Improvements to Consumption Prediction

This repository contains the paper Improvements to Consumption Prediction: Machine Learning Methods and Novel Features as well as reference code and associated data.

Improvements to Consumption Prediction: Machine Learning Methods and Novel Features was published in the SMU Data Science Review in 2018. Link: https://scholar.smu.edu/datasciencereview/vol1/iss4/

These Jupyter Notebooks contain a portion of the associated code for the paper.

Economic variable stationarity: pce_econ_stat.ipynb
Sentiment variable stationarity: pce_sent_stat.ipynb
Vector Autoregression (VAR) models: pce_var_model.ipynb
Random Forest models: pce_rf_model.ipynb

Requirements

Python 3.*

Anaconda Python distribution (recommended)

Conda Environments

Additional Python packages (included with Anaconda distribution) -jupyter -matplotlib -numpy -pandas -seaborn -sklearn -statsmodels

Installation

1. Install Anaconda

Reference:
macOS: https://docs.continuum.io/anaconda/install/mac-os.html
Windows: https://docs.continuum.io/anaconda/install/windows

2. Create conda environment from the pce-requirements.yml file

Reference: https://docs.conda.io/projects/conda/en/latest/user-guide/tasks/manage-environments.html

Example Code: conda env create -f pce-requirements.yml --name impr-pce-env

3. Activate the new environment

Example Code:
macOS: conda activate impr-pce-env
Windows: activate impr-pce-env

4. Launch Jupyter Notebooks

Reference: https://jupyter-notebook.readthedocs.io/en/stable/

Example Code: jupyter notebook

gkoswald / Consumption_Prediction