ENHANCEMENT OF THE PREDICTION ACCURACY OF GREY SYSTEM MODEL USING A PARTICLE SWARM OPTIMIZED INITIAL CONDITION

Abstract

The grey system model has seen vast application in many fields due to its good accuracy in predicting systems with a limited dataset. In this paper, the prediction accuracy of the traditional grey system model is enhanced using the particle swarm optimization algorithm. The enhancement is mainly the use of PSO to predict an optimum initial condition value based on the input dataset to improve the prediction accuracy of the original grey model that uses the first data of the input dataset as its initial condition. The performance of the enhanced model was tested against the traditional grey model and another model that seeks to enhance the initial condition using the average-minimum-maximum absolute error and the mean absolute percentage error to prove its adaptability. Sample monotonic increasing and decreasing datasets, value of lost load and value of lost load per GDP datasets were used as testing data to prove the accuracy of the proposed model. The proposed model predicted optimum initial condition values of 18.9241, 5.9160, 5.0203 and 3120012789 that resulted in the lowest MAPE OF 0.1798%, 0.1799%, 2.1359% and 11.2813% for the monotonic increasing and decreasing, value of lost load and value of lost load per GDP datasets respectively. It was shown that the proposed model outperforms the traditional grey system model and an improved initial condition model in literature.

Keywords

• particle
• optimization
• initial condition
• prediction
• performance