i wanted to ask in assistance on modeling a problem we want to solve. We would like to calculate the probability of finding a free parking space in 2 parking lots given that:
- we have online real value of free parking spaces in both parking lots (can't get each separately) - it's not 100% accurate but its not bad. - of course if we have more than X slots available, we can consider the probability to be 100% or close to 100%.
- we know the difficult hours (that usually the parking lot is full) - this should look like a bell (around 10:00 AM its very likely that the parking lots will be full, so manual updated - see later - should be more reliable)
- we know the rate of people getting in and leaving the parking lot (after not finding a parking space) - there's also a probability that someone entered, and another one exited - we can't know that for sure (lets say we have 6 free parking spaces. in 11:35 - its 5 (someone entered), in 11:37 it back to 6 (someone didn't find a parking), etc... The rate is usually relevant to parking lot A, because Parking lot B's barrier will not be opened if the parking lot is full (of course if someone else which is not registered parked, the barrier will open). (so we can guess those people came in, didn't find anything, and left the parking lot)
- manual updates - someone enters the parking lot and can update if he sees enough parking or not (seperated into A and B parking lots) - the updates are relevant mostly when they are closed to "Now" (meaning, if i see update 20 minutes ago, i should consider it less than update from 3 minutes ago).
i can try to model it with weights for each parameter - the weights should change dynamically depending on the current inputs (for example, manual update of "full" in parking A will probably receive weight of 0 when the hour is 21:00 at night), etc...
but! as i'm not mathematician, i'd love to consult with you what's the best approach to model it?
thanks