Automobile warranty forecasting is made challenging by ‘maturing data’ (also called warranty growth) phenomena that causes warranty performance at specific months-in-service (MIS) values to change with time. In this paper a wavelet transform based technique is proposed for forecasting the warranty performance in presence of the ‘maturing data’ phenomena. The wavelet transform helps to cut the warranty performance data into different frequency components, and then allow study of each component with a resolution matched to its scale. In other words it allows the use of long time intervals (or longer window widths) for low-frequency information and shorter regions (or shorter window width) for higher-frequency information. The shorter the window width, the better is the resolution. The window widths of the wavelet transform can be adjusted automatically, and provide better time resolution for high frequency components and better frequency resolution for low frequency components of the warranty performance pattern. The proposed approach first decomposes the warranty performance patterns into several scales at different levels of resolution using wavelet transform analysis. The decomposition involves an approximate part from which higher frequency information has been filtered and detail parts from which lower frequency information have been filtered. The chosen approximate and detail parts are extended using neural networks. Subsequently original and the forecasted signals are reconstructed to obtain forecast values for the warranty performance. A real-world application example is used to illustrate the use of the proposed methodology.

This content is only available via PDF.
You do not currently have access to this content.