本研究主要目的在於利用遺傳演算法(Genetic Algorithms,GAs)來發展模糊預測模式以提昇預測的準確度,遺傳演算法是一種以目標為導向的平行搜尋技術,主要應用許多最佳化問題上以尋找全域或近似全域之極值,在本研究中我以銷售預測為例,提出了一個動態的預測技術模式,其主要運用遺傳演算法以加強在建構模糊時間序列的銷售量預測過程中,用來搜尋最佳之語意變數個數與分割區間的多寡,以及找出最適合各不同模糊時間序列之模式推論基底數值w。另外本研究也提出了融合專家意見之領先指標於模糊時間序列之預測值,結果發現其預測準確度顯著高,證明本研究所提之模糊預測模式之有效性。 The major purpose of this study is applying Genetic Algorithms(GAs) to developing fuzzy forecasting in order to increase the accuracy of forecasting. Genetic algorithm is a parallel goal-oriented search technique for optimization and can be used to easily find out the global or nearly global optima for optimization problems. In this study, we focus on sales forecasting and propose a dynamic forecasting modeL By using Genetic Algorithms in searching the optimal linguistic variables and partition intervals,and finding out the most fitness model basis w of fuzzy time series in different cases. Finally, we propose adding the expert opinions served as leading indicators in the fuzzy time series for forecasting value. Results show that the accuracy of the forecasting results is significantly improved,it proved the effectiveness of the fuzzy forecasting model we proposed.
Relation:
第六屆國際資訊管理學術研討會論文集=Proceedings of the 6th International Conference on Information Management,頁205-212