分類問題在金融業、電商業抑或是醫療業無處不在。舉例來說，金融業透過儲戶的年齡、年收入、教育和歷史還款紀錄來預測其信用評等，而這些信用評等屬於類別型變數。此外，深度學習模型的蓬勃發展也反映出分類問題的重要性。另一方面，在電腦資源的限制下，伴隨著資料量的快速成長，多樣的資料縮減方法不斷地被提出。在本篇論文中，我們利用資料縮減的概念發展出適用於分類問題的預測模型，此外，也透過模擬與實際案例以展示我們提出的方法。;In financial, telecom, or medical industry, classification problems are ubiquitous. For example, the financial industry predicts a depositor′s credit rating based on several input variables such as age, annual income, education, and repayment history, where the responses are qualitative. More and more deep learning models are developed for such purposes, reflecting the importance of classification problems. On the other hand, with the rapid growth of data size given limited computer resources, various data reduction methods have been innovated. In this thesis, we utilize a concept of data reduction to develop a classification predictor. We illustrate the proposed method through simulations and real examples.
