[Paper Review] Visual Analytics of Movement Pattern Based on Time-Spatial Data: A Neural Net Approach

Zhenghao Chen, Jianlong Zhou+1 more|arXiv (Cornell University)|Jul 9, 2017

Data Visualization and Analytics8 references3 citations

TL;DR

This paper proposes a neural network-based visual analytics framework for time-spatial movement data, integrating a back-end database, Self-Organizing Map (SOM) for clustering, and Recurrent Neural Network (RNN) for prediction. The system enables interactive visualization through 2D heatmaps, 3D time-spatial cubes, and animations, achieving high accuracy in predicting mobility patterns and detecting abnormal traffic behavior in real-world applications such as commercial promotion and wildlife impact assessment.

ABSTRACT

Time-Spatial data plays a crucial role for different fields such as traffic management. These data can be collected via devices such as surveillance sensors or tracking systems. However, how to efficiently an- alyze and visualize these data to capture essential embedded pattern information is becoming a big challenge today. Classic visualization ap- proaches focus on revealing 2D and 3D spatial information and modeling statistical test. Those methods would easily fail when data become mas- sive. Recent attempts concern on how to simply cluster data and perform prediction with time-oriented information. However, those approaches could still be further enhanced as they also have limitations for han- dling massive clusters and labels. In this paper, we propose a visualiza- tion methodology for mobility data using artificial neural net techniques. This method aggregates three main parts that are Back-end Data Model, Neural Net Algorithm including clustering method Self-Organizing Map (SOM) and prediction approach Recurrent Neural Net (RNN) for ex- tracting the features and lastly a solid front-end that displays the results to users with an interactive system. SOM is able to cluster the visiting patterns and detect the abnormal pattern. RNN can perform the predic- tion for time series analysis using its dynamic architecture. Furthermore, an interactive system will enable user to interpret the result with graph- ics, animation and 3D model for a close-loop feedback. This method can be particularly applied in two tasks that Commercial-based Promotion and abnormal traffic patterns detection.

Motivation & Objective

• To address the challenge of analyzing massive time-spatial mobility data for pattern discovery and prediction.
• To improve upon traditional visualization and clustering methods by integrating deep learning techniques for scalable and accurate feature extraction.
• To develop an interactive front-end system that enables users to explore and interpret complex movement patterns through graphics, animations, and 3D models.
• To apply the framework to real-world problems, including commercial user promotion and wildlife impact detection in nature preserves.

Proposed method

• The framework uses a hybrid back-end combining relational and dynamic non-relational databases to manage time-spatial data and generate temporal databases on demand.
• Self-Organizing Map (SOM) is employed for clustering movement patterns, identifying regular and abnormal visitor behaviors based on spatial and temporal features.
• Recurrent Neural Network (RNN), specifically Long Short-Term Memory (LSTM), is used for time-series prediction of future locations of moving objects.
• The front-end system visualizes results using 2D heatmaps, 3D time-spatial cubes, relational maps, and animations to support user interpretation and feedback.
• The pipeline integrates data ingestion, feature extraction (clustering and prediction), and interactive visualization into a closed-loop analytics system.
• The system was validated on two real-world datasets: mobile user activity in China and vehicle trajectories in a nature preserve (VAST 2017).

Experimental results

Research questions

• RQ1 How can neural network-based methods improve the clustering and prediction of large-scale time-spatial movement data compared to traditional statistical and visualization techniques?
• RQ2 What role does interactive visualization play in enhancing user understanding of complex mobility patterns and abnormal behaviors?
• RQ3 Can RNN-based models outperform classical machine learning models in predicting future movement patterns in time-spatial data?
• RQ4 How can clustering via SOM detect abnormal traffic patterns that may indicate high-impact events, such as prolonged vehicle presence in sensitive ecological zones?

Key findings

• RNN achieved the highest prediction accuracy of 92.3% on the VAST 2017 nature preserve dataset, outperforming SVM (87.6%), KNN (84.2%), and Random Forest (79%).
• The SOM clustering method successfully identified four distinct and abnormal vehicle patterns, including one vehicle that remained in the preserve for 92 days, indicating prolonged ecological impact.
• The interactive 3D time-spatial cube and heatmaps clearly visualized the trajectories and check-in frequencies of abnormal clusters, enabling rapid detection of high-impact vehicles.
• RNN predictions closely matched actual traffic patterns from 19:00–20:00 pm, correctly outlining four major movement routes starting from General Gate 6.
• The system enabled actionable insights for commercial promotion, such as identifying high-traffic mobile stations in advance to guide targeted advertising.
• The framework demonstrated strong performance on both commercial and ecological applications, validating its versatility and scalability for real-world time-spatial analytics.