在貝氏分類法中，簡易貝氏分類器與樹狀貝氏分類器都被廣泛的使用。兩者差別僅在於簡易貝氏分類器為屬性間彼此條件獨立，樹狀貝氏分類器則為每個屬性至多有一其它屬性與其相關，相較來看，樹狀貝氏分類器運算複雜度只有些微增加，正確率卻有顯著提升，被提出以來就逐漸受到重視；而兩分類器一般使用狄氏分配做為事前機率，但狄氏分配有兩個性質，第一為變數兩兩之間為負相關，第二為所有變數的正規化後變異數均相同，過去已有學者指出，簡易貝氏分類器使用狄氏分配為先驗分配是需要懷疑的，而樹狀貝氏分類器是否也有相同情形，即為本研究之方向；樹狀貝氏分類器建立資料結構的方法很多，在考量運算複雜度與分類正確率下，本研究採用SP-TAN為架構法則；另外廣義狄氏分配和羅氏分配也可作為貝氏分析先驗分配，而且比狄氏分配更為一般化，因此本研究分別實驗狄氏、廣義狄氏與羅氏分配為SP-TAN的先驗分配，並藉由12個資料檔不同參數組合所得的分類正確率作為分析的依據，結果發現在廣義狄氏分配與羅氏分配實驗中，在許多資料檔分類正確率比起狄氏分配實驗有所提升，顯示狄氏分配的兩個性質對於SP-TAN是有可能不合理的。

　　The naïve Bayesian classifier and the tree augmented naïve Bayesian classifier are two widely used classifiers. All attributes must be conditional independent in a naïve Bayesian classifier, while each attribute can have at most one correlated attribute in a tree augmented naïve Bayesian classifier. Thus, the tree augmented naïve Bayesian classifier has a higher computational complexity, but its prediction is generally more accurate. Both classifiers assume Dirichlet priors for attributes in calculating classification probabilities. However, the variables in a Dirichlet random vector will be negatively correlated and must have the same confidence level measured by normalized variance. A study has pointed out that the Dirichlet assumption can affect the prediction accuracy of the naïve Bayesian classifier. In this research, we will identify whether the Dirichlet assumption is feasible for the tree augmented naïve Bayesian classifier. The SP-TAN method is adopted to learn the structure of the tree augmented naïve Bayesian classifier. The generalized Dirichlet and the Liouville distributions are different extensions of the Dirichlet distribution and can be used as the prior distributions for the attributes in a tree augmented naïve Bayesian classifier. The experimental results on 12 data sets demonstrate that the Dirichlet assumption is infeasible to the tree augmented naïve Bayesian classifier.

Bier, V. M. and Yi, W. (1995). A Bayesian method for analyzing dependencies in precursor data, International Journal of Forecasting, 11, 25-41.

Blake, C. and Merz, C. (1998). UCI machine learning repository:
http://www.ics.uci.edu/~mlearn/MLRepository.html .

Cestnik, B. and Bratko, I. (1991). On estimating probabilities in tree pruning, Proceedings of the 5th European working session on learning on Machine learning, 138-150, Porto, Portugal.

Chinnasamy, A. and Sung, W. K. (2005). Protein structure and fold prediction using tree-augmented naive Bayesian classifier, Computer Science, 3(4), 803-820.

Chow, C. and Liu, C. (1968). Approximating discrete probability distributions with dependence trees, IEEE Transactions on Information Theory, 14(3), 462-467.

Cormen, T. H., Leiserson, C. E., and Rivest, R. L. (1990). Introduction to algorithms, MIT Press, Combridge, MA.

Dougherty, J., Kohavi, R., and Sahami, M. (1995). Supervised and unsupervised discretization of continuous features, Proceedings of the 12th International Conference on Machine Learning, 192-202, San Francisco.

Fang, K. T., Kotz, S., and Ng, K. W. (1990). Symmetric Multivariate and Related Distributions, New York: Chapman and Hall.

Friedman, N. and Goldszmidt, M. (1996). Building classifiers using Bayesian networks, Proceedings of the13th National Conference on Artificial Intelligence, 1277–1284, Menlo Park, CA.

Friedman, N., Geiger, D., and Goldszmidt, M. (1997). Bayesian network classifiers, Machine Learning, 29, 131–163.

Helman, P., Veroff, R., Atlas, S. R., and Willman, C. (2004). A Bayesian network classification methodology for gene expression data, Computational Biology, 11(4), 581-615.

Hsu, C. N., Huang, H. J., and Wong, T. T. (2003). Implications of the Dirichlet assumption for discretization of continuous attributes in naïve Bayesian classifiers, Machine Learning, 53, 235-263.

Jiang, L., Zhang, H., Cai, Z., and Su, J. (2005). Learning tree augmented naive Bayes for ranking, Proceedings of the 10th International Conference on Database Systems for Advanced Applications, 688-698, Berlin: Springer.

Keogh, E. J. and Pazzani, M. J. (1999). Learning augmented Bayesian classifiers: A comparison of distribution-based and classification-based approaches, Proceedings of 7th International Workshop on Artificial Intelligence and Statistics, 225–230, Ft. Lauderdale, FL.

Kohavi, R. and Sahami, M. (1996). Error-based and entropy-based discretization of continuous features, Proceedings of the 2nd International Conference on Knowledge Discovery and Data Mining, 114-119, Portland, OR.

Kononenko, I. (1991). Semi-naive Bayesian classifier, Proceedings of the 6th European Working Session on Learning on Machine Learning, 206-219, Porto, Portugal.

Pernkopf, F. and O’Leary, P. (2003). Floating search algorithm for structure learning of Bayesian network classifiers, Pattern Recognition Letters, 24, 2839-2848.

Pernkopf, F. (2004). Detection of surface defects on raw steel blocks using Bayesian network classifiers, Pattern Analysis and Applications, 7, 333-342.

Pudil, P., Novovicova, J., and Kittler, J. (1994). Floating search methods in feature selection, Pattern Recognition Letters, 15, 1119–1125.

Wilks, S. S. (1962). Mathematical Statistics, New York: John Wiley.

Wong, T. T. (1998). Generalized Dirichlet distribution in Bayesian analysis, Applied Mathematics and Computation, 97, 165-181.

Wong, T. T. (2005). The feasibility of the Dirchlet assumption in naive Bayesian Classifiers, A working paper.

Zhang, H. and Ling, C. (2001). An improved learning algorithm for augmented naive Bayes, Proceedings of the 5th Pacific-Asia Conference on Knowledge Discovery and Data Mining, 581 – 586, Berlin: Springer.