oddA

oddA is a modular open-source framework designed for time-series analysis and anomaly detection. It provides tools for data ingestion, preprocessing, model construction, evaluation, and deployment, enabling practitioners to build scalable anomaly-detection workflows across heterogeneous data sources.

oddA originated from the OddA Consortium, formed by researchers from universities and industry in 2022 to address fragmentation in time-series analytics. The first public release appeared in 2023, with ongoing development and community contributions through 2025.

The framework uses a dual-channel architecture with odd and even processing streams that run in parallel and can be fused for robust detection. It includes data connectors for CSV, JSON, SQL databases, and streaming platforms such as Kafka; preprocessing steps for missing data, normalization, and windowing; and a model zoo that allows mixing conventional statistical models with neural architectures.

Supported algorithms include robust regression, isolation forests, autoencoders, and Bayesian change-point detectors. Evaluation supports cross-validation, backtesting, and multiple precision-recall metrics. It can export models to standard formats and deploy to cloud or edge environments.

OddA is used in industrial monitoring, energy management, and financial services for detecting anomalies, faults, and fraud, especially when data arrive irregularly or at varying rates.

The project has an active user community, tutorials, and contributor guidelines. Critics note a learning curve and resource demands for complex models, though modularity allows scaling.

See also: anomaly detection; time-series analysis; data science software.