Pre-training Spatial-Temporal Trajectories from Time-Aware Location Embeddings PROJECT TITLE : Pre-training Time-Aware Location Embeddings from Spatial-Temporal Trajectories ABSTRACT: In recent years, location-based Data Mining has received a significant amount of attention due to the growing amount of spatial-temporal trajectory data that has been accumulated. Learning the embedding vectors of locations through self-supervised pre-training is a fundamental research topic in this field of study. Embedding vectors that have been pre-trained can make use of the abundantly available unlabeled trajectory data, which is beneficial to subsequent tasks in a number of different ways. However, the vast majority of the currently available techniques disregard the temporal information that is concealed within the visited times of locations in trajectories. It is necessary to combine temporal information with location embedding vectors given that human activities are highly regulated by particular times of the day. This is due to the fact that certain periods of the day have a significant impact on how people behave in different locations. In this paper, we propose a Time-Aware Location Embedding (TALE) pre-training method that is based on the CBOW framework. This method is able to incorporate temporal information into the learned embedding vectors of locations and is able to do so by using the CBOW framework. During the process of calculating Hierarchical Softmax, a brand new temporal tree structure was developed with the purpose of extracting temporal information. We apply the learned embedding vectors into three downstream location-based prediction tasks, namely location classification, location visitor flow prediction, and user next location prediction, in order to validate the efficacy of TALE. These tasks are: location classification; location visitor flow prediction; and user next location prediction. Experiments are run on four datasets based on real-world user trajectories, and the results of these experiments demonstrate that our TALE model can obviously assist downstream tasks in gaining better performance. Did you like this research project? To get this research project Guidelines, Training and Code... Click Here facebook twitter google+ linkedin stumble pinterest Concept of Semi-Supervision Learning through Concept Space and Concept-Cognitive Learning Using a Bayesian Model for Social Networks to Predict Hot Events in the Early Period