How to utilize

asav_yu

I have historical auction sales data of various products. I set the prices (reserve), some products sell, some do not. I control whether to let the product go for less than reserve price. Some products sell for more than reserve.

Then I use whatever sold to train a model to predict & set prices for future sales. However, whatever did not sell I cannot use. I feel like I can extract insight from not sold data as well.

Ideally I would love to have not only predicted price but also probability of sale at that price.

This is a conceptual question as I have no idea how to approach it and would love to pick your brilliant minds. Any ideas are very much appreciated!

rfuentealba

Hello @asav_yu

How does your data look like? I am venturing something like:

id | name | category | price | reserve

You should first check if you can get stuff like "it's from the original owner", "it's historical", "it's red", "it's restored to original condition" (I'm inventing based in a TV programme that shows auctions). Everything you have can help. Even the day of the auction (people go to auctions on Saturdays but not on Mondays) might help you.

Once you get your historical data in shape, you should train something like a decision tree, a neural network, a logistic regression... try different algorithms to check which one will help you predict first if the product will be sold at price of reserve, higher or lower, or not at all. Check if there are variables that influence the decision (like: "historical products that belong to Luxembourg Army are always sold if the price is under USD $ 10.000", or "Swiss products always sell", or "Chinese products sell only if they are culturally bond").

The process here isn't something you could solve in an hour. But let's start with what data you have first.

All the best,

Rodrigo.