Machine Learning in Inflation Prediction for the Finnish Economy
Pant, Sukrit (2023)
Julkaisun pysyvä osoite on
https://urn.fi/URN:NBN:fi-fe2023051042804
https://urn.fi/URN:NBN:fi-fe2023051042804
Tiivistelmä
With growing inflation faced by countries around the world impacting the lives of citizens for each economy, there is a need for a model that can accurately predict inflation for better decision-making. Machine learning techniques have evolved in the past few years, with newer algorithms being invented and improving on previous models.
Most of the research focused on larger economies such as the USA, the UK, Germany, and China. There is a demand for studies that are focused on other economies. Similar studies have not been conducted in Finland. At least those that are available for researchers and economists. The research focuses on improved accuracy of machine learning algorithms compared to traditional econometrics models. A multivariate analysis using a machine learning algorithm is performed to determine the best-performing model for the Finnish economy.
Similarly, the impact of adding further variables in inflation forecasting models is analyzed to understand the change in inflation accuracy. The study suggests that multi-layer perceptron performs the best for both sets of analysis, while cross-validation results in support vector regression for smaller datasets and LASSO for larger datasets. Also, the study reveals that adding more variables impacts the accuracy negatively for most of the algorithms.
Most of the research focused on larger economies such as the USA, the UK, Germany, and China. There is a demand for studies that are focused on other economies. Similar studies have not been conducted in Finland. At least those that are available for researchers and economists. The research focuses on improved accuracy of machine learning algorithms compared to traditional econometrics models. A multivariate analysis using a machine learning algorithm is performed to determine the best-performing model for the Finnish economy.
Similarly, the impact of adding further variables in inflation forecasting models is analyzed to understand the change in inflation accuracy. The study suggests that multi-layer perceptron performs the best for both sets of analysis, while cross-validation results in support vector regression for smaller datasets and LASSO for larger datasets. Also, the study reveals that adding more variables impacts the accuracy negatively for most of the algorithms.
Kokoelmat
- 512 Liiketaloustiede [433]