Improving forecast accuracy by a segmented rate of change in technology forecasting using data envelopment analysis (TFDEA)

Abstract

Technology forecasting using data envelopment analysis (TFDEA) captures technological advancement from the evolution of the state-of-the-art (SOA) frontier. Within this process, TFDEA combines rates of changes (RoC) from past technologies that have been superseded by superior technologies. However, it was occasionally observed in previous applications that forecasting based on a single aggregated RoC did not consider the unique growth patterns of each technology segment, which resulted in a conservative or aggressive forecasting. This study proposes a procedure to improve the forecasting accuracy by identifying local rates of change for each frontier segment that may represent different product families. This approach is applied to six previously published applications using a rolling origin hold-out sample tests to validate its performance compared to the traditional TFDEA approach. The results indicate that the segmented rate of change approach determines different rates of change for product niches that result in more accurate forecasts