Comparison with other outlier detection methods

In this example, we compare RRCF with the outlier detection methods included in scikit-learn. We use the same test datasets used in the scikit-learn outlier detection documentation.

Import modules

import numpy as np
import pandas as pd
import matplotlib.pyplot as plt
import seaborn as sns
from sklearn.datasets import make_moons, make_blobs
from sklearn import svm
from sklearn.covariance import EllipticEnvelope
from sklearn.ensemble import IsolationForest
from sklearn.neighbors import LocalOutlierFactor
import time
import rrcf

Run comparisons

rng = np.random.RandomState(42)


# Example settings
n_samples = 300
outliers_fraction = 0.15
n_outliers = int(outliers_fraction * n_samples)
n_inliers = n_samples - n_outliers



# Outlier detectors from sklean plot
anomaly_algorithms = [
    ("Robust covariance",
     EllipticEnvelope(contamination=outliers_fraction)),
    ("One-Class SVM",
     svm.OneClassSVM(nu=outliers_fraction,
                     kernel="rbf",
                     gamma=0.1)),
    ("Isolation Forest",
     IsolationForest(contamination=outliers_fraction,
                     behaviour='new')),
    ("Local Outlier Factor",
     LocalOutlierFactor(n_neighbors=35,
                        contamination=outliers_fraction))]

# Define datasets
blobs_params = dict(random_state=0,
                    n_samples=n_inliers,
                    n_features=2)
datasets = [
    make_blobs(centers=[[0, 0], [0, 0]],
               cluster_std=0.5,**blobs_params)[0],
    make_blobs(centers=[[2, 2], [-2, -2]],
               cluster_std=[0.5, 0.5],**blobs_params)[0],
    make_blobs(centers=[[2, 2], [-2, -2]],
               cluster_std=[1.5, .3],**blobs_params)[0],
    4. * (make_moons(n_samples=n_samples,
                     noise=.05, random_state=0)[0]
                     - np.array([0.5, 0.25])),
    14. * (np.random.RandomState(42).rand(n_samples, 2)
           - 0.5)]

# Add outliers to the data sets
outliers = []  # record keeping
data = []
for i in datasets:
    out = rng.uniform(low=-6, high=6,
                      size=(n_outliers, 2))
    outliers.append(out)
    data.append(np.concatenate([i, out], axis=0))

# Set forest params
avg_codisp = []
num_trees  = 100
tree_size = 256
plot_num = 1

fig = plt.figure(1, figsize=(12, 12))

for d in range(len(data)):
    forest = []
    n = len(data[d])
    tr1 = time.time()
    while len(forest) < num_trees:
        # Select random subsets of points uniformly from point set
        ixs = np.random.choice(n,
                               size=(n // tree_size, tree_size),
                               replace=False)
        # Add sampled trees to forest
        trees = [rrcf.RCTree(data[d][ix],
                             index_labels=ix) for ix in ixs]
        forest.extend(trees)

    # Compute average CoDisp
    avg_codisp_d = pd.Series(0.0, index=np.arange(n))
    index = np.zeros(n)
    for tree in forest:
        codisp = pd.Series({leaf : tree.codisp(leaf)
                            for leaf in tree.leaves})
        avg_codisp_d[codisp.index] += codisp
        np.add.at(index, codisp.index.values, 1)
    avg_codisp_d /= index

    avg_codisp.append(avg_codisp_d)
    tr2 = time.time()


    for name, algorithm in anomaly_algorithms:
        t0 = time.time()
        algorithm.fit(data[d])
        t1 = time.time()
        plt.subplot(5, len(anomaly_algorithms) + 1,
                    plot_num)
        if d == 0:
            plt.title(name, size=16)

        # fit the data and tag outliers
        if name == "Local Outlier Factor":
            y_pred = algorithm.fit_predict(data[d])
        else:
            y_pred = (algorithm.fit(data[d])
                      .predict(data[d]))

        colors = np.array(['#377eb8', '#ff7f00'])
        plt.scatter(data[d][:, 0], data[d][:, 1], s=10,
                    color=colors[(y_pred + 1) // 2])
        plt.text(.99, .01,
                 ('%.2fs' % (t1 - t0)).lstrip('0'),
                 transform=plt.gca().transAxes, size=15,
                 horizontalalignment='right')
        plot_num += 1

    plt.subplot(5, len(anomaly_algorithms) + 1, plot_num)
    avg_cod = avg_codisp[-1]
    mask = np.percentile(avg_cod, 85)
    avg_cod[avg_cod < mask] = 1
    avg_cod[avg_cod > mask] = 0
    c = ['#377eb8' if i == 0 else '#ff7f00'
         for i in avg_cod]
    plt.scatter(data[d][:,0], data[d][:,1], s=10, c=c)
    if d == 0:
        plt.title("RRCF", size=16)

        plt.text(.99, .01,
                 ('%.2fs' % (tr2 - tr1)).lstrip('0'),
                  transform=plt.gca().transAxes, size=15,
                  horizontalalignment='right')
    plot_num += 1