Ocean Current Prediction Based on Machine Learning for Deciding Handover Priority in Underwater Wireless Sensor Networks

In Underwater Wireless Sensor Network(UWSN), there are nodes moving by themselves like AUVs(Autonomous Underwater Vehicles), or nodes moving passively by currents and other environmental affects. For the passive nodes, normally there exists difficulties in estimating the nodes' locations. Herein, handover technologies for the nodes are necessary but the researches for handover technologies in UWSN are barely being processed. Because of the environmental characteristics, it is difficult to realize underwater handover. This paper presents the model which predicts the ocean current in specific timelines through the practical data learned in various machine-learning methodologies at the southeastern sea of the Korean peninsula. By assessing the performance, the appropriate model would be selected for predicting current directions. Assume that unfixed underwater sensor nodes are moved by currents, and handover priorities will be decided by the predicted current directions. The result of deciding handover priorities suggested with dividing sectors by stations, DTC(Decision Tree Classifier) had the highest prediction performance which is Rank 1 accuracy 55.79% and Rank 3 accuracy 85.60%.