NEXT LOCATION PREDICTION VIA DEEP LEARNING SQUEEZE AND EXCITATION BIDIRECTIONAL GATED RECURRENT UNIT

Next location prediction has recently attracted much attention from researchers due to its application in various domains. Many variables usually affect moving objects, including time, distance, and user configuration. This makes it difficult to predict where moving items will go when these factors are considered. This research proposes a deep learning-based next-location prediction network (DL-NLocP) to increase the accuracy of next-location prediction. Initially, the datasets are pre-processed to enhance the data quality and employ term frequency-inverse document frequency (TF-IDF) with glove word embedding approaches to convert the textual data into real-valued vectors. Afterward, multi-head CNN extracts the vector data's temporal, location, and user behavior features. Finally, squeeze and excitation with the BiGRU network are developed to predict the following location in each trajectory with contextual information. The proposed DL-NLPN model was tested on the Ningbo AIS and Geolife dataset, and experimental results supported the model's validity. The proposed model consistently outperforms current state-of-the-art approaches by 93.75 % for Geolife and 94.75 % for Ningbo AIS on average accuracy@20. The results show that the proposed approach performs better in Next location prediction than the existing methods.