AI for homeseer

[alphatech]

I was wondering, there is no much talk here on this forum integrating AI with HS, is there a need but hard to implement?

Sent from my SM-G973U1 using Tapatalk

[Eman]

Try here : https://www.google.com/search?client...&q=ai+for+home

If you spend some time and craft your using data logging ( https://grafana.com/blog/2019/10/29/...-your-kitchen/ and more here #41: Datalogging with MQTT, Node-RED, InfluxDB, and Grafana – SuperHouse Automation) mqtt and node-red then I think the values stored can trigger devices...


Eman.

[alphatech]

Interesting, I have all of those products in place, trying to figure out how I would use AI for home automation. As an example, I collected and still collecting all the sensors data in a sql database, theoretically I could feed those sensors for AI processing.

Sent from my SM-G973U1 using Tapatalk

[puma]

I think it’s too ambitious and unrealistic to see meaningful results. Companies like Google, Tesla, Apple invest $billions in AI and still not much to brag about.

[Eman]

Interesting, I have all of those products in place, trying to figure out how I would use AI for home automation. As an example, I collected and still collecting all the sensors data in a sql database, theoretically I could feed those sensors for AI processing.

Sent from my SM-G973U1 using Tapatalk

Right here : https://forums.homeseer.com/forum/ul...chine-learning

I tried a while back to get people get excited about this or people had most of the stuff in place to make HomeSeer interesting but found that people are most likely to fancy their graphs and data but not make use of it, say to add plugins like EasyTrigger, Smart Device or mcsMQTT and have the system use the data and trigger events intelligently or based on averages ...etc

It doesn't have to be big-data but half way there.

If you want to play start here : https://developer.ibm.com/components...-tensorflowjs/



Eman.

[alphatech]

Thanks for sharing that article, the thermostat example is a perfect example for it. I know aws offer several tools but to be honest they are all above my brain power.
Hopefully one day someone here will write a nice Plugin for it

Sent from my SM-G973U1 using Tapatalk

[kideon]

I’ve seen a few projects over the years where people used machine learning to predict behavior and automate devices

[Kigmatzomat]

Its a slow pain for most people.. Even thermostat training requires you to go through the process of letting the thermostat be wrong repeatedly for a week while you make constant adjustments until it learns. That is a total SAF barrier. I hate to say it, but lots of ML really needs cloud levels of data to make inferences across populations in semi-realtime. I.e. these 500 people program in holidays so if most of them disable routines on a weekday, it must be a holiday and offer to make that change for all users.

If you can live without the "realtime" aspect, a hint system to provide event programming guidance could work. The problem is, it will take weeks or months to provide the first insight. "You adjust the thermostat to be 2F warmer most weekend nights in winter; would you like to adjust the hvac schedule?" "The doors are manually locked at 11pm. Would you like an autolock event?"

people will get bored and uninstall it.

Alternately, you could get a lot of HS backups to extract events vs devices to build a dictionary of common "if device a and b then event c" logic. Like "you installed a smoke detector. Would you like to add an event to send emails and/or turn on lights on an alarm?" Or "you installed a light switch in a bedroom, would you like a wake-up alarm?"

that would provide an immediate benefit and then the ML aspect would be a once-a-month notice of "new insight found".

[alphatech]

Great explanation and examples, I see great points here.

Sent from my SM-G973U1 using Tapatalk