sklearn-docs/IsolationForest-Model-for-Anomaly-Detection

1import os2 3import pandas as pd4from sklearn.ensemble import IsolationForest5 6import numpy as np7from sklearn.model_selection import train_test_split8import gradio as gr9import matplotlib.pyplot as plt10from skops import hub_utils11import pickle12import time13 14 15 16#Data preparation17n_samples, n_outliers = 120, 4018rng = np.random.RandomState(0)19covariance = np.array([[0.5, -0.1], [0.7, 0.4]])20cluster_1 = 0.4 * rng.randn(n_samples, 2) @ covariance + np.array([2, 2])  # general deformed cluster21cluster_2 = 0.3 * rng.randn(n_samples, 2) + np.array([-2, -2])  # spherical cluster22outliers = rng.uniform(low=-4, high=4, size=(n_outliers, 2))23 24X = np.concatenate([cluster_1, cluster_2, outliers]) #120+120+40 = 280 with 2D25y = np.concatenate(26    [np.ones((2 * n_samples), dtype=int), -np.ones((n_outliers), dtype=int)]27)28 29def load_hf_model_hub():30    '''31    Load the directory containing pretrained model  32    and files from the model repository33    '''34    repo_id="sklearn-docs/anomaly-detection"35    download_repo = "downloaded-model"36    hub_utils.download(repo_id=repo_id, dst=download_repo)37    time.sleep(2)38    loaded_model = pickle.load(open('./downloaded-model/isolation_forest.pkl', 'rb'))39    return loaded_model40 41#Visualize the data as a scatter plot42 43def visualize_input_data():44    fig = plt.figure(1, facecolor="w", figsize=(5, 5))45    scatter = plt.scatter(X[:, 0], X[:, 1], c=y, s=20, edgecolor="k")46    handles, labels = scatter.legend_elements()47    plt.axis("square")48    plt.legend(handles=handles, labels=["outliers", "inliers"], title="true class")49    plt.title("Gaussian inliers with \nuniformly distributed outliers")50    return fig51 52 53 54 55 56title = " An example using IsolationForest for anomaly detection."57description1 = "The isolation forest is an Ensemble of Isolation trees and it isolates the data points using recursive random partitioning."58description2 = "In case of outliers the number of splits required is greater than those required for inliers."59description3 = "We will use the toy dataset as given in the scikit-learn page for Isolation Forest."60 61with gr.Blocks(title=title) as demo:62    63    gr.Markdown(f" # {title}")64    gr.Markdown(65    """66    The isolation forest is an ensemble of isolation trees and it isolates the data points using recursive random partitioning.67    In case of outliers, the number of splits required is greater than those required for inliers.68    We will use the toy dataset for our educational demo as given in the scikit-learn page for Isolation Forest.69    70    """)71 72    gr.Markdown("You can see the associated scikit-learn example [here](https://scikit-learn.org/stable/auto_examples/ensemble/plot_isolation_forest.html#sphx-glr-auto-examples-ensemble-plot-isolation-forest-py).")73    74    loaded_model = load_hf_model_hub()75 76    with gr.Tab("Visualize Input dataset"):77        btn = gr.Button(value="Visualize input dataset")78        with gr.Row():79        80            btn.click(visualize_input_data, outputs= gr.Plot(label='Visualizing input dataset') )81            # out = gr.Textbox(label="explaination of the loss function")82            gr.Markdown(83            """84            ## Data Generation85            We generate 2 clusters one spherical and the other slightly deformed, from Standard Normal distribution86            For the sake of consistency inliers are assigned a ground label of 1 and outliers are assigned a label -1.87            The plot is a visualization of the clusters of the input dataset.88 89            """)90 91    with gr.Tab("Plot Decision Boundary"):92      # btn_decision = gr.Button(value="Plot decision boundary")93      # btn_decision.click(plot_decision_boundary, outputs= gr.Plot(label='Plot decision boundary') )94      with gr.Row():95        image_decision = gr.Image('./downloaded-model/decision_boundary.png')96        gr.Markdown(97        """98        ## Plot the Discrete Decision Boundary99        We plot the discrete decision boundary. 100        The background colour represents whether a sample in that given area is predicted to be an outlier or not. 101        The scatter plot displays the true labels102 103        """)104    105    with gr.Tab("Plot Path"):106      with gr.Row():107        image_path = gr.Image('./downloaded-model/plot_path.png')108        gr.Markdown(109        """110        ## Plot the path length of the decision boundary111        By setting the `response_method="decision_function"`, the background of the `DecisionBoundaryDisplay` represents 112        the measure of the normality of an observation. 113 114        Normality of Observation = Path Length/Number of Forests of Random Trees115 116        117        The RHS of the above equation is given by the number of splits required to isolate a given sample118        Such score is given by the path length averaged over a forest of random trees, which itself is given by the depth of 119        the leaf (or equivalently the number of splits) required to isolate a given sample.120 121        When a forest of random trees collectively produces short path lengths for isolating some particular samples, 122        they are more likely to have anomalies, and the measure of normality is close to 0. 123        Similarly, large paths correspond to values close to 1 and are more likely to be inliers.124 125        """)126 127demo.launch()