應用詞頻以改良多元貝氏定理於文件分類之研究

CCUR > College of Business > Department of Information Management & Graduate Institute of Information Management > Thesis > Item 987654321/23850

Please use this identifier to cite or link to this item: https://irlib.pccu.edu.tw/handle/987654321/23850

Title:	應用詞頻以改良多元貝氏定理於文件分類之研究
Authors:	羅仁君
Contributors:	資訊管理學系
Keywords:	多元貝氏定理 multimembership Bayesian 文件分類 document classification 文件自動分類 automatic document classification 基因演算法 genetic algorithm
Date:	2011
Issue Date:	2012-12-04 10:30:30 (UTC+8)
Abstract:	多元貝氏定理(multimembership Bayesian，簡稱MMB)近幾年曾運用於醫療、網站、郵件和文件的自動分類與推論，到目前MMB的在知識分類推論領域上的相關研究一直都持續進行著。而本研究是針對MMB文件自動分類提出以基因演算法為概念的改良方法，使用動態產生篩選門檻來達到真正完全的自動分類。以基因演算法為概念的MMB改良方式「自動調適篩選門檻」來提取重要字詞進行MMB分類，最後也以「自動評估」取得最佳結果。經實驗發現，當類別彼此差異度大時，最佳分類準確率為83.93%；類別彼此差異度小時，其分類準確率為70.60%。 In recent years, multimembership Bayesian (MMB) has had a wide application for medical, website, E-mail and other document processing use the practices list above utilize the automatic classification and knowledge inference function of MMB to im-prove efficiency. Given MMB’s practicality and popularity across all walks of lives, the research around MMB remains a constant focus academically. To further improve the strength of MMB’s core automatic document classification function, our study proposes the additional application of genetic algorithm before traditional MMB. Based on the law of probability, the extra step of genetic algorithm helps develop the "automatic adaptable screening threshold" mathematically, thus with more accuracy. Such calculation pin-points the significant, frequently-used words to form the threshold for further MMB classification. Since the application of genetic algorithm acts as the in-itial screen, consequently, the extracted leftovers are more precise for any document classification. Further, based on the mathematic results, the threshold leads to automatic assessment, which selects the most desirable word choice automatically. The research presents significantly improved results. When class differences are relatively in large degree, its classification accuracy achieves 83.93%. Even when class differences are to a lesser extent, its classification accuracy rate of is able to reach 70.60%.
Appears in Collections:	[Department of Information Management & Graduate Institute of Information Management] Thesis

Files in This Item:

File	Size	Format
http___thesis.lib.pccu.edu.tw_cgi-bin_cdrfb3_gsweb1.pdf	149Kb	Adobe PDF	192	View/Open
http___thesis.lib.pccu.edu.tw_cgi-bin_cdrfb3_gsweb2.pdf	683Kb	Adobe PDF	379	View/Open
http___thesis.lib.pccu.edu.tw_cgi-bin_cdrfb3_gsweb3.pdf	6271Kb	Adobe PDF	889	View/Open

View Licence

Loading...