A case study: The prediction of Taiwan’s export of polyester fiber using small-data-set learning methods

摘要

During the past four decades, the textile industry has been the industry earning the largest amount of foreign exchange in Taiwan. Notably, polyester fibers are one of the most outstanding industries in Taiwan on the global economic stage. The productivity of the polyester fiber industry in Taiwan has remained steady since the 1980s, and the midstream and downstream industry is also tied in with this development. However, starting from 2000, Taiwan’s export of polyester fibers has changed dramatically owing to the rapid economic rise of China. Since this sudden change occurred only 5 years ago, it is hard for researchers to predict the amount of future exports accurately using the trends of historical data over the past 20 years. This research adopts the methodology GIKDE (General Intervalized Kernel Density Estimator), which is a newly developed method for small-data-set prediction, to predict the amount of future exports and expects to obtain a more accurate estimation to use as a reference to help managers make plans for products, capacity and markets.