Forecasting Inflation (Consumer Price Index) using LSTM (Long Short-Term Memory)

Crux of the Project

• The project is carried out on the Economic Data of the United States of America because of the dominating and stable value of $.
• This repository contains the code for the Machine Learning & Deep Learning-based Economy Inflation Prediction Model. The main parameter considered for measuring inflation is CPI (Consumer Price Index).
• For reference, if CPI '24 = x and CPI '25 = y, then Inflation Rate from 2024-2025 = [(x-y)/x]*100.
• The project is developed using algorithms like Linear Regression, K-Nearest Neighbour and LSTM (Long Short-Term Memory) with each depicting certain accuracy in diverse conditions.
• Various methods have been applied in the LSTM-based model: Manual Analysis | Stop-loss | Keras Tuner | Weighted Average Ensembling - Among these, Manual Analysis proved to give the best result.

Data Sources

• CPI: Link
• BIE: Link
• Oil Prices: Link
• PPI: Link
• NEER: Link
• US Unemployment Data: Link

References

1. The Impact of Machine Learning on Economics by Susan Athey
2. Inflation Prediction Method Based on Deep Learning by Cheng Yang and Shuhua Guo
3. Prediction of Economic Growth by Using Machine-Learning Algorithms through Sentiment Index Analysis in the Economy of Pakistan by Asad Tanveer, Idrees Afzal, Kainat Fatima, Sahira Bano.
4. FinTech and the future of financial services: What are the research gaps? by Anil Savio Kavuri and Alistair Milne