I have a project in mind that I would like to see if I can use ML.net to return data.

The data exists in multiple tables in a DB, but it is relational.

On Microsoft’s page: https://learn.microsoft.com/en-us/dotnet/machine-learning/how-to-guides/load-data-ml-net

It looks like I cannot load multiple models in at the same time. So, I was considering writing a big query to pull all the data in the tables in one big query, joining the tables together.

To simplify this, I’ll use another example model similar to what I was looking at working with.

public class house
{
    public int id;
    public int addressNumber;
    public string street;
    public string color;
    public string directionFacing;
    public int numberOfDoors;
    public bool isWalkOut;
    public string ownerFirstName;
    public string ownerLastName;
}

What I want to do is query the ml.net data model with something like:

“What is the color and facing direction at address 1212 23rd Avenue”

“What is the walkout style at address 2312 Oak Street”

“What is the owners name at address 4008 Fordham Drive”

Again, this is just sample data.

I know I can do this using SQL queries, but the questions would need to be exact and the person asking the question would need to be trained “how” to ask the question.

I have played with Catalyst to flatten the questions. That in the combination with SQL queries, I’ve been able to get a workable app. But, if the questioner asks things out of order, it gets more complex. And I would have about 90 columns I need to be able to query. That’s where I’d like to see if I can get ml.net to work with this. I need to return data from any of the columns.

All the videos and examples I’ve seen thus far refer to returning only the likely row(s) that are being queried, not the data within the row.

I assume this is possible somehow with ML.net…