There are several dealer-to-dealer automobile auctions in the United States and more worldwide. Each day tens of thousands of used vehicles are posted by dealers for purchase by other dealers.
For the dealers that are buying cars, they want to purchase used vehicles they can easily and profitably resell. Each dealer is unique but some combination of make, model, year (years old), mileage, asking price, and used car value from services such as NADA and Kelly Blue Book (KBB) and perhaps even seasonal variations.
Automating this process to provide recommendations across multiple auctions to a specific dealer taking into account that dealers’ specific need could be an excellent application of machine learning.
The set of cars for sale change each day so that all the cars have to be loaded into the recommendation system, used car values found and a machine learning model run for each dealer to come up with a recommendation.
This system would have an area where dealers could fill out a profile with parameters on which to build a machine learning model.
Test and training data would have to be automatically obtained and updated on a regular basis as automotive models change over time.
Once a model is created and trained for the dealer it can be run each day against the current day’s set of cars and the recommendations communicated with the dealer.
Future expansion could expand this into automatically bidding (possibly another machine learning task) using the daily recommendations.