查詢結果分析
相關文獻
- 類神經網路之預測--以臺灣地區積體電路產量為例
- 類神經網路之預測與信賴區間之建構--以臺北市失業率為例
- Time Series Forecasting of Kaohsiung Unemployment Rate Using Neural Network Model
- A Prediction-Based Recurrent Neural Fuzzy Network for Temporal Sequence Recognition
- 門市管理銷售預測模式建構
- 結合獨立成份分析與類神經網路於財務時間序列預測模式之建構
- 支援向量機與支援向量迴歸於財務時間序列預測之應用
- 時間序列法與類神經網路於股票漲跌預測比較之研究
- 人工智慧應用於室內空氣汙染物濃度預測之研究
- 臺北市失業率之預測--模糊時間序列之應用
頁籤選單縮合
題名 | 類神經網路之預測--以臺灣地區積體電路產量為例=Forecasting Using Neural Networks--A Case Study of the Production Volume of Integrated Circuits in Taiwan |
---|---|
作者 | 李天行; 邱志洲; 徐步鯤; Lee, Tian-shyug; Chiu, Chih-chou; Hsu, Smiler; |
期刊 | 統計與資訊評論 |
出版日期 | 19980900 |
卷期 | 4 1998.09[民87.09] |
頁次 | 頁47-69 |
分類號 | 448.552 |
語文 | chi |
關鍵詞 | 類神經網路; 積體電路; 時間序列; 多元迴歸; 預測; ARIMA; Integrated circuits; Time series; Multiple regression; Forecasting; Neural network; |
中文摘要 | 鑑於全球積體電路產業之蓬勃發展,國內積體電路廠商也投入大筆資金,積極興建晶圓廠,以求能進一步分食積體電路市場的大餅。然而,廠商間一味競賽式的投資設廠,帶來的可能是價格的崩盤,甚至投資難以回收、進而破產的地步。因此,考慮積體電路的需求面以預測未來產業所需之產能,能提供重要的訊息而有利於廠商做出正確的投資決策。本研究嚐試利用積體電路下游20種應用產品自1993年4月到1997年11月之歷史資料,以多元迴歸模式、ARIMA模式與類神經網路模擬進行積體電路產量預測之實証結果顯示,類神經網路模式顯著有較好的預測能力。 |
英文摘要 | The semi-conductor industry has been experiencing the double-digit growth rate during the past two decades. And hence billions of dollars have been invested n the industry due to the forecasts of continuously high demand. However, starting in 1996, the supply of the integrated circuits has far more exceeded the demand due to the capacities can not be absorbed by the market. Several leading companies expected record-low rate of returns and some manufacturing facilities even face the problem of being shut down. The semi-conductor industry has encountered the most serious recession which they believe will still last for a while. Valuable information can be obtained through the analysis of the demand pattern and this will definitely contribute to the success of making the correct decision. The purpose of this research is to forecast the production volume of the integrated circuits in Taiwan through the analysis of those precuts using integrated circuits. Multiple regression analysis. ARIMA model, and neural networks were used to build up the forecasting model. The result shows that neural networks has significantly better forecasting results in terms of MSE, normalized MSE, MAD, and MAPE. |
本系統之摘要資訊系依該期刊論文摘要之資訊為主。