This thesis develops a composite likelihood-based approach for multiple change-points estimation in general multivariate time series models. Specifically, we derive a criterion function based on pairwise likelihood and minimum description length principle for estimating the number and locations of change-points and performing model selection in each segment. By the virtue of pairwise likelihood, the number and location of change-points can be consistently estimated under mild conditions. The computation can be conducted efficiently with a pruned dynamic programming algorithm. Simulation studies and real data examples are presented to demonstrate the statistical and computational efficiency of the proposed method. / 本論文目的為開發一套以複合似然為基礎的多變點估計方法，該方法可應用於一般多變量時間序列模型。具體而言，我們在最小描述長度原理及成對似然的基礎上推導出一個準則函數，用於估計變化點的數量及位置，並在各段進行模型選擇。憑藉成對似然，在適當條件下變點的數量和位置可以一致地估計。透過使用修剪動態規劃算法，相關的運算能有效地進行。模擬研究及真實數據實例都演示出該方法在統計及運算效率。 / Ma, Ting Fung. / Thesis M.Phil. Chinese University of Hong Kong 2014. / Includes bibliographical references (leaves 51-54). / Abstracts also in Chinese. / Title from PDF title page (viewed on 05, October, 2016). / Detailed summary in vernacular field only.