[Talks@IDSIA] 3 Aug, 12:15 Forecasting prices in dynamic heterogeneous product markets using multivariate prediction methods

3 Aug, 12.15 Room 222, DSAN SPEAKER: G. Lucchese, University of Bergamo TITLE: Forecasting prices in dynamic heterogeneous product markets using multivariate prediction methods Accurate forecasting of product prices is essential for competitive bidding in consumer markets. Price developments of different products often move in parallel possibly due to shared components. Real time pricing puts a tremendous strain on efficient decision making. Research questions -How forecast prices in dynamic markets for heterogeneous products that share common components? -Can co-dependencies among product prices be used to efficiently forecast product prices? -Can adequate multivariate forecasts be obtained, which satisfy conditions of precision and available time for decision making? It is proposed a dynamic model, which makes use of Kalman filter and EM to estimate the parameters; its results are compared to alternative univariate and multivariate autoregressive methodologies.