# Librairies
```python
import numpy as np
import matplotlib.pyplot as plt
import seaborn as sns
import pandas as pd
from sklearn.neighbors import LocalOutlierFactor
```
# Load data
```python
# SImulation d'un jeu de données
np.random.seed(0)
m=100
X = np.random.randn(m).reshape(-1,1)
y = 3 * x - 2 + np.random.randn(m).reshape(-1,1)
X = np.concatenate((x, [[2]]), axis=0)
y = np.concatenate((y, [[-8]]), axis=0)
plt.scatter(X, y)
```
<matplotlib.collections.PathCollection at 0x7029325cd610>
```python
# méthode IQR
```
```python
fig, ax = plt.subplots(1, 2, figsize=(9, 4))
ax[0].boxplot(y)
ax[0].set_title("y")
ax[1].boxplot(X, vert=False)
ax[1].set_title("X")
plt.show()
```
on ne détecte pas d'outliers avec les boxplots --> la méthode du IQR ne suffit pas
```python
data = np.concatenate((X, y), axis=1)
```
```python
detection_model = LocalOutlierFactor(contamination=0.01, n_neighbors=5)
outliers = detection_model.fit_predict(data)
plt.scatter(X, y, c=outliers, cmap='bwr_r')
```
<matplotlib.collections.PathCollection at 0x702931979310>
