Conclusion
This study explores the k-means machine learning algorithm for building core inflations measures. As an empirical example, a core inflation measure, the k-vol core, is built based on reducing volatility. This indicator is evaluated on two criteria: its ability to grasp the inflation signal and its capacity to forecast future inflation. The results show that the k-vol core outperforms trimmed mean indicators and core inflation excluding food and energy. The study aims to contribute to the inflation literature and to provide a useful tool to policy makers when taking monetary policy decisions.
