yzhao062/pyod

Features

• Anomaly Detection - Provides a comprehensive collection of algorithms to identify outliers and assign binary labels in datasets.
• Agentic Planning - Provides an AI-driven orchestration layer that translates natural language into structured detection plans.
• Anomaly Degree Scoring - Provides raw numerical values for each data point to quantify the degree of outlierness.
• Graph Anomaly Detectors - Identifies structural and attribute outliers in graph-structured data using connectivity analysis.
• Graph-Based Outlier Detection - Identifies outliers in datasets by analyzing data structures through a graph-based approach.
• Multi-Modal Detection - Provides the ability to identify outliers across various data formats including unstructured text and images.
• Unstructured Data Analysis - Chains foundation model encoders with outlier detectors to identify anomalies in text and images.
• Unified Interfaces - Standardizes input and output signatures across numerous anomaly detection algorithms for consistent model swapping.
• Multi-Modal AI Toolkits - Offers a toolkit for detecting anomalies across diverse formats including unstructured text, images, and graph data.
• Outlier Detection - Implements deep learning architectures to distinguish between inliers and outliers.
• Time Series Detection - Identifies outliers within sequential data by analyzing observation windows to determine if a point is anomalous.
• MLOps Pipeline Automation - Translates natural-language requests into automated detection pipelines using AI agents.
• Agentic Workflow Automation - Provides a framework that uses AI agents to translate natural language into automated detection pipelines.
• Anomaly Detection Metrics - Compares predicted outlier labels and scores against ground-truth data to quantify the performance of a detection model in the project.
• Anomaly Scoring - Calculates raw anomaly scores and probability estimates for data samples using fitted models.
• Ensembles - Combines multiple outlier detection models into an ensemble to improve processing speed and prediction accuracy.
• Histogram-Based Outlier Detection - Identifies anomalies in datasets using histogram-based outlier detection to assign labels and scores.
• Influence-Based Outlier Detection - Identifies anomalous data points by calculating the influence of each observation on a regression model.
• Kernel Principal Component Analysis - Uses Kernel PCA to project data into higher-dimensional spaces to identify non-linear patterns and outliers.
• Median Absolute Deviation - Detects anomalies by calculating the Median Absolute Deviation to determine distance from the median.
• Minimum Covariance Determinant - Utilizes the Minimum Covariance Determinant algorithm to produce binary labels and anomaly scores.
• One-Class SVM - Isolates outliers by fitting a boundary around the majority of training data using One-Class SVM.
• Sampling-Based Detection - Analyzes the distance or density of samples to distinguish outliers from normal observations.
• Transductive Graph - Identifies outliers in graph data using transductive learning and decomposition.
• Dual-Autoencoder Graph Detectors - Identifies anomalies in graph data by comparing structural connectivity and node attributes through symmetric reconstruction.
• Cluster-Based Outlier Detection - Identifies outliers using a cluster-based local outlier factor approach to calculate deviation scores.
• HDBSCAN Density-Based Detection - Identifies anomalous data points using density-based clustering to separate inliers from outliers.
• Proximity and Density Analysis - Separates outliers from inliers using proximity-based and density-based detection models.
• Bagging Ensembles - Uses a feature bagging meta-framework to combine multiple detection models for improved robustness.
• Feature Bagging Outlier Detection - Identifies anomalous data points by training an ensemble of detectors on random subsets of features.
• Ensemble Execution Acceleration - Combines multiple outlier detectors into a single model to increase processing speed through parallel execution.
• Ensemble Result Aggregation - Combines multiple anomaly models into a single ensemble to improve efficiency through parallel result aggregation.
• Contamination Threshold Optimization - Determines optimal contamination levels for outlier scores using data-driven thresholding approaches.
• Contamination Threshold Optimizers - Includes logic to determine optimal contamination levels by analyzing the statistical distribution of raw anomaly scores.
• Prediction Confidence Assessment - Assesses model certainty for each single-sample prediction to reject highly uncertain results.
• Feature Importance Attribution - Calculates the contribution of each input feature to the detection process to identify which variables influence results in the project.
• Gaussian Mixture Model Detection - Identifies anomalous data points by fitting a Gaussian Mixture Model and calculating sample likelihood.
• Generative Adversarial Network Detection - Identifies anomalous data points by training a GAN to learn normal data distributions.
• Outlier Detection - Identifies anomalous data points by training gradient boosting models on feature sets.
• k-Nearest Neighbors Outlier Detection - Identifies anomalous data points by calculating the distance to the nearest neighbors.
• Outlier Detection - Implements kernel density estimation to flag data points in low-density regions as anomalies.
• Outlier Labeling - Assigns binary labels to distinguish between normal observations and outliers based on learned thresholds.
• Performance Metrics - Computes performance metrics by comparing predicted outlier labels and scores against known ground truth data in the project.
• Model Performance Evaluators - Compares predicted outlier scores against ground truth labels to quantify the accuracy of the detection model in the project.
• Model Prediction Evaluation - Quantifies detection accuracy by comparing predicted outlier scores against ground truth labels.
• Multi-modal Embedding Generation - Transforms diverse data formats into a common representation for unified outlier scoring.
• Performance Evaluation Tools - Provides tools for calculating accuracy metrics and visualizing the distribution of detected anomalies.
• Performance Metrics - Calculates performance metrics by comparing predicted outlier scores against known ground truth labels in the project.
• Accuracy Calculators - Compares predicted outlier scores against ground truth labels to quantify the accuracy of a detection algorithm in the project.
• Foundation-Model Encoder Chains - Pairs pretrained deep learning encoders with outlier detectors to extract features from unstructured text and image data.
• Spectral Residual Analysis - Identifies outliers in sequential data using spectral residual analysis to find abnormal patterns.
• Severity Scoring - Quantifies the degree of anomaly for data points by calculating raw decision scores.
• Structural Graph Anomaly Detection - Identifies outliers in graph-structured data by analyzing structural connectivity and node attributes using dual autoencoders.
• Variational Autoencoders - Implements anomaly detection using Variational Autoencoder architectures to identify outliers via reconstruction error.
• Connectivity-Based Outlier Analysis - Identifies anomalous data points by analyzing the local density and connectivity of points.
• Distance-Based Outlier Detection - Identifies anomalous data points by calculating distances between neighbors to detect deviations.
• PCA-Based Outlier Detection - Identifies anomalous data points by analyzing principal components and calculating deviation scores.
• Quasi-Monte Carlo Discrepancy Analysis - Identifies anomalous data points by calculating the Quasi-Monte Carlo Discrepancy to determine distribution deviation.
• Parallel Processing Pipelines - Combines multiple detection models into a parallel execution pipeline to improve processing throughput.
• Statistical Distribution Visualizers - Plots the distribution of inliers and outliers in a dataset to provide a visual representation of the model in the project.
• General Machine Learning - Toolkit for detecting outliers in multivariate data.
• Data Quality and Validation - Outlier and anomaly detection.
• Data Validation and Anomaly Detection - Toolbox for scalable outlier and anomaly detection.

Star history