AutoML

RECAP AutoML is a library for the automated optimisation of machine learning (ML) models. It has been designed for a target audience with limited; technical knowledge, previous experience, or knowledge of statistics or mathematics but who would like to develop their own ML models. RECAP AutoML currently provides support for the optimisation of three common classes of models: ARIMA-type time series models; decision trees (as a baseline for more complex models such as random forests and boosted trees); and multilayer perceptron neural networks.

The main task of the RECAP AutoML tool is to relieve the inexperienced ML user from the burden of feature selection, model selection, as well as hyperparameter optimisation. RECAP AutoML provides such optimisation ‘out-of-the-box’ so that the user does not need the background knowledge required to select the model’s parameters or to select feasible models. This substantially reduces the learning curve for the supported classes of models. In
addition, optimal models found by the tool can be served as a web service that can be queried using a REST interface. The latter makes it easier to integrate the models with more complex environments. RECAP AutoML is integrated with the Python library, scikit-learn.

Benefits: RECAP AutoML facilitates the adoption of machine learning techniques by SMEs and institutions that do not have sufficient in-house knowledge in data science, and/or the capacity to hire data science professionals. The tool enables the extraction of higher value from collected data without a large initial investment in specialised human resources and computing infrastructure. The initial coverage of algorithms provided by RECAP AutoML includes the most popular choices for time series forecasting and classification tasks.

Licence: Open Source

Download: https://github.com/rleiva/fastautoml

Contact: IMDEA Networks Institute, Spain | Rafael García Leiva | [email protected]