Mobile Menu

First end-to-end autoML pipeline for time-series data

The van der Schaar lab has announced the release of their alpha version of clairvoyance, an autoML complete package for time series analysis. The product comes as a result of years of R&D and real-world testing, to aid clinical research and support decision making.

The lab, based in the University of Cambridge, specialise in developing ML and AI theory and methods primarily for improving healthcare and medical understanding. In a blog post, the team say that their newly-released pipeline is the first of its kind in producing personalised and interpretable decisions and recommendations using time-series data.

Time-series data is an essential element of evidence-based clinical decision making, providing more nuanced insights than static data. As electronic health records become more accessible for research purposes, there is real opportunity to fully leverage time-series data for predictive modelling and drawing accurate insights.

However, the complexity of these data sets has presented researchers with challenges in developing ML approaches, with many failing to account for the intricacies and inter-dependencies that exist in the real world. Existing ML methods for time-series analysis, have therefore, been unable to demonstrate their effectiveness in supporting clinical decision making.

Clairvoyance, however, offers “a systematic and automated approach to personalized dynamic predictions, personalized information acquisition, personalized monitoring, and personalized treatment plans, while also offering interpretations.”

Although developed for clinical contexts, it also has applicability in non-medical domains, say the team, “thanks to its ability to facilitate complex inference workflows in a transparent, reproducible, and efficient manner.”

The package addresses three key problems: the engineering problem: with a single, consistent interface the software toolkit simplifies collaboration and code-sharing; the evaluation problem: the complete benchmarking environment provides realistic and systematic context for evaluating; and the efficiency problem: where past models have been resource-heavy is overcome by the use of automated ML.

Lab head, Professor Mihaela van der Schaar, concluded on the utility of clairvoyance: “I have no doubt that clairvoyance will prove useful in driving clinical decision-making research, and I also believe it can offer a lot of benefits to the machine learning community.”

The alpha version of clairvoyance is open access, for researchers to download and test, with the release of a beta version expected in the near future.

The summary blog post can be found here and the full publication here

Image Source:

Share this article