Delphi Anomaly Detection Library

A comprehensive library for detecting anomalies in business data using statistical and machine learning approaches. Designed for fraud detection, system monitoring, data validation, and quality control.

For complete documentation, examples, and API reference, visit: danieleteti.it/delphianomalydetection

Features

• 6 Detection Algorithms - Three Sigma, Sliding Window, EMA, Adaptive, Isolation Forest, DBSCAN
• Production Ready - Thread-safe, memory-efficient, thoroughly tested
• Real-time & Batch - Streaming data and historical analysis
• Easy Integration - Single unit, no external dependencies
• Performance Monitoring - Built-in metrics and benchmarking
• Hyperparameter Tuning - Grid/Random search with cross-validation

Algorithms

Algorithm	Best For	Speed
Three Sigma	Quality control, baseline analysis	Fast
Sliding Window	Real-time monitoring, dashboards	Medium
EMA	Financial data, trending patterns	Very Fast
Adaptive	Evolving patterns, learning systems	Fast
Isolation Forest	Fraud detection, multi-dimensional	Medium
DBSCAN/LOF	Cluster-based anomalies	Slow

Quick Start

uses
  AnomalyDetection.Types,
  AnomalyDetection.ThreeSigma,
  AnomalyDetection.Factory;

var
  Detector: IAnomalyDetector;
  Result: TAnomalyResult;
begin
  // Create detector
  Detector := Factory.CreateDetector(adtThreeSigma, 'QualityControl');

  // Add data and build
  Detector.AddValues([100, 105, 98, 102, 107, 99, 103, 101]);
  Detector.Build;

  // Detect anomaly
  Result := Detector.Detect(150);
  if Result.IsAnomaly then
    WriteLn('Anomaly detected! Z-score: ', Result.ZScore:0:2);
end;

Real-time Monitoring

uses AnomalyDetection.SlidingWindow;

var
  Detector: TSlidingWindowDetector;
begin
  Detector := TSlidingWindowDetector.Create(100);  // 100-value window
  try
    while HasIncomingData do
    begin
      Value := GetNextReading;
      if Detector.IsAnomaly(Value) then
        TriggerAlert('Spike detected: ' + Value.ToString);
      Detector.AddValue(Value);
    end;
  finally
    Detector.Free;
  end;
end;

Fraud Detection (Multi-dimensional)

uses AnomalyDetection.IsolationForest;

var
  Detector: TIsolationForestDetector;
begin
  Detector := TIsolationForestDetector.Create(100, 256, 5);  // 100 trees, 5 features
  try
    // Train on normal transactions
    for Transaction in NormalTransactions do
      Detector.AddTrainingData([Amount, Hour, Day, Category, Age]);
    Detector.Train;

    // Detect fraud
    Result := Detector.DetectMultiDimensional([5000, 3, 2, 5, 35]);
    if Result.IsAnomaly then
      FlagForReview('Suspicious transaction');
  finally
    Detector.Free;
  end;
end;

Factory Pattern

uses AnomalyDetection.Factory;

// Pre-configured detectors for common use cases
WebDetector := Factory.CreateForWebTrafficMonitoring;
FinanceDetector := Factory.CreateForFinancialData;
IoTDetector := Factory.CreateForIoTSensors;
FraudDetector := Factory.CreateForHighDimensionalData;

Documentation

Full documentation with all algorithms, examples, and tuning guide:

danieleteti.it/delphianomalydetection

Installation

1. Download or clone from GitHub
2. Add the src folder to your Library Path
3. Add the appropriate units to your project

Requirements

• Delphi 10.3 Rio or later
• No external dependencies

License

Apache 2.0 - Free for commercial and personal use.

Professional Support

• Training & Consulting: Bit Time Professionals
• Email: professionals@bittime.it

Support

• Documentation: danieleteti.it/delphianomalydetection
• Issues: GitHub Issues
• Community: Facebook Group