ml_anomaly_detector 0.1.0

A lightweight and extensible Dart library for anomaly detection in univariate and multivariate data. Includes Z-Score and LOF algorithms.

📊 ml_anomaly_detector

A simple and extensible Dart package for anomaly detection using classical algorithms like Z-Score, with support for pluggable strategies and detailed result reporting.

---

🚀 Features

• ✅ Z-Score based anomaly detection
• 📦 Designed for extensibility (LOF, IQR, DBSCAN coming soon)
• 🔍 Returns detailed results: value, score, confidence, explanation
• 📊 Compatible with time-series or single-point anomaly detection
• 💡 Easy to use, production-ready

---

🧠 Example

import 'package:ml_anomaly_detector/ml_anomaly_detector.dart';

void main() {
  final data = [1.0, 1.1, 0.9, 1.2, 10.0];
  final detector = ZScoreDetector(threshold: 2.0);
  final results = detector.detect(data);

  for (var result in results) {
    print(
        'Value: \${result.value}, Score: \${result.score.toStringAsFixed(2)}, '
        'Anomaly: \${result.isAnomaly}');
  }
}

---

📦 Installation

Add the following line to your pubspec.yaml:

dependencies:
  ml_anomaly_detector: ^1.0.0

Then run:

dart pub get

---

📁 API Overview

ZScoreDetector

ZScoreDetector({
  double threshold = 3.0,
  String algorithmName = 'z_score',
});

AnomalyResult

Field	Type	Description
value	double	Original input value
score	double	Anomaly score (Z-score, etc.)
isAnomaly	bool	Whether it's an anomaly
index	int?	Optional index in dataset
confidence	double?	Confidence level (0.0 to 1.0)
algorithm	String?	Name of algorithm used
explanation	String?	Human-readable decision explanation

---

🧪 Tests

Run unit tests using:

dart test

---

📌 Roadmap

• ✅ Z-Score detector
• ❌ IQR-based anomaly detection
• ❌ Local Outlier Factor (LOF)
• ❌ DBSCAN
• ❌ Visual anomaly plotting (with Flutter)

---

📃 License

MIT © 2025 Mehmet Çelik

Contributions are welcome! PRs & feedback appreciated.

---