Read more
This book uses R package iForecast to conduct financial economic time series forecasting with machine learning methods, especially the generation of dynamic forecasts out-of-sample. Firstly, the machine learning methods cover, for example, enet, random forecast, gbm, and autoML etc., including high binary economic time series. Secondly, I will explain the problem about the generation of recursive forecasts in machine learning framework, under which, there are no covariates, namely, input (independent) variables. This case is pretty common in real decision environment, for example, the decision-making wants 6-month forecasts in the real future, under with, there are no covariates available; therefore, what we can use is recursive, or multistep, forecasts. Besides, macro-econometric modelling uses VAR (vector autoregression) to overcome the problem of multivariate regression, this book offers a Machine-Learning VAR routine, which is found to improve the performance of multistep forecasting.
List of contents
Preface.- Chapter 1 Time Series Basics in R.- Chapter 2 Predictive Time Series Modelling.- Chapter 3 Forecasting using Machine Learning Methods.- Chapter 4 Special Topics.- Chapter 5 Predictive Case Studies Training by Rolling.- References.
About the author
Tsung-wu Ho is a professor at National Taiwan Normal University. His research interests are Asset Pricing, Machine Learning, Economic and Decision Making.
Summary
This book uses R package iForecast to conduct financial economic time series forecasting with machine learning methods, especially the generation of dynamic forecasts out-of-sample. Firstly, the machine learning methods cover, for example, enet, random forecast, gbm, and autoML etc., including high binary economic time series. Secondly, I will explain the problem about the generation of recursive forecasts in machine learning framework, under which, there are no covariates, namely, input (independent) variables. This case is pretty common in real decision environment, for example, the decision-making wants 6-month forecasts in the real future, under with, there are no covariates available; therefore, what we can use is recursive, or multistep, forecasts. Besides, macro-econometric modelling uses VAR (vector autoregression) to overcome the problem of multivariate regression, this book offers a Machine-Learning VAR routine, which is found to improve the performance of multistep forecasting.
