scafati98/meli-data-challenge-2021

Machine learning model to forecast item inventory days based on Mercado Libre historical data.

meli-data-challenge-2021's Issues

Ideas to implement

• Merge train data set with .jl dictionaries.
• sum minutes active when sold quantity = 0 until sold quantity = 1 to have the amount of time the product was listed without being sold, divide this by the amount of days. This feature can be highly predictive.
• Watch how it distributes for different the item_id, item_domain_id and other cohorts.
• Make random split of the train set to make evaluations and hyperparameters optimizations on dev set.