/boxed_lunch_sales_forecasting

CP1&CP2 for ARIH remote training project

Primary LanguageJupyter Notebook

Boxed Lunch Sales Forecasting

This repository provides a solution for a practice problem on SIGNATE: Boxed Lunch Sales Forecasting. The sales of boxed lunch at a certain company O in Chiyoda-ku are possibly related to many factors such as the menu, the weather, or the date. Our goal is to forecast the sales of the boxed lunch.

Dependency

  • Python 3.9.2
  • Numpy 1.19.2
  • Pandas 1.2.4
  • Scikit-learn 0.24.1
  • Feature-engine 1.0.2
  • Sklearn-pandas 2.1.0
  • Matplotlib 3.4.1
  • Seaborn 0.11.1

1. Exploratory Data Analysis

To perform exploratory data analysis and check the results, run code in notebooks/exploratory_data_analysis.

2. Machine Learning Modeling

To perform machine learning modeling, you can execute

  1. Jupyter Notebook code in notebooks/modeling, or

  2. Python code in train_valid_models.py.

Reference

[1] Walk-forward Validation

[2] Recursive Method for Multi-step Forecasting