Abstract During the voyage, predicting fuel consumption of ships under different sea-states and weather conditions has been a challenging and far-reaching topic, because there are a great number of feature variables affecting the fuel consumption, including main-engine status, cargo weight, ship draft, sea-states and weather conditions, etc. Data driven statistical models have been employed to model the relationship between fuel consumption rate and these voyage parameters. However, some of the feature variables are highly correlated, e.g. wind speed and wave height, air pressure and wind force, cargo weight and draft etc., thus a typical multiple collinearity problem arises so that the fuel consumption cannot be accurately calculated by using the traditional multiple linear regression. In this study, the LASSO (Least Absolute Shrinkage and Selection Operator) regression algorithm is employed to implement the variable selection for these feature variables, additionally, it guides the trained predictor towards a generalizable solution, thereby improving the interpretability and accuracy of the model. On the basis of the LASSO, a novel ship fuel consumption prediction model is proposed. Experimentally, the superiority of the proposed method was confirmed by comparing it with some existing methods on predicting the fuel consumption. The proposed method is a promising development that improves the calculation of the fuel consumption.