Shipping Optimization Challenge my solution

how to run

RUN

python src/prophet_v1.py

My environment

Python

• Python 3.7.9
• conda
• library(requirements.txt)

Computer

• Ubuntu 20.04
• CPU: i9 9900K
• Memory: 16 * 2 (GiB)

solution

Features

Since there is no data for days when there was no trade, I have added such data that such days would also be zero. (Adding it increased the score compared to not adding it.)

I didn't do features engineering just to prepare the data.

Modeling

I did the modeling based on prophet.

parmaeters

I used the data about holidays in the UK to train.

year_list = [2019, 2020]
holidays = make_holidays_df(year_list=year_list, country='UK')
params = {
    "growth": "logistic",
    # "changepoint_prior_scale": 0.05,
    # "seasonality_prior_scale": 10,
    # "mcmc_samples": 0,
    "seasonality_mode": "multiplicative",
    "daily_seasonality": True,
    "weekly_seasonality": True,
    # "changepoints": ["2020-03-23", "2020-05-11"],
    "holidays": holidays
    #  "prophet_pos": multiplicative
    # "likelihood": "NegBinomial"
}

What didn't work

• LGBM modeling(ex. https://www.kaggle.com/kneroma/m5-first-public-notebook-under-0-50)
• complex parameters
• features(on Prophet)