Location prediction: a deep spatiotemporal learning from external sensors data

Modeling Trajectories Obtained from External Sensors for Location Prediction via NLP Approaches

“…The EST prediction model investigated, which was also based on the previous papers [ 14 , 15 , 16 ] is a recurrent based model. The first layer (1) is an Embedding Layer, which applies a linear transformation on the high-dimensional input vectors to reduce their dimensionality while trying to preserve the similarity between instances from the original space of features in the new feature space.…”

“…Sensors may fail to capture the license plate and then produce incomplete trajectories with missing observations [ 14 , 15 ]. Furthermore, sensors are spatially sparse and not equally distributed, producing sparse trajectories.…”

“…We present some definitions to support our problem statement in the same way as in our previous work [ 14 , 16 ].…”

“…To deal with the problems of incompleteness and scarcity, different data imputation methods are used and compared. The paper [ 14 ] presents a multitask neural network model based on space and time to tackle the same problem addressed by [ 15 ]. Next-location prediction in an external sensor network is also addressed in [ 34 ], which introduces a deep learning model for the historical trajectories obtained by traffic surveillance devices, deployed along with a city’s street network.…”

“…State-of-the-art papers include the adoption of neural embeddings for model locations, points of interest (PoIs) [ 10 , 11 , 12 ] or to learn the temporal interactions between users and items [ 13 ]. The papers [ 14 , 15 , 16 ] also focus on the next location prediction task from trajectories obtained by external sensors and assume that a road network constrains object movements. Their trajectories report predefined positions (i.e., sensors’ location) on the road network, which makes trajectories much more sparse than the usual GPS raw trajectories.…”