Cloud-Based IoT Solution for Predictive Modeling of Ship Fuel Consumption

Abstract

The need for savings in ship fuel consumption has led to the proliferation of various cloud-based service-oriented approach towards predicting and optimizing ship operation. However, majority of the cloud-based services are generally designed for general purpose prediction where ship owners do not have the liberty to select and customize machine learning algorithms and parameters that they desire to experiment with for their specific datasets. In this paper, the feasibility of a novel Do-It-Yourself (DIY) approach towards performing predictive modeling and analytics of ship fuel consumption based on out-of-the-box cloud-based Azure Machine Learning (ML) Studio tool sets is demonstrated. The POC system implementing multiple regression model (MLR) model may provide insight into ship operational fuel consumption based on historical operational IoT data collected from ships operated under various operational parameters. The derived predictive model is validated with coefficient of determination, R2 for goodness of fit. The coefficient of determination, R2 result at 0.9707 indicates the good fitness of regression.