Modeling engine for continuous intelligence

There are many exciting use cases for machine learning where time is not part of the modeling challenge: image recognition, movie recommendation, sentiment analysis…

Datapred's modeling engine is designed for challenges where capturing the information contained in the sequence of events is critical (in addition to the information conveyed by the events themselves).

It streamlines everything that is specific about machine learning for time series, at every stage of the machine learning pipeline — preprocessing, modeling and post-processing.

Datapred gives you complete control and endless possibilities at every modeling step:

• Integrate and/or aggregate any predictive model, from moving averages to neural networks.
• Choose one of our built-in aggregation and stacking procedures, or implement your own.
• Easily combine machine learning, human expertise, physical models, management rules and operations research.
• Adapt your batch Keras, Scikit-learn, TensorFlow or PyTorch algorithms to the specificities of time series.
• Define and optimize your own cost function - however complex.

Datapred streamlines the implementation of explainability best practices:

• It is model agnostic, and makes it easy to combine linear, autoregressive and machine learning models.
• It dynamically quantifies the contribution of every model you are using to overall performance.
• It facilitates the application of model-specific explainability techniques to streaming data.
• It dynamically assesses the predictive power of every feature.

Datapred accelerates the transition from good machine learning model and a production-ready machine learning application:

• It packages over 30 DevOps open-source components to help you kick-start, deploy and maintain your applications, without having to install and configure everything by yourself.
• It uses a graph structure that facilitates collaborative development, parallel vs distributed computations, continuous improvement and component recycling for your application.
• It includes built-in connectors to standard databases, connector templates for non-standard databases, and is natively Linux/Docker-compatible.