基於屬性向量進行二元分類是一種長期存在的問題，而線性判別分析可以利用組間變異數以及組內變異數之最大化比值找到一組線性組合向量進行分類，此研究是學習線性判別分析方法之經驗，特別是特徵選擇問題。在高維度以及高資料量的分類中，我們必須使用訓練資料集產生線性判別的相對應之係數，並且使用測試資料集加以證明精準度以確保可達到高效能的分類。本研究利用NSL-KDD99入侵偵測資料集進行分類，研究實驗指出線性判別分析法可以達到相當高的檢測精度。

Binary classification based on a feature vector is a long-standing problem. Linear discriminant analysis proposed by Fisher finds a linear combination of features such that the ratio of the variation between class to the variation within class is maximized. This thesis studies empirically the behavior of the linear discriminant analysis method, specifically on the feature selection problem.

[1] Fisher, R. A.(1936), The Use of Multiple Measurements in Taxonomic Problems, Blackwell
[2] Max Welling, Fisher Linear Discriminant Analysis, Department of Computer Science University of Toronto
[3] Petros Xanthopoulos, Panos M. Pardalos, Theodore B. Trafalis(2012), Robust Data Mining Part of the series SpringerBriefs in Optimization pp 27-33, Springer New York
[4] Webb, A. R. and Copsey, K. D. (2011) Linear Discriminant Analysis, in Statistical Pattern Recognition, Third Edition, John Wiley & Sons, Ltd, Chichester, UK. doi: 10.1002/9781119952954.ch5