A fuzzy predictive model with confidence interval estimation for alloy property assessment

It is very important to provide reliable assessment and prediction of mechanical properties for alloy steel products. To build reliable models for alloy steel property prediction, this paper presents a fuzzy predictive model with its own confidence interval estimation. The proposed confidence interval calculation not only involves the accuracy of the trained model with the given input-output data, but also incorporates the influence of the training data distribution. The focus is on the calculation of confidence bounds to provide an indication of the reliability of the resulting model predictions. Simulation results show that the proposed confidence interval estimation correctly reflects the availability of the training data in the region where the prediction is being made, and it has been effectively applied to high-dimensional sparse data set of alloy steel property prediction.