Classification Based Reliability Growth Prediction on Data Generated by Multiple Independent Processes

Abstract

Reliability Growth is a modeling process for product quality characterization over the lifespan for both hardware and software products and has been explained by multiple models like Duane, Crow-AMSAA, Lloyd Lipow etc. Our research proposes a framework for case-based/scenario based model estimation and prediction, by supervised learning of historical data. In this proposed framework, the case base is generated from historical data and Crow Model is applied in a novel sense to extract information from the historically labeled occurrences. With our framework, we draw in a comparative advantage over the traditional predictive modeling using a Crow’s Growth Model.