Forecasting Inflation Using Machine Learning

Specialeforsvar: Zhuorong Lin

Titel: Forecasting Inflation Using Machine Learning

Abstract: Due to the high inflation rate recently, inflation has become the ’it’ topic everywhere. So is it possible to predict inflation? This paper will try to explore the possibility of predicting inflation using machine learning and compare it with the more classic time series models such as ’AR’, ’MA’ and ’ARMA’. The goal here isn’t about optimizing one specific model, rather it is about investigating different models to see if there are any potential. The main models in focus are ’Adaptive Boosting’, ’Multi Linear Regression’, ’k-nearest neighbors’, ’Decision Tree’, ’Random Forest’, ’Epsilon-Support vector Regression’ and ’long short-term memory’. There will only be focus at regression problems and not classification problems. The best models found through machine learning will hereby be used in an attempt to predict inflation 3 years into the future, where we will compare these forecasting results with the model explained in ”Inflation Derivatives Explained” by Jeroen Kerkhof [4]. We will come to the conclusion that it is possible to predict inflation using machine learning, though not all machine learning models are equally good. Long short-term memory will turn out to be the best model based on future potentials.

vejleder:  Rolf Poulsen
Censor:    Lasse Engbo Christiansen, DTU