應用統計迴歸與類神經網路建立早期公開與核准專利的關聯性模型

合適的技術預測可輔助經營者掌握現代技術的最新市場趨勢，在產品與製程研發相關方面亦能輔助擬定最佳決策。在專利制度中，早期公開專利經常比核准專利更早呈現技術的特性與涵義，企業與產品管理層亦廣用類神經網路與統計迴歸方法建立模組和預測系統。此研究目的首要為建立早期公開與核准專利之間的關聯性，決定公開至核准的時間差選擇方法並依此設為預測用週期。採用美國專利為預測資料，進行多點自變數預測工作找出最佳預測時間間隔點，並以早期公開專利數預測核准專利數；應用統計迴歸與倒類神經演算(BPNN)兩種方法針對MRAM、OLED、LCD三種技術進性預測工作，結果均可獲得合理的模組與架構。然而統計迴歸與類神經網路的相異特性，為類神經網路預測工作較不受輸入自變項數量影響，在預測工作上所需耗費的工作時間較長，但因仍是尚未成熟的技術，架構是否合理的檢定僅能用經驗判斷，在本研究中則以MAPE〈誤差百分比〉作為架構組成的合理性判定依據；統計迴歸方式採用過多自變項將造成無限多解的結果，除了預測準確度產生偏差及實際工作上產生不必要的時間與金錢浪費，但預測工作時間較快，而且擁有成熟的結果檢定方式，如此研究是以T、F與P值進行結果檢定。

關鍵字

專利預測；技術預測；倒傳遞類神經；多項式迴歸；迴歸分析

並列摘要

A suitable technology forecasting method can help managers grasp the latest trends of market in specific technologies and make the best decisions on product and process developing policies. The number of published patent applications on specific technologies can reflect the significance of those technologies before they are granted. There are considerable amount of researches that use the methods of neural network and statistical regression on modeling and forecasting systems in the domains of business and production management. The objective of this research is to first study the relationships between published patent applications and patent grants, and determine the publish-to-grant time lag. A second step was then conducted to choose the most appropriate modeling time span based on the best fitting of data acquired from USPTO. Finally the models established can be used to estimate the future numbers of patent grants and carry out technology forecasting. Two methods, statistical regression and neural network, were used in the implementation of the proposed methodologies, and three case studies were conducted for presentation. On BPNN, it used MAPE as the effective or ineffective determination of modeling. On Statistic regression analysis, it used T, P and F test as the modeling determination.

並列關鍵字

technological forecasting ； backpropagation ； BPNN ； multinomial regression analysis ； polynomial regression analysis

參考文獻

[33] Yang, T. J., 2004, “Application of Back-propagation Neural Network in Storm Surge Forecasting,” Department of Resource and Environment, Leader University.

[1] Amjady, N., 2000, “Short-Term Hourly Load Forecasting Using Time-Series Modeling With Peak Load Estimation Capability,” IEEE Transactions on Power System, Vol. 16, No. 4, pp. 798-805.

[2] Amjady, N., 2007, “Short-Term Bus Load Forecasting of Power Systems by a New Hybrid Method,” IEEE Transactions on Power Systems, Vol. 22, No. 1, pp. 333-341.

[3] Bershad, N. J., 1997, “Statistical Analysis of a Two-Layer Backpropagation Algorithm Used for Modeling Nonlinear Memoryless Channels: The Single Neuron Case,” IEEE Transactions on Signal Processing, Vol. 45, No. 3. pp. 747-756.

[5] Bishop, C. M., 1995, “Neural Networks for Pattern Recognition,” Oxford: Oxford University Press.

國際替代計量

應用統計迴歸與類神經網路建立早期公開與核准專利的關聯性模型

全文下載

主題瀏覽