| Summary: | 碩士 === 國立臺灣大學 === 企業管理碩士專班 === 107 === The term “recession”is not only a sensitive topic to the workers, investors, and businesses that suffer immense loss during this time, but also economists whom continually struggle to predict them. Through the rise of Big Data, Machine Learning potentially gives statisticians and economists alike a new tool for predicting recessions. In order to explore this field, this paper asks two fundamental questions:
• Does machine learning help classify and forecast recessions within the business cycle?
• Which models are most effective in predicting and classifying recessions?
Applying the most common machine learning classification algorithms, we perform out-of-sample performance evaluations using a self-selected sample of macroeconomic indicators. Our findings imply that Random Forest, KNN, and Support Vector Machine models best classify and predict recessions. The analysis results suggest that machine learning has incredible potential in improving prediction accuracy. However, we believe these models can be further developed through additional research and application within the Deep Learning field of machine learning.
|
|---|
