Suggestion for algorithm to use

[Paul B]

Hello, I've been a software developer with .NET for decades but am just now experimenting with machine learning. All of the options are a little overwhelming so I was hoping someone could point me in the proper direction.

We have sensor data for light, motion, temperature and sound that we record from a small device in a meeting room. We then observe the people in the room and also record that information and then later join it up to the sensor data. This will be our training and then eventually we would like to predict the number of people in the room based on just the sensor data. So for example, a single meeting would have data every few seconds that looks like:

Sound,Motion,Temp,Light,MeetingId,People

0,0,74,80,M1,2

23,1,74,80,M1,2

...

So we'll have a bunch of rows per meeting but we want to associate the number of people (2 in this case) with all those rows of data which is why we've added the meetingid to the data.

I've thought of doing this via clustering (with the categories being numbers from 0 to 10), linear regression (where the number of people is the solution?) or a random forest as suggested by a friend. Unfortunately I'm not sure where to start on any of these or if there is a better/simpler way to go about it.

Any advice?