ImputeGAP Tutorial - KDD’25¶
In this tutorial, we will provide an engaging hands-on tutorial where you will learn time series imputation using the powerful Python library, ImputeGAP. The tutorial is divided into two parts. In the first part, we will dive into building an end-to-end imputation workflow with the library, where you will explore real-world missingness patterns simulation, leverage automated tuning for optimal imputation, and benchmark imputation techniques—all with extensive customization options. In the second part, we will unlock advanced functionalities, including assessing the impact of imputation on downstream analytics and understanding how time series features influence imputation outcomes. The ImputeGAP library is accessible at: https://imputegap.readthedocs.io.
Learning Outcomes¶
The Interactive part of the tutorial will pro vide participants with experience in applying time series imputation techniques to real-world datasets and missingness scenarios. They will also know how to (1) deploy ImputeGAP to build a full imputation pipeline for time series with various customization options, (2) create a common test-bed for comparing imputation algorithms, (3) assess the effects of data imputation on downstream applications, and (4) provide insights into the imputation behavior. We expect that the attendees will gain a deep understanding of time series imputation techniques and their underlying mechanisms, including their theoretical foundations, practical implementation, and real-world implications.
Target Audience and Prerequisites¶
Our tutorial is intended for both beginners and intermediate-level time series practitioners, such as data scientists and software engineers, who frequently engage in data cleaning tasks. It also serves time series researchers, especially those who focus on improving data quality for machine learning applications.
Engaging Experience¶
Attendees will benefit from in-depth demonstrations and carefully designed step-by-step hands-on materials. Given that most attendees are expected to have access to their personal computers, they will have the opportunity to explore and experiment with ImputeGAP firsthand during the tutorial.