|
Developement of neuro-fuzzy predictors for new product market forecasts.
Prediction of the demand for a new product is based on a number of parameters such as pre-launch, test marketing, feedback from distributors, attributes of the product, demand for previous products with subsets of attributes that are similar, as well as feedback from the market.
Metacomp was used to develop nested neural models that used ~ 40 input parameters from the pre-launch period, plus market feedback postlaunch. It was found that very significant improvments were obtained for the period after launch, both on a nation wide basis and on the demand experienced regionally by the distribution network.
With time after launch, the improvments compared to the regression models previously in the use diminished, mainly because all models become quite accurate will sufficient market feedback.
|