Massive Online Analysis
| Company type | Open-source software project |
|---|---|
| Industry | Data mining, Machine learning software |
| Headquarters | Hamilton , New Zealand |
| Products | Software for data stream mining and concept drift handling |
| Owner | University of Waikato |
| Website | moa |
Massive Online Analysis (MOA) is an open-source, cost-free software initiative dedicated to data stream mining with an emphasis on handling concept drift. Created in Java at the University of Waikato, New Zealand, MOA delivers a robust platform for designing and executing machine learning experiments on continuously evolving data streams through a graphical user interface, command-line tools, or Java API.
Description
MOA encompasses a broad array of machine learning methods specifically designed to manage extensive data streams and adapt to real-time concept drift, featuring:
- Classification
- Bayesian methods: Naive Bayes, Multinomial Naive Bayes
- Decision tree models: Decision Stump, Hoeffding Tree, Hoeffding Option Tree, Hoeffding Adaptive Tree
- Ensemble classifiers: Bagging, Boosting, ADWIN-based Bagging, Adaptive-Size Hoeffding Trees Bagging, Perceptron Stacking, Leveraging Bagging, Online Accuracy Updated Ensemble
- Functional classifiers: Perceptron, Stochastic Gradient Descent (SGD), Pegasos
- Drift detection classifiers: Self-Adjusting Memory, Probabilistic Adaptive Windowing
- Multi-label and active learning classification methods
- Regression
- FIMTDD
- AMRules
- Clustering
- StreamKM++, CluStream, ClusTree, D-Stream, CobWeb
- Outlier detection
- STORM, Abstract-C, COD, MCOD, AnyOut
- Recommendation systems
- BRISMFPredictor
- Frequent pattern mining
- Mining of itemsets and graphs
- Change detection techniques
These tools facilitate scalable machine learning operations on large-scale data streams subject to evolving concepts.
MOA supports two-way integration with Weka and is distributed under the terms of the GNU General Public License.
See also
- ADAMS Workflow: A workflow engine compatible with MOA and Weka
- Streams: A modular platform for crafting and running data stream experiments
- Vowpal Wabbit
- List of numerical analysis software