Multi-task deep learning of time record events

Abstract

A method, computer system, and computer program product are provided for managing time record events. Time record events are collected for a number of users. Each time record event includes a geolocation of one of a number of users. The time record events and geolocations for each of the number of users are models via machine learning. A current geolocation for a given user is identified. A suggested event is predicted based on the current geolocation and a current time. The suggested event is pushed to the user.

1 - 21 . (canceled) 
     
     
         22 . A system, comprising:
 one or more processors, coupled with memory, to:
 predict, using a first machine learning model trained using time record events corresponding to geolocations, events according to a sequence of the time record events; 
 determine, using the first machine learning model, a probability for a timing of each of the events based on the sequence of the time record events; 
 identify a first geolocation associated with a profile using a client device associated with the profile; 
 select, according to the first machine learning model, an event of the predicted events based on the first geolocation and a first time corresponding to the timing; 
 display the event on the client device associated with the profile; 
 receive feedback regarding the event from the client device associated with the profile; and 
 retrain the first machine learning model based on the time record events and the feedback using a second machine learning model. 
   
     
     
         23 . The system of  claim 22 , wherein the geolocations corresponding to the time record events comprise a clock-in or clock-out location of a plurality of profiles associated with a plurality of client devices. 
     
     
         24 . The system of  claim 22 , wherein the first machine learning model comprises multimodal multi-task learning. 
     
     
         25 . The system of  claim 22 , comprising the one or more processors to display the event on the client device associated with the profile responsive to the first geolocation corresponding to a geolocation of the event. 
     
     
         26 . The system of  claim 22 , wherein the first machine learning model trained using the time record events corresponding to the geolocations is trained for each of a plurality of profiles associated with client devices. 
     
     
         27 . The system of  claim 22 , comprising the one or more processors to determine, using the first machine learning model, the probability for the timing of each of the events based on the sequence of the time record events and based on the profile. 
     
     
         28 . The system of  claim 22 , wherein the feedback comprises a change in the first geolocation. 
     
     
         29 . The system of  claim 22 , wherein the feedback comprises a change in the timing. 
     
     
         30 . The system of  claim 22 , comprising the one or more processors to:
 establish the event is a first event; and   select, responsive to receiving the feedback, according to the second machine learning model, a second event of the predicted events based on the first geolocation, the first time corresponding to the timing, and the first event.   
     
     
         31 . The system of  claim 22 , wherein the first machine learning model is a recurrent neural network. 
     
     
         32 . A method, comprising:
 predicting, by one or more processors, coupled with memory and using a first machine learning model trained using time record events corresponding to geolocations, events according to a sequence of the time record events;   determining, by the one or more processors using the first machine learning model, a probability for a timing of each of the events based on the sequence of the time record events;   identifying, by the one or more processors, a first geolocation associated with a profile using a client device associated with the profile;   selecting, by the one or more processors according to the first machine learning model, an event of the predicted events based on the first geolocation and a first time corresponding to the timing;   displaying, by the one or more processors using the first machine learning model, the event on the client device associated with the profile;   receiving, by the one or more processors, feedback regarding the event from the client device associated with the profile; and   retraining, by the one or more processors, the first machine learning model based on the time record events and the feedback using a second machine learning model.   
     
     
         33 . The method of  claim 32 , wherein the geolocations corresponding to the time record events comprise a clock-in or clock-out location of a plurality of profiles associated with a plurality of client devices. 
     
     
         34 . The method of  claim 32 , comprising displaying the event on the client device associated with the profile responsive to the first geolocation corresponding to a geolocation of the event. 
     
     
         35 . The method of  claim 32 , wherein the first machine learning model trained using the time record events corresponding to the geolocations is trained for each of a plurality of profiles associated with client devices. 
     
     
         36 . The method of  claim 32 , comprising determining, using the first machine learning model, the probability for the timing of each of the events based on the sequence of the time record events and based on the profile. 
     
     
         37 . The method of  claim 32 , wherein the feedback comprises a change in the first geolocation. 
     
     
         38 . The method of  claim 32 , wherein the feedback comprises a change in the timing. 
     
     
         39 . The method of  claim 32 , comprising:
 establishing the event is a first event; and   responsive to receiving the feedback, selecting, by the one or more processors according to the second machine learning model, a second event of the predicted events based on the first geolocation, the first time corresponding to the timing, and the first event.   
     
     
         40 . A non-transitory computer-readable medium, comprising instructions embodied thereon, the instructions to cause one or more processors to:
 predict, using a first machine learning model trained using time record events corresponding to geolocations, events according to a sequence of the time record events;   determine, using the first machine learning model, a probability for a timing of each of the events based on the sequence of the time record events;   identify a first geolocation associated with a profile using a client device associated with the profile;   select, according to the first machine learning model, an event of the predicted events based on the first geolocation and a first time corresponding to the timing;   display the event on the client device associated with the profile;   receive feedback regarding the event from the client device associated with the profile; and   retrain the first machine learning model based on the time record events and the feedback using a second machine learning model.   
     
     
         41 . The non-transitory computer-readable medium of  claim 40 , comprising the instructions to cause the one or more processors to:
 establish the event is a first event; and   select, responsive to receiving the feedback, according to the second machine learning model, a second event of the predicted events based on the first geolocation, the first time corresponding to the timing, and the first event.