Uncertainty-aware visual analytics for exploring human behaviors from heterogeneous spatial temporal data

Abstract

When analyzing human behaviors, we need to construct the human behaviors from multiple sources of data, e.g. trajectory data, transaction data, identity data, etc. The problems we’re facing are the data conflicts, different resolution, missing and conflicting data, which together lead to the uncertainty in the spatial temporal data. Such uncertainty in data leads to difficulties and even failure in the visual analytics task for analyzing people behavior, pattern and outliers. However, traditional automatic methods can not solve the problems in such complex scenario, where the uncertain and conflicting patterns are not well-defined. To solve the problems, we proposed a semi-automatic approach, for users to solve the conflicts and identify the uncertainties. To be general, we summarized five types of uncertainties and solutions to conduct the tasks of behavior analysis. Combined with the uncertainty-aware methods, we proposed a visual analytics system to analyze human behaviors, detect patterns and find outliers. Case studies from the IEEE VAST Challenge 2014 dataset confirm the effectiveness of our approach.

Introduction

Data recording human behavior becomes more and more in volume and diversity. With the development of the techniques, the GPS can record people’s position and movement, the transaction system in the bank can record people’s purchase and billing behavior, and more social media data would reflect people’s attitude towards public affairs, or even eating preference. Facing the heterogeneous data, we can adopt visual analytics to understand the people behavior, find patterns and detect outlier events.

Directly using heterogeneous data analyzing process could lead to difficulties and even failure for the visual analytics tasks. This is because the data are often heterogeneous and imperfect. There can be various uncertainties in the data, including errors, data missing and conflicts. The data can also be in different resolutions. However, traditional automatic methods can’t solve the problems in such complex scenario, where the uncertain and conflicting patterns are not well-defined. Our approach combines both algorithmic methods together with interactions in visualization, to enable users to identify, mark and refine such uncertainty issues. Together with such uncertainty-aware methods, we proposed a visual analytics system for supporting human’s spatial temporal behavior analysis from the heterogeneous data.

In this paper, we report different kinds of uncertainties that we identified in a visual spatial data analysis and demonstrate how we refine them with the semi-automatic methods. Generally, our methods are data-driven reliability improvement methods. For different types of data, we have proposed different solutions and adopt cross referent of multiple sources of data. As there are heterogeneous data sharing the same attributes, but with different granularity, we can get finer resolution data with uncertainty from other types of data. With these approaches, we can better understand people’s behavior in different dimensions and mark the reliability for further analysis. Uncertainty identification and analysis vary much and are challenging to solve through pure computation methods. So in our work, combined with visual identification and automatic preprocess methods, our methods have users in the visual analytics loop. Thus, users can better explore the different reliability of the data and further analyze outlier events.

Throughout this work, we use the fictitious datasets from IEEE VAST Challenge 2014 Mini Challenge 2 [1]. Combined with the uncertainty-aware approach, our proposed visual analytics system is able to summarize the general movement patterns of a group of people, and help analysts detect abnormal events, with various visualization view and multiple filters. In summary, our contribution is as follows.

• Semi-automatic Uncertainty Refinement Methods. We summarized five general types of uncertainty and proposed novel solutions for each. To solve the ill-defined uncertainty problems, we combine users’ capability and algorithmic methods and allow human in the analysis loop.

• Uncertainty-aware Visual Analytics System. We have developed a comprehensive visual analytics system, incorporating the uncertainty-aware approaches and multiple coordinated visualization views, thus providing a full solution for understanding the human behaviors and detect interesting patterns and outliers.

This paper is structured as follows. Section 2 reviews related work. After introducing the data in Section 3, we present the uncertainty summary and general description of solutions in Section 4. We present the details of uncertainty-aware approach in Section 5. We present the visual analytics procedure and technical details in Sections 6 and 7. We demonstrate the use of our tools in four case studies. Finally, we discuss the limitations, future work, and conclusion.