search

artc-dsc / Tasks

For recording purpose.
0 stars 0 forks source link

Establish accuracy benchmarking as baseline #47

Open ARTC-Doris opened 4 days ago

ARTC-Doris commented 4 days ago

  1. Models: (Seasonal)ARIMA + Prophet + Naive + Exponential smoothing + Croston for baseline see select_model_n_param() in subprocess_model_evaluation.py

  2. Start from SKU-monthly to SKU-weekly

  3. Error measurement on 13-week and first month MAPE with zero adjustment see mape_adj() function in evaluation.py Input: sales_resampled_sku_code_m.csv

  4. Feature extraction: COKE_02_demand_profiling.ipynb