I have a MySQL database that contains datetime for every check-in with an RFID card. I have millions of records in it. It shouts machine learning prediction for me.

So I'd like to predict popular times to see when most people use the terminal. I'd like to represent it the same way as Google does at places:

Google's popular times panel at location search

This is the rare case when I do not ask for code, but keywords.

I assume I have to predict number of check-ins between time ranges. Also I am sure I have to make a dataset like this:

                        (These are the number of check-ins in hour ranges)
month | day | day-name | 0-1 | 1-2 | ... | 6-7 etc. 
jan   | 1   | Monday   |  0  |  1  |     |  12
jan   | 2   | Tuesday  |  1  |  3  |     |  15

So it can predict today's popular times based on what day it is, also by day of given month (as Saturdays and Sundays will be dead, also 25th of December. A good prediction should know these from the dataset).

As I wrote this question I solved most of it it seems. The only thing I need is a keyword as I have little experience in this. What model fits this best?

  • $\begingroup$ An average hourly check-in on the day of the week would not be enough for a forecast? $\endgroup$ – Guilherme IA Dec 1 '17 at 18:45

"Regression", is to estimate a continuous value as a function of some other parameters.

There are different forms of regression, such as linear regression and logistic regression, that differ in the assumptions they make on the data.

You could try looking up those for a start.

  • $\begingroup$ I used Accord.net's Multivariate Linear Regression for it and it beautifully and very accurately predicts the numbers, even on the weekends when there it predicts accurately less people in given hours. Thank you again for the answer! $\endgroup$ – Ákos Nikházy Dec 5 '17 at 13:48

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.