Dead or Alive: Continuous Data Profiling for Interactive Data Science

Dakuo Wang Erick Oduor Justin D. Weisz + 2 lainnya

13 Januari 2021

Data science (DS) projects often follow a lifecycle that consists of laborious tasks for data scientists and domain experts (e.g., data exploration, model training, etc.). Only till recently, machine learning(ML) researchers have developed promising automation techniques to aid data workers in these tasks. This paper introduces AutoDS, an automated machine learning (AutoML) system that aims to leverage the latest ML automation techniques to support data science projects. Data workers only need to upload their dataset, then the system can automatically suggest ML configurations, preprocess data, select algorithm, and train the model. These suggestions are presented to the user via a web-based graphical user interface and a notebook-based programming user interface. Our goal is to offer a systematic investigation of user interaction and perceptions of using an AutoDS system in solving a data science task. We studied AutoDS with 30 professional data scientists, where one group used AutoDS, and the other did not, to complete a data science project. As expected, AutoDS improves productivity; Yet surprisingly, we find that the models produced by the AutoDS group have higher quality and less errors, but lower human confidence scores. We reflect on the findings by presenting design implications for incorporating automation techniques into human work in the data science lifecycle.

Zezhong Wang Lovisa Sundin Dave Murray-Rust + 1 lainnya

18 Januari 2020

This paper introduces the concept of 'cheat sheets' for data visualization techniques, a set of concise graphical explanations and textual annotations inspired by infographics, data comics, and cheat sheets in other domains. Cheat sheets aim to address the increasing need for accessible material that supports a wide audience in understanding data visualization techniques, their use, their fallacies and so forth. We have carried out an iterative design process with practitioners, teachers and students of data science and visualization, resulting six types of cheat sheet (anatomy, construction, visual patterns, pitfalls, false-friends and well-known relatives) for six types of visualization, and formats for presentation. We assess these with a qualitative user study using 11 participants that demonstrates the readability and usefulness of our cheat sheets.

J. Freire J. Comba K. Gaither + 2 lainnya

1 Maret 2021

Interactive visualizations are at the core of the exploratory data analysis process, enabling users to directly manipulate and gain insights from data. In this article, we present three different ways in which interactive visualizations can be included in Jupyter Notebooks: 1) matplotlib callbacks; 2) visualization toolkits; and 3) embedding HTML visualizations. We hope that this article will help developers to select the best tools to build their interactive charts in Jupyter Notebooks.

J. Horcas J. Galindo David Benavides

12 September 2022

Data visualization aims to effectively communicate quantitative information by understanding which techniques and displays work better for different circumstances and why. There are a variety of software solutions capable of generating a multitude of different visualizations of the same dataset. However, data visualization exposes a large space of visual configurations depending on the type of data to be visualized, the different displays (e.g., scatter plots, line graphs, pie charts), the visual components to encode the data (e.g., lines, dots, bars), or the specific visual attributes of those components (e.g., color, shape, size, length). Researchers and developers are not usually aware about best practices in data visualization, and they are required to learn about both the design practices that make communication effective and the low level details of the specific software tool used to generate the visualization. This paper proposes a software product line approach to model and materialize the variability of the visualization design process, guided by feature models. We encode the visualization knowledge regarding the best design practices, resolve the variability following a step-wise configuration approach, and then evaluate our proposal for a specific software visualization tool. Our solution helps researchers and developers communicate their quantitative results effectively by assisting them in the selection and generation of the visualizations that work best for each case. We open a new window of research where data visualization and variability meet each other.

Michael J. Muller Dakuo Wang Amy X. Zhang

18 Januari 2020

Today, the prominence of data science within organizations has given rise to teams of data science workers collaborating on extracting insights from data, as opposed to individual data scientists working alone. However, we still lack a deep understanding of how data science workers collaborate in practice. In this work, we conducted an online survey with 183 participants who work in various aspects of data science. We focused on their reported interactions with each other (e.g., managers with engineers) and with different tools (e.g., Jupyter Notebook). We found that data science teams are extremely collaborative and work with a variety of stakeholders and tools during the six common steps of a data science workflow (e.g., clean data and train model). We also found that the collaborative practices workers employ, such as documentation, vary according to the kinds of tools they use. Based on these findings, we discuss design implications for supporting data science team collaborations and future research directions.