Ensembles voor outlier-classificatie

Anomaly Detection in Python

Bekhruz (Bex) Tuychiev

Kaggle Master, Data Science Content Creator

Terug naar Airbnb

outliers = airbnb_df[iforest.labels_ == 1]

outlier_probs = iforest.predict_proba(outliers)
outlier_probs[:10]

array([[0.51999538, 0.48000462],
       [0.61789522, 0.38210478],
       [0.61802032, 0.38197968],
       [0.35184434, 0.64815566],
       [0.57533286, 0.42466714],
       [0.59038933, 0.40961067],
       [0.57677613, 0.42322387],
       [0.54158826, 0.45841174],
       [0.49118093, 0.50881907],
       [0.21387357, 0.78612643]])
Anomaly Detection in Python

Best practice: kandsdrempel

Drempel:

  • 75% bij laag risico

  • 90% bij hoge kosten, zoals:

    • geneeskunde
    • cyberbeveiliging
    • weinig data
Anomaly Detection in Python

Wat is een ensemble?

  • Combinatie van twee of meer classifiers
  • Voorspellingen zijn stabieler
Anomaly Detection in Python

Bekijk de data

google.head()

              Open    High     Low   Close    Volume  day_of_week  month  day
Date                                                                         
2006-01-03  211.47  218.05  209.32  217.83  13137450            1      1    3
2006-01-04  222.17  224.70  220.09  222.84  15292353            2      1    4
2006-01-05  223.22  226.00  220.97  225.85  10815661            3      1    5
2006-01-06  228.66  235.49  226.85  233.06  17759521            4      1    6
2006-01-09  233.44  236.94  230.70  233.68  12795837            0      1    9
Anomaly Detection in Python

Numerieke features schalen

from sklearn.preprocessing import QuantileTransformer


# Definieer de te schalen kolommen
to_scale = ['Open', 'High', 'Low', 'Close', 'Volume']


# Initialiseer de transformer
qt = QuantileTransformer(output_distribution="normal")

# Schaal en schrijf de kolommen terug
google.loc[:, to_scale] = qt.fit_transform(google[to_scale])
Anomaly Detection in Python

Arrays maken

# Maak een lijst met estimators
estimators = [KNN(n_neighbors=20), LOF(n_neighbors=20), IForest()]


# Maak een lege array
shape = (len(google), len(estimators))
probability_scores = np.empty(shape=shape)
Anomaly Detection in Python

In de loop

estimators = [KNN(n_neighbors=20), LOF(n_neighbors=20), IForest()]

shape = (len(google), len(estimators))
probability_scores = np.empty(shape=shape)


# Loopen en fitten
for index, est in enumerate(estimators):
    est.fit(google)


    # Maak waarschijnlijkheden
    probs = est.predict_proba(google)


    # Sla de probs op
    probability_scores[:, index] = probs[:, 1]
Anomaly Detection in Python

Aggregatie - gemiddelde

mean_scores = np.mean(probability_scores, axis=1)

mean_scores

array([0.20699869, 0.21455413, 0.17166271, ..., 0.31255075, 0.33553513,
       0.32217186])
Anomaly Detection in Python

Aggregatie - mediaan

median_scores = np.mean(probability_scores, axis=1)

median_scores

array([0.20699869, 0.21455413, 0.17166271, ..., 0.31255075, 0.33553513,
       0.32217186])
Anomaly Detection in Python

Kansfilter

# Maak een mask met 75%-drempel
is_outlier = median_scores > 0.75


# Filter de outliers
outliers = google[is_outlier]


len(outliers)

3
Anomaly Detection in Python

Stappenoverzicht

# Maak een lijst met estimators
estimators = [KNN(n_neighbors=20), LOF(n_neighbors=20), IForest()]
probability_scores = np.empty(shape=(len(google), len(estimators)))


for index, est in enumerate(estimators):
    # Fit en genereer probabiliteiten
    est.fit(google)
    probs = est.predict_proba(google)

    # Sla de probabiliteiten op
    probability_scores[:, index] = probs[:, 1]
Anomaly Detection in Python

Stappenoverzicht

# Neem het gemiddelde van de scores
mean_scores = np.mean(probability_scores, axis=1)

# Filter met 75%-drempel
outliers = google[mean_scores > 0.75]

print(len(outliers))

3
Anomaly Detection in Python

Laten we oefenen!

Anomaly Detection in Python

Preparing Video For Download...