Prediction Intervals for Mechanical Property Forecasting with Improved ANFIS

Abstract

Application of Adaptive-Network-based Fuzzy Inference System (ANFIS) for point forecast has burgeoned in recent years. While the performance of ANFIS drops when the uncertainty level increases in dataset, and point forecasts also suffer from unreliable and uninformative problems. Prediction interval (PI) has been proposed as a powerful tool to address these drawbacks. In spite of plentiful application of ANFIS for constructing PIs, the use in mechanical property forecasting is limited. Compared with traditional computational expensive algorithms, in this paper, we combine Particle Swarm optimization (PSO) and Genetic Algorithm (GA) to minimize the cost function and optimize the parameters of ANFIS. The proposed novel approach is applied to a real mechanical property forecasting case study. The outcomes show modified PSO-based ANFIS not only computational less expensive, but also with good quality of constructing better PIs.