BrainJS
interval between dates
Guys, can someone please help me to set up a network to train the frequency at which the user visits the establishment?
If the user goes X days without going to the establishment, he receives a notification. I would like to assemble something to train between the ranges of dates, I already have the data of dates and times of users' frequency in a database.
Not sure if you're still working on this, but sounds like you'd like to try to infer normal user visitation patterns and trigger a notification when this deviates? Or am I misunderstanding the use case?
Hey, thanks for the feedback! That's exactly what I'm still looking for.
I see! Well, I'm no expert, but I'd imagine the best starting place would be to consider the problem like this:
- Each customer's historic visitations represents a separate set of inputs. That is, Alice's visitation pattern has no bearing on Bob's. Consequently, each customer would have their own model to derive visitation predictions from.
- It may be easier to conceptualize and to normalize training data by looking at visitation as an array of intervals between visits instead of datetimes.
- Each customer would have their own training data, which would be appended to with each subsequent visit and the model retrained. With that you could predict their next N visits, and store that denormalized value in your customer database for the purpose of notifications.
- Hypothetically, you could also fabricate a data set from existing customers and demographics/heuristics as a sort of 'customer profile', and use that for new customers meeting the criteria, until that customer has enough actual data to create accurate predictions.
Am I barking up the right tree here?