基於軟體品質在軟體業界越來越被重視,許多應用資料探勘與人工智慧的軟體品質預測技術已被提出,但尋求ㄧ個適當方法來建構軟體品質預測模式依然是個艱鉅的任務。本研究提出非線性粒子群分類樹(CTPSO)來建構軟體品質預測模型,此方法使用粒子群演算法來搜尋非線性函數其參數,產生分類規則及分類樹之節點,進而建構出樹狀分類模型來改善決策樹隱含性分類問題。模式驗證使用RSER benchmark中的KC2資料集進行實驗並與C5.0、CART、CHAID、QUEST、ANN、LR、SVM及GP等方法作效能測試,經訓練及測試證明在KC2資料中,CTPSO所建立之預測模式有較高的預測能力。
Software quality in the software industry has gradually taken seriously. Therefore, many techniques From the data mining methods and artificial intelligence use to establish software quality classification models have been proposed. But finding a suitable method of establishing Prediction model is still a difficult task. The study is published in CTPSO to build software quality prediction model. This method uses the PSO to search for Non-linear function of Parameter. Using this approach produce Classification rules and Nodes. Further development of the Classification model to improve Decision Tree is hidden problem. This study used experimental data sets for the KC2. This method for comparison to C5.0, CART, CHAID, QUEST, ANN, LR, SVM and GP. The results showed that KC2 used CTPSO to produce better prediction results.