Home automation seems to be in it's infancy -- we use an arcane, platform specific language to write hard-coded rules that attempt to describe how we want automations to behave.
In some circumstances, HE collects enough data to make it possible to generate new rules that describe who we actually use our smart devices. For example, each time a switch (dimmer, button, etc) is activated -- whether manually or as the result of an application within HE, that provides a datapoint. These logs could be processed to report a series of actions with a high correlation.
Let me offer a simple example....we've got Lutron Caseta switches for lights. Some of these are tied into HE, with fairly simple rules (ie., if the hallway motion detector is active, at night, when the lights are off, turn on the hallway lights). However, there are complex sequences of activity that aren't currently programmed into HE, but which could easily be inferred from log data resulting from human beings manually activating smart devices. For example, if the living room motion detector is activated, then the first floor lights are turned on after 10pm, 90% of the time the 2nd floor stairway lights are turned on 15 seconds later, and 75% of the time the bathroom light is turned on, followed by all stairway lights & bathroom being turned off and the bedroom light being turned on.
It would be great to process those logs asyncronously, outside the Hub, and output suggested rules to be (manually) entered in HE. The implementation wouldn't necessarily need to produce output that is in the form of correct or complete HE syntax, but a human-readable representation of suggested rules.
