Filter and Refine Approach
Spatiotemporal co-occurrence patterns (STCOPs) represent the subsets of event types that occur together in both space and time. However, the discovery of STCOPs in data sets with extended spatial representations that evolve over time is computationally expensive because of the necessity to calculate interest measures to assess the co-occurrence strength, and the number of candidates for STCOPs growing exponentially with the number of spatiotemporal event types. In this work, we introduce a novel and effective filter and-refine algorithm to efficiently find prevalent STCOPs in massive spatiotemporal data repositories with polygon shapes that move and evolve over time. We provide theoretical analysis of our approach, and follow this investigation with a practical evaluation of our algorithm effectiveness on three real-life data sets and one artificial data set.
Pattern Growth-Based Approach
Spatiotemporal co-occurrence pattern (STCOP) mining refers to discovering the subsets of event types whose instances frequently co-locate in a spatial context and coincide in a temporal context. STCOP mining is the spatiotemporal extension to Frequent Itemset Mining (FIM). Unlike the classical FIM approaches, which are applied on transactional databases, STCOP mining is applied on the spatiotemporal datasets comprised of event instances which are represented by evolving region trajectories. Previous STCOP mining algorithms are Apriori-based, where the number of candidate patterns can grow exponentially with the number of event types. In this work, we present a pattern growth-based approach for mining STCOPs, which allows us to discover STCOPs without computationally expensive candidate generation processes. We experimented our algorithm with four real-life solar event datasets and compared its performance with the earlier Apriori-based approach
Mining spatiotemporal co-occurrence patterns requires assessing the strength of co-occurrences among the instances of different feature types. Currently, a spatiotemporal version of the Jaccard measure is used for measuring the strength of spatiotemporal co-occurrences. We present an extended spatiotemporal version of the Jaccard measure (J*) that is more relevant and efficient for the task of STCOP mining. We also demonstrate the space and time efficiency of the J* with experimental evaluation.
STCOP in Non-relational Databases
In the era of big spatiotemporal data, it is unavoidable to use distributed non-relational databases for storing spatiotemporal event instances. Current relational database management systems offer rich functionality for geometric operations (e.g. PostGIS); but, for mining massive spatiotemporal datasets, newly popular non-relational database systems can perform much better. In this project, we designed data models for storing spatiotemporal event instances, and implemented a spatiotemporal join procedures for mining spatiotemporal co-occurrence patterns. Our system uses Accumulo database and a Java client for implementation of mining algorithms.
Visit the project page: https://grid.cs.gsu.edu/~baydin2/proj/nonrelstcop.html
ST index for STCOP
We have developed a framework for mining spatiotemporal co-occurrence patterns. Two well-known trajectory-based indexing tecniques are successfully utilized and the mining process, now, can be done even in real-time for solar data.
The raster data collected from Solar Dynamic Observatory is huge. 17 feature finding teams and their modules generate vector data. Using this vector data, we perform data-driven analysis of solar phenomena. However, previously used database settings were not satisfactory and feasible for large-scale data analysis. In this project, we have indexed solar vector data for mining solar co-occurrences, in order to increase the efficiency.
Indexing techniques, we have utilized for the STCOP-mining framework are Scalable and Efficient Trajectory Index (SETI) and Chebyshev Polynomial Index (CPI). The framework mimics a simple database environment using binary files. The source code for this project can be found on BitBucket. The framework is implemented in C++, using Boost libraries.
Visit the project page: https://grid.cs.gsu.edu/~baydin2/proj/sdst.html