Towards customer evaluation based product performance modeling

Abstract

In the front-end customer driven design process, a number of design alternatives need to be evaluated based on customer expectations. A systematic method is presented in this paper to rank design alternatives by formulating a future rating model. Projections to Latent Structures (PLS), which is a multivariate analysis technique with proven efficiency, is employed to model experts' ratings in terms of product attributes for each time step, and screen unimportant attributes using Variable Importance (VIP) Scores. The PLS model parameter time series is used to formulate the future rating model, by means of an optimization algorithm embedded with forecasting. A case study is conducted using ratings of 16 major mobile phone technological forums collected from 2006 first quarter to 2009 third quarter. The future model predictions of 2009 fourth quarter, with Mean Absolute Percentage Error (MAPE) of 1.6, show the potential of proposed concept screening method. ©2010 IEEE.