Enhanced Accuracy in Mining Spatiotemporal Sequential Patterns Using Mapreduce

The aim is to improve spatiotemporal sequential pattern mining by employing the STS-miner within a MapReduce system for better efficiency and accuracy. It aims to prove better scalability and accuracy when dealing with big data. There were two key elements utilized in the research. Group 1 employed the GSP algorithm on a large-scale spatiotemporal dataset from sensors for traffic, social media, and sensor networks. Group 2 utilized STS-Miner with pattern extraction emphasis on data preprocessing and discretization. The computing system was a 5 -node cluster (1 master +4 workers) having Intel Xeon 2.6 GHz processors, 64 GB memory, and 2 TB SSD storage, using MapReduce for parallel processing for mining spatiotemporal patterns. STS-Miner achieved $\mathbf{9 0. 2 \%}$ accuracy, outperforming GSP with $\mathbf{7 9. 8 \%}$. It also showed better precision (89.5 %) and recall (92.7 %) than GSP on F1-score. The model cut down execution time by 27.4 %, which worked much better on large-scale spatiotemporal data. STS-Miner successfully extracted spatial hotspots and temporal trends, and therefore was sufficiently applicable to traffic monitoring and epidemic surveillance. STS-Miner improves spatiotemporal sequence mining by cutting down execution time and improving accuracy. It efficiently handles enormous data sets and is best suited for applications like city mobility, and upcoming research focused on privacy and adaptive learning.