政府採購電子領投標系統招標文件閱覽之客戶行為評量與分析

Evaluation and Analysis of Customers Behavior on Browsing Bid Documents in Government Electronic Procurement System

關天盛(Tien-Sheng Kuan)

指導教授 ： 李錫捷

共同指導教授 ： 盧以詮

元智大學/資訊學院/資訊管理學系/碩士(2006年)

摘要

本研究架構於電子採購領投標系統之上。招標文件閱覽為該系統之加值服務功能之一，目的在於能使廠商對於機關招標標案，將所有標案之不確定性，降至最低，有助於廠商是否投標之決策。
於本研究中，使用Kohonen Model針對一百四十萬筆資料作分群，且先決條件為同屬相同標案類別，如工程類、勞務類與財物類。群組化後，便得知所有群組的詳細資訊。根據分群後的資料，選定此群組中的組成標案類別，預設組成標案類別擷取範圍為群組總和百分比大於80%，組成標案類別的細項標案類別百分比，每項必須要大於8%。利用組成標案類別的細項標案類別算出群組向量，再由群組向量推衍出訓練向量值與測試向量值。
最後，將利用相似度評量方法。在本研究中，使用的標籤相似度法(SimLabels)、SimNet、空間向量模型(Vector Space Model)中的Cosine coefficient等評量方法，並由程式將此三種評量方法，交由電腦運算批次處理，算出最終之訓練向量與測試向量之比值。藉此，從比值大小，便可輕易得知訓練向量與測試向量之間的相似度，更能回推此Kohonen Model是否合適，與研究目標是否契合。

關鍵字

電子採購 ； 領投標系統 ； 招標文件閱覽加值服務 ； Kohonen ； 群聚分析 ； Label Similarity ； SimNet ； Space Vector Model

並列摘要

This study is based on Government Electronic Procurement System. Browsing bid documents is one of the value-added services which is the best way to improve what cases is merchant needed. At the same time, all clients also want to spend the least expenses on getting right bid cases to tender bids.
In this study, I use Kohonen model to segment the whole over 1.4 million data into ten to twelve groups after the whole data segments to the right bid classes. For example, there are three bid classes; construction, labor service, and assets showing in my area. I realize all the details after being segmented. But there are some conditions that I have to design here. For instance, it picks up the groups that added is greater than 80%, and each class in the group has to be greater than 8%. Then, it is got vector by groups and each class of the group. Finally, we could reach all the values of vector of the training and testing.
As the result, I use three methods of similarity, like SimLabels, SimNet, SimCos. I use these methods to help me to calculate the three values of similarity and find out what happened in my study between training and testing vectors. According to the values of similarity, we also find out the similarity easily between two vectors in the end.

並列關鍵字

Electronic Purchasing ； Government Electronic Procurement System ； Browsing Bid Documents Service ； Kohonen Model ； Clustering Analysis ； Label Similarity ； SimNet ； Vector Space Model

參考文獻

[8]J. Han and M. Kamber, Data Mining Concepts and Techniques, Morgan Kaufmann Publishers, 2001.

[10]T. Hastile, R. Tibshirani, and J. Friedman, “The elements of statistical learning:Data Mining, inference and prediction”, Spring-Verlag, New York, 2001.

[12]C. Rygielski, J. C. Wang, and D. C. Yen, ”Data mining techniques for customer relationship management”, Technology in Society 24 (2002) 483-502.

[14]C. Bounsaythip and E. Rinta-Runsala, ”Overview of Data Mining for Customer Behavior Modeling”, VTT Information Technology, 2001.

[15]P. Berkhin, ”Survey of Clustering Data Mining Techniques”, Accrue Software Inc, 2002.