3, 1)}
sns.set(rc={'figure.figsize':(11,8)}, font_scale=1.5, style='whitegrid')
mean=df_apurer['duration'].mean()
median=df_apurer['duration'].median()
mode=df_apurer['duration'].mode().values[0]
duration = sns.boxplot(data=df_apurer, x="duration", y="y", ax=ax_box, order = df_apurer["y"].value_counts().index)
duration.set(xscale="log")
ax_box.axvline(mean, color='r', linestyle='--')
ax_box.axvline(median, color='g', linestyle='-')
ax_box.axvline(mode, color='b', linestyle='-')
sns.histplot(data=df_apurer, x="duration", ax=ax_hist, kde=True)
ax_hist.axvline(mean, color='r', linestyle='--', label="Mean")
ax_hist.axvline(median, color='g', linestyle='-', label="Median")
ax_hist.axvline(mode, color='b', linestyle='-', label="Mode")
ax_hist.legend()
ax_box.set(xlabel='')
plt.show()
Age Feature¶
In [203]:
mean=df_apurer['age'].mean()
median=df_apurer['age'].median()
mode=df_apurer['age'].mode().values[0]
sns.boxplot(data=df_apurer, x="age", y="y")
sns.histplot(data=df_apurer, x="age",kde=True)
plt.show()
Job feature¶
In [146]:
sns.countplot(x="job", data = df_apurer, hue = "y", order = df_apurer["job"].value_counts().index)
plt.title("Analyse de la relation entre 'job' et 'y'")
plt.show()
Default feature¶
In [148]:
default = sns.countplot(x="default", data = df_apurer, hue = "y", order = df_apurer["default"].value_counts().index)
plt.title("Analyse bivariée entre 'default' et 'y'")
plt.show()
4- Modélisation¶
In [150]:
trainset, testset = train_test_split(df_apurer, test_size=0.2, random_state=0)
In [152]:
trainset['y'].value_counts()
Out[152]:
y 0 29223 1 3727 Name: count, dtype: int64
In [154]:
testset['y'].value_counts()
Out[154]:
y 0 7325 1 913 Name: count, dtype: int64
In [156]:
def preprocessing(df):
df = encodage(df)
df = deleteInconherenteVal(df)
#df = imputation(df)
X = df.drop('y', axis=1)
y = df['y']
print(y.value_counts())
return X, y
In [158]:
X_train, y_train = preprocessing(df_apurer)
y 0 36548 1 4640 Name: count, dtype: int64
In [160]:
X_test, y_test = preprocessing(testset)
y 0 7325 1 913 Name: count, dtype: int64
In [162]:
model = LogisticRegression(solver='liblinear')
model.fit(X_train, y_train)
model.score(X_test, y_test)
Out[162]:
0.8983976693372178
In [164]:
def evaluation(model):
model.fit(X_train, y_train)
ypred = model.predict(X_test)
print(confusion_matrix(y_test, ypred))
print(classification_report(y_test, ypred))
5- EVALUATION DU MODELE¶
In [166]:
evaluation(model)
[[7217 108]
[ 729 184]]
precision recall f1-score support
0 0.91 0.99 0.95 7325
1 0.63 0.20 0.31 913
accuracy 0.90 8238
macro avg 0.77 0.59 0.63 8238
weighted avg 0.88 0.90 0.87 8238