行動環境下之使用者行為樣式研究—以二維度序列型樣進行探勘

The Research of User Behavior Patterns in Mobile Service Environments－By A Two Dimentional Sequential Pattern Mining

林柏伸(Bo-Shen Lin)

指導教授 ： 李維平

中原大學/商學院/資訊管理研究所/碩士(2004年)

https://doi.org/10.6840/cycu200400558

摘要

了解行動環境下的使用者行為樣式一直是近來熱門的研究議題，因其具有應用於行動商務的商業價值。然而，過去對此方面的研究大都侷限在單純的使用者移動樣式分析，如此並無法充分表達出行動環境下使用者所隱含的行為意圖。因此，本論文提出了一套適用於行動環境下的探勘方法，稱做二維度序列型樣探勘法。此方法對於行動環境下的地點、服務與間隔時間的關係有較充分的考慮。
在以往一般的序列探勘方法中，大都只能得知樣式中項目之間的先後關係，卻無法得知項目之間的詳細間隔時間。雖然也有學者透過事先設定固定區間的方式來挖掘時間資訊，但仍然受限於所訂定區間的大小。有鑑於此，本論文將時間點可能非常分散的特性透過分群的原理給區分出來，讓挖掘出來的序列樣式其時間資訊更富彈性。
在實作方面，我們發展出兩種新的探勘方法─M-PrefixSpan和M-Spade。在最後的實驗分析中，我們測試比較了這兩種演算法在精確性（Accurancy）、完整性（Completeness）及效率性（Efficiency）各項指標上的表現。從實驗結果得知M-PrefixSpan在各項效率上皆勝過M-Spade，為一較佳的二維度序列型樣演算法。

關鍵字

資料探勘、序列型樣、行動服務

並列摘要

Understand that the behavior patterns of user under mobile environments has been a hot research topic in recent year all the time, because it has commercial value applied to the mobile commerce. But the study on this respect mostly confined to simple user’s moving patterns analysis in the past, it is so unable to fully express user’s behavior patterns under mobile environments. So this thesis has proposed one data mining method that is suitable for mobile environments, call Two Dimentional Sequential Pattern Mining. This method has more abundant relation consideration between location, service and time interval under mobile environments.
The past sequential pattern mining methods can only mostly learn the order between items, but can’t learn the detailed interval time. Though some scholars mining time information by establishing fixed time interval, but still limited to the size of them. In view of this, this thesis uses clustering method to distinguish out the scattered of time and let the time information of sequential patterns full of more elasticity.
In making in fact, we develop two kinds of new mining methods - M-PrefixSpan and M-Spade. In the last experimental analysis, we have tested and compared these two kinds of algorithms of performing in the accuracy, completeness and efficiency. Learn from the experimental result that M-PrefixSpan all surpasses M-Spade on every efficiency, for one better two dimentional sequential pattern mining algorithm.

並列關鍵字

Data Mining、Sequential Pattern、Mobile Service

參考文獻

[崔青福] “行動網絡環境下之服務樣式探勘機制,“ 國立成功大學資訊工程研究所碩士論文, 2002。

[AMHU＋99] I. F. Akyildiz, J. Mcnair, J. S. M. Ho, H. Uzunalioglu, and W. Wang, “Mobility Management in Next-Generation Wireless System,“ Proc. Of the IEEE, Vol. 87, No. 8, August 1999.

[CS99] J. Chan and A. Seneviratne, “A Practical User Mobility Predicition Algorithm for Supporting Adaptive Qos in Wireless Networks“, IEEE ICON 99, January 1999.

[KMT99] M. Klemettinen, H. Mannila, and H. Toivonen, “Interactive Exploration Of Interesting Findings In The Telecommunication Network Alarm Sequence Analyzer(TASA),“ Information and Software Technology, Vol. 41, No. 9, pp.557-567, 1999.

[LT02] W. Lin and S. M. Tseng, “A Data Generator for Mobile Web Environment,“ Technical Report, CSIE Dept, National Cheng Kung University, Taiwan, 2002.