Forecasting in Multi Agent Dynamic Supply Chains: A Hedonics Switching Regimes Model Tested on TAC SCM


Speaker


Abstract

Almost any stage of supply chain management is based on a forecast model. The latter is created depending on the kind of network of the supply chain. In our case we are interested in multi agent supply chains where the actors are the manufacturers-assemblers of a range of end-products sharing several  components, like in the case of personal computers and mobile phones. Hence, our model is designed taking into account the correlations between the hedonic price for a component and the product price. It is an alternative of VAR models depending on the historical data. We name it Multivariate Hedonic Switching Regime model (MHSR). For its characteristics MHSR is easily adaptable in Inventory Control to forecast demand depending on sub-item design, but also in other fields of SCM. In our paper we include regime considerations respect on our previous one system model. We retrace the steps of a variant of Kim's algorithm to estimate regimes offline. We show test results of MHSR model in TAC SCM. It provides us with data about product prices and their historical behaviors, the input of our algorithm.
 
We introduce the first results in regime detection for MHSR model showing cluster techniques and optimization methods in multivariate parameter case. This talk is overall an opportunity to identify weak and strong points of the model together with auditors. Furthermore, it may be an opportunity for future researches about application of MHSR model in the field of Inventory Control and Transportation Logistics Management.
 
Contact information:
dr. Wolf Ketter
Email