If a person enters a dark room they are very familiar with it is highly probable that they will turn on the light. If a person enters a well lit room they are not familiar with it is highly improbable that they would turn on the light. I'm sure there are a lot of situations where my generic baseline wouldn't apply.

100% Familiar, 100% Dark
999:1000

50% Familiar, 50% Dark
500:1000 >> 1:2

0% Familiar, 0% Dark
1:1000

If you don't need to do this dynamically, or need to do the calculations yourself, then you can use data-mining software like Weka 3 to extract the knowledge you need from your history data.
You can use Weka to train NNs, construct decision trees (which you could then hard-code into an app with nested 'if's), etc...

I actually work in the access control business and one of my colleagues wrote a prototype program that analysed all past access events to determine if any new given event was unusual. However I don't have any details on it and wouldn't be allowed to share them if I did.

Good luck though, it is certainly achieveable with some degree of accuracy.

Quote: Original post by Hodgman
If you don't need to do this dynamically, or need to do the calculations yourself, then you can use data-mining software like Weka 3 to extract the knowledge you need from your history data.
You can use Weka to train NNs, construct decision trees (which you could then hard-code into an app with nested 'if's), etc...

yes, my problem is thinking on which data to feed to this algorithms
I belive that just something like the lines I pasted on my first post won't make any of this algorithms find complex rules like the ones I told before

Quote: Original post by kavelot
the problem I find is that the event aren't exactly the same
for example, person A may come from work everyday around 6:30pm and turn the lights on, but some days he may come 6:35pm, other days 6:21pm, etc. I could try to use some stats to determine something like "I want a time interval where I'll be correct 90% of the times", but an algorithm like that won't make more complex rules (for example, "it just turn the lights on if there isn't a person on the sector for more than 8 minutes")

What you're describing is the need for a Fuzzy Logic system. With a 6:15 to 6:45 fuzzy triangle, 6:35 would be a 0.666~ truth value. That number can be used as part of the calculations for determining events.

For instance, with the 6:30AM Event having a higher fuzzier value than the 6:00AM or the 7:00AM events, the 6:30AM event could be the one picked as a reasonable event to base triggers off of. So, if the lights should turn off about 12 hours after being turned on, but aligned to a half-hour mark, the 6:30PM mark would be the target. Like-wise, having a 9:14AM event would push the 9:00PM trigger.