국내 폐자동차 발생 예측을 위한 모형의 선택 방법

Abstract

The number of End-of-Life Vehicle(ELV) increases with the development of automobile industry and leads to environmental pollution. In the European Union, the bill was amended to increase the ELV recycling rate of the existing 85% to 95% by 2015. Republic of Korea is also planning to raise the ELV recycling rate up to 95% legally. To improve the ELV recycling rate, studies on the efficient organization and operation of the ELV dismantling system is in progress. The generation of ELV is an important factor influencing the capacity and size of the ELV dismantling system. The aim of this paper is to present an efficient methodology to predict the amount of ELV. The Bayesian variable selection method presented in this paper consists of two steps. In the first step, by applying for Occam's Window Algorithm and selecting the models consisting of the main factors in the whole set of predictive models. In the second step, using the Bayes' Theorem to find a high posterior probability model from the selected models in step one. The performance of the model selection methodology proposed in this paper is verified by analyzing its performance compared with the forecast methodology presented in previous studies with actual data. Therefore, methodology presented in this paper is expected to reduce the complexity of the problem when analyzing the time-series data in the big data environment.