[Paper Review] Spatio-Temporal Data Mining: A Survey of Problems and Methods

Gowtham Atluri, Anuj Karpatne+1 more|arXiv (Cornell University)|Nov 13, 2017

Data Management and AlgorithmsComputer Science245 references86 citations

TL;DR

This survey maps spatio-temporal data mining (STDM) problems and methods, classifying literature into six categories and detailing ST data types, instances, and challenges.

ABSTRACT

Large volumes of spatio-temporal data are increasingly collected and studied in diverse domains including, climate science, social sciences, neuroscience, epidemiology, transportation, mobile health, and Earth sciences. Spatio-temporal data differs from relational data for which computational approaches are developed in the data mining community for multiple decades, in that both spatial and temporal attributes are available in addition to the actual measurements/attributes. The presence of these attributes introduces additional challenges that needs to be dealt with. Approaches for mining spatio-temporal data have been studied for over a decade in the data mining community. In this article we present a broad survey of this relatively young field of spatio-temporal data mining. We discuss different types of spatio-temporal data and the relevant data mining questions that arise in the context of analyzing each of these datasets. Based on the nature of the data mining problem studied, we classify literature on spatio-temporal data mining into six major categories: clustering, predictive learning, change detection, frequent pattern mining, anomaly detection, and relationship mining. We discuss the various forms of spatio-temporal data mining problems in each of these categories.

Motivation & Objective

• Define the landscape of spatio-temporal data mining (STDM) and motivate its study across domains.
• Provide a comprehensive taxonomy of ST data types, instances, and similarity measures.
• Survey STDM approaches organized around core problem categories (clustering, predictive learning, change detection, frequent pattern mining, anomaly detection, relationship mining).
• Highlight how spatial and temporal dependencies (auto-correlation) and heterogeneity affect STDM methods and evaluation.

Proposed method

• Review and synthesize literature on STDM from diverse domains (climate science, neuroscience, epidemiology, transportation, Earth sciences).
• Classify STDM literature into six major problem categories based on the nature of the data mining task.
• Describe data types (event, trajectory, point reference, raster) and data instance definitions (points, trajectories, time series, spatial maps, ST rasters).
• Explain challenges posed by auto-correlation and non-stationarity in ST data for modeling and evaluation.
• Discuss data conversion among ST data types to leverage different STDM tools.
• Present a framework linking data types, instance definitions, similarity measures, and STDM tasks.

Experimental results

Research questions

• RQ1 What are the fundamental data types and representations in spatio-temporal data that drive STDM formulations?
• RQ2 How do ST properties like auto-correlation and heterogeneity influence modeling choices and evaluation?
• RQ3 What are the main STDM problem categories, and how can literature be organized to reveal commonalities and differences?
• RQ4 How can ST data be converted across types to enable cross-domain application of STDM methods?

Key findings

• ST data mining for spatio-temporal data can be meaningfully categorized into six problem areas: clustering, predictive learning, change detection, frequent pattern mining, anomaly detection, and relationship mining.
• There is a rich taxonomy of ST data types (events, trajectories, point references, rasters) and multiple ways to define data instances, affecting problem formulations.
• Auto-correlation and heterogeneity in space and time challenge traditional data mining assumptions and evaluation methods, necessitating specialized STDM approaches.
• ST data enable novel problem formulations by coupling spatial and temporal information, enabling cross-domain transfer of methods and insights.
• A unified framework and survey can bridge classical data mining with STDM, guiding researchers to select appropriate formulations for their data.