Application of ARIMA to Forecast Spatio-Temporal Ship Density

Abstract

With the increased marine traffic in the recent decade, it has become essential to monitor the amount of traffic in the territorial sea because of maritime accidents and overfishing. The state-of-art technologies such as marine telecommunication and satellite observation have been drastically interested in researchers and have enhanced maritime surveillance power. Despite that, the forecasting of ship density is required to arrange the human resources of the coast guard efficiently. Therefore, this paper introduces Auto-Regressive Integrated Moving Average (ARIMA), an analysis model for time-series data, to forecast ship density. Prior to the application of the model, V-Pass data is preprocessed, and then the number of fishing ships within 12 hours (day/night each) and within specific spatial grids is counted. ARIMA model has been generated after checking the stationarity of the data for every spatial grid. The proposed method is revealed that forecast ship density within reasonable error bounds.