Future prediction of regional city based on causal inference using time-series data

Regional cities in Japan have a lot of social issues. Various measures are being considered to solve these social issues, but it is difficult to ascertain and implement practical and effective measures to address them. In this study, we proposed a method for selecting indicators that have a causal relation to solve the social issues based on a causal inference. If there was a causal relation between two sets of time-series data, the slope of the approximation line of the time-shifted correlation coefficients at the base time returned a negative value. The causal inference was verified by using samples of time-series data and we constructed a network of the causal indicators based on the causal inference. In addition, we achieved future predictions via the vector autoregressive model using the network of causal indicators. The model was verified using the actual time-series data of the 87 regional cities. As a result, it was possible to simulate future predictions by introducing practical and effective measure that originated from the social issue.