PROJECT TITLE :

Deep Learning for Spatio-Temporal Data Mining: A Survey

ABSTRACT:

The availability of spatio-temporal data has increased significantly in recent years as a result of the rapid development of a variety of positioning methods, including the Global Positioning System (GPS), mobile devices, and remote sensing. The extraction of useful information from spatio-temporal data is of paramount significance for a wide range of real-world applications, including the comprehension of human mobility, the development of intelligent transportation, urban planning, public safety, health care, and the administration of environmental management. Traditional data mining methods, particularly statistically-based methods for dealing with such data, are becoming increasingly ineffective as the number of spatio-temporal data records, their volume, and their resolution continue to rapidly increase. Deep learning models, such as recurrent neural networks (RNN) and convolutional neural networks (CNN), have recently achieved remarkable success in many different domains due to their powerful ability in automatic feature representation learning. These models are also widely applied in various spatio-temporal data mining (STDM) tasks, such as predictive learning, anomaly detection, and classification. In this paper, we provide a comprehensive review of recent progress that has been made in applying techniques of deep learning to STDM. After briefly discussing the deep learning models that are frequently utilized in STDM, we begin by classifying the spatio-temporal data into the five distinct types that are most commonly used. Next, we will categorize the existing literature according to the different types of spatio-temporal data, the data mining tasks, and the deep learning models. After that, we will discuss the applications of deep learning for STDM in a variety of different fields, such as transportation, on-demand services, climate and weather analysis, human mobility, location-based social networks, crime analysis, and neuroscience. Finally, we draw a conclusion about the shortcomings of the previously conducted research and discuss potential future directions for research.