Forecasting CPI Inflation with Hierarchical Recurrent Neural Networks

This Repository contains implementation of novel model we suggested in our paper.

1. Hierarchical GRU

In addition as a contribution for further research we provide he data set used in this work, which is taken from the BLS after parsing and pre-processing.

The data could be found on data/cpi_us_dataset.csv note: this is a sample from the all data

A brief description of the data set:

The data set contains the following columns:

• Date (differs by month)

• Category- item name

• Category id- item unique id

• Price- Seasonally adjusted CPI-U for the given month

• Weight- Relative importance of the item from the total aggregated index (=100)

*Indent- The hierarchy level (total aggregated index has indent 0, lowest level is 8)

*Parent- Parent’s item name

*Parent ID- Parent’s item ID

In order to run the code please make sure all Prerequisites are met (Pandas==0.22 in particular)

Prerequisites

pip install -r requirements.txt

To execute the code please run one of the following

1. hierarchical_gru.py

Article

@article{barkan2023forecasting, title={Forecasting CPI inflation components with hierarchical recurrent neural networks}, author={Barkan, Oren and Benchimol, Jonathan and Caspi, Itamar and Cohen, Eliya and Hammer, Allon and Koenigstein, Noam}, journal={International Journal of Forecasting}, volume={39}, number={3}, pages={1145--1162}, year={2023}, publisher={Elsevier} }