機器學習模型被廣泛應用於不同行業。模型的準確性取決於我們採樣的數據。在現實世界中，數據不可能一直是穩定不變的，有時環境會發生一些變化，可能會導致模型失效。模型切割可以幫助我們提升模型的準確度。先前有許多方法已經討論過要如何檢測數據集中改變的點，但很少有人在設計方法時考量到效率問題。我們認為當處於時間資源有限的情況下，時間效率可能會成為一個問題。換一種說法如果我們能在短時間內得到同樣的結果，那這樣的方法肯定是更好的方法。我們在這項研究中的貢獻是提出了一種基於貝氏優化器的分割方法，可以更快速有效地提高模型的預測能力。;Machine learning model is widely used in different industry. The accuracy of a model is depending on the data we sampled. In the real world, data cannot be static all the time, sometimes there will be some changes in the environment that may cause the failure in model. Segmentation is the answer to this question. Many works have discussed how to detect and determine the changing point in the dataset, and yet very little of them pay attention to efficiency while designing their method. In our opinion, time efficiency could be an issue if we are in the situation with limited time resource. In another way of thinking, if we can obtain the same result with short amount of time, the solution will definitely be the better way. Our contribution in this study is to propose a segmentation method based on Bayesian optimization which can improve the predictive ability of a model more efficiently.
