ml_edm package - a Python toolkit for Machine Learning based Early Decision Making

Aug 4, 2026

Upload by Oliver

ml_edm is a Python 3 library built for early decision making in learning tasks that involve temporal and sequential data. The toolkit is modular, enabling researchers to implement custom triggering strategies for classification, regression, or other machine learning tasks. It provides efficient implementations of many state-of-the-art Early Classification of Time Series (ECTS) algorithms using parallel computation. The API follows scikit-learn conventions, ensuring estimator and pipeline compatibility, and the software is released under the BSD-3-Clause license.

LanguageEnglish

CategoryTechnology

FormatPDF

Pages5

Reading Time13 min

AccessFree to download

1 Introduction

2 Implementation details

3 An API for Early Time Series Classification

3.1 Cost setting

3.2 Classification

arXiv :2408 . 12925v1 [ cs .LG] 23 Aug 2024 ml_edm package: a Python toolkit for Machine Learning based Early Decision Making Aurélien Renault Youssef Achenchabe Édouard Bertrand Alexis Bondu Antoine Cornuéjols Vincent Lemaire Asma Dachraoui [aurelien.renault@orange.com](aurelien.renault@orange.com)[ ](aurelien.renault@orange.com)youssef.achenchabe@ [edouard.bertrand100@gmail.com](edouard.bertrand100@gmail.com)[ ](edouard.bertrand100@gmail.com)[alexis.bondu@orange.com](alexis.bondu@orange.com) [antoine.cornuejols@agroparistech.fr](antoine.cornuejols@agroparistech.fr)[ ](antoine.cornuejols@agroparistech.fr)[vincent.lemaire@orange.com](vincent.lemaire@orange.com)[ ](vincent.lemaire@orange.com)asma.dachraoui@ Editor: my editor Abstract ml_edm is a Python 3 library, designed for early decision making of any learning tasks involving temporal/sequential data. The package is also modular, providing researchers an easy way to implement their own triggering strategy for classi􀀜cation, regression or any machine learning task. As of now, many Early Classi􀀜cation of Time Series (ECTS) state-of-the-art algorithms, are e􀀞ciently implemented in the library leveraging parallel computation. The syntax follows the one introduce in scikit-learn, making estimatorsand pipelines compatible with ml_edm. This software is distributed over the BSD-3-Clause license, source code can be found at [https://github.com/ML-EDM/ml_edm](https://github.com/ML-EDM/ml_edm). Keywords: Machine Learning, Python, Time Series, Sequential data, Cost-sensitive learning 1 Introduction In hospital emergency rooms (Mathukia et al. (2015)), in the control rooms of national or international power grids (Dachraoui et al. (2015)), in government councils assessing critical situations, there is a time pressure to make early decisions. On the one hand, the longer a decision is delayed, the lower the risk of making the wrong decision, as knowledge of the problem increases with time. On the other hand, late decisions are generally more costly, if only because early decisions allow one to be better prepared. For example, a cyber-attack that is not detected quickly enough gives hackers time to exploit the security 􀀝aw found. A number of applications involve making decisions that optimizes a trade-o􀀛 between accuracy of the prediction and its earliness. The problem is that favoring one usually works against the other. Greater accuracy comes at the price of waiting for more data. Such a compromise between the Earliness and the Accuracy of decisions has been particularly studied in the 􀀜eld of Early Classi􀀜cation of Time Series (ECTS) (Gupta et al. (2020)), and introduced by Xing et al. (2008) . In this paper, the ml_edm package is presented, it gathers many state-of-the-art ECTS algorithms ina modular way, clearly di􀀛erencing the classi􀀜cation part from the triggering part, when possible. Thus, it allows researchers to easily reproduce past results, test well-know methods in di􀀛erent settings and implement new ECTS algorithms. 2 Implementation details Dependencies: The ml_edm package depends on numpy (Van Der Walt et al. (2011)), scipy (Virtanen et al.(2020)) and pandas (Wes McKinney (2010)) for classic array operations. It also depends on scikit-learn for its API, utilities as well as some classical Machine Learning models. The package is also dependent of aeon (Middlehurst et al. (2024)), for some time series specialized features extraction algorithms. 3 An API for Early Time Series Classi􀀜cation One of the main 􀀜eld of research when it comes to Early Decision Making is ECTS. The package has been primarily built to reproduce results from this literature, working only with univariate time series for now (Renault et al. (2024)) . In what follows, the di􀀛erent modules are presented through an introductory example. Please note that, as Early Decision Making can be broaden to other ML tasks, the ml_edm package can also be used to address these. 3.

FAQs

What is the ml_edm package designed to do?

It is a Python 3 library for early decision making in learning tasks involving temporal or sequential data, particularly Early Classification of Time Series (ECTS).

How does ml_edm support customization for triggering strategies?

The package is modular and allows researchers to implement their own triggering strategy for classification, regression, or other machine learning tasks.

What API compatibility does ml_edm provide?

The syntax follows scikit-learn, making its estimators and pipelines compatible with scikit-learn-style workflows.

Aug 4, 2026

Upload by Oliver