Accurate total consumer price index forecasting with data augmentation, multivariate features, and sentiment...

Area of Science:

• Economics
• Econometrics
• Data Science

Background:

• The Consumer Price Index (CPI) is a critical economic indicator influencing monetary policy and economic stability in Korea.
• Accurate CPI forecasting is essential for policymakers, but faces challenges due to data heterogeneity, sparsity, and external volatility.

Purpose of the Study:

• To develop a novel framework for enhancing the accuracy of Korean Total CPI forecasting.
• To address the limitations of existing methods in capturing complex price dynamics and external influences.

Main Methods:

• A hybrid Convolutional Neural Network-Long Short-Term Memory (CNN-LSTM) model was developed to capture intricate patterns in CPI data.
• Multivariate inputs, including CPI component indices and auxiliary variables, were utilized for richer contextual information.
• Data augmentation techniques (linear interpolation) converted monthly to daily data, and a sentiment index from news articles was incorporated.

Main Results:

• The proposed framework demonstrated superior performance in CPI prediction compared to existing approaches.
• Lower Root Mean Square Error (RMSE) values indicated improved estimation accuracy.
• The model effectively integrated diverse data sources, including sentiment analysis, for enhanced forecasting.

Conclusions:

• The novel hybrid deep learning framework significantly improves the accuracy of Korean Total CPI forecasting.
• Enhanced prediction accuracy supports the development of more effective and timely economic policies.
• This approach offers a robust solution for navigating the complexities of inflation monitoring and economic management.