New Method for Structural Change Detection of Time Series as an Optimal Stopping Problem

In general, an appropriate prediction expression and/or model is constructed to fit a time series though, the model begins to unfit (or not to fit) the time series from some time point, especially in the field that relates to human activity and social phenomenon. In such case, it will be important not only to quickly detect the unfitting situation but also to rebuild the prediction model after the detection as soon as possible. In this paper, we formulate the structural change detection problem in time series as an optimal stopping problem, using the concept of DP (Dynamic Programming) with a cost function that is the sum of unfitting (or not fitting) loss and action cost to be taken after detection. And we propose a method for optimal solution and show the correctness by proving a theorem. Also we clarify the effectiveness by showing the numerical experimentation.