Time forecast - Predict a persons chances of login tomorrow, with a predicted time

python

#1

My data set contains the worker information for 3 year. it contains their login day, and time. assumptions the workers may take leave or some days are public holidays.
But when ever the worker is logged in the time and date of that person is recorded.

So we have fields with ,

workerid , login time, login date

for all workers are available

Friend using only this data how we can predict, worker AAA will login tomorrow at … (this) time?


#2

Hi,
You will have to feature engineer the data in order to arrange the dataframe to suit your requirement.

Some of the ways could be

  1. Simply arranging the data grouped by the workerID and finding patterns of login/date on it.
    And then perhaps indicating a flag column forecasting his next day’s action.

  2. You should also try and cluster the workers into certain groups based on their login behaviour and then predict their probability of logging in the next day.

If the exact time prediction is difficult, you can subset the time column into 3-4 groups (viz. morning hours, afternoon hours, evening, night, etc)

So, in a nut shell, this may not be strictly a time series problem, but rather a classification problem

Hope this helps.


#3

Hi ,

Asmit , Thanks for your points.
in my case, the clustering will not work, for each worker there are some task waiting, depending on the prediction i want to assign the task later. So i am looking for a prediction model with time , from their historical login time. Arima model is taking so much time. Any other samples available?