Network computer system to implement predictive time-based determinations for fulfilling delivery orders

Abstract

A network computer system selects a service provider for individual order requests by predicting an order preparation time of the respective supplier for the order request. During a time interval that precedes the order preparation time, the computer system matches an arrival time of a service provider to the respective supplier of an order request. The network computer system estimates an order delivery time for the requester based at least in part on the predicted order preparation time and on a location of the supplier relative to a location of the requester.

1 - 20 . (canceled) 
     
     
         21 . A computing system comprising:
 one or more processors; and   one or more memory resources to store a set of instructions that, when executed by the one or more processors, cause the computing system to develop a plurality of predictive models using a machine learning process, wherein the machine learning process comprises:
 obtaining first training information indicative of time durations for which service providers waited at a respective supplier for a delivery order; 
 training, based on the first training information, a first predictive model to predict a time characteristic for preparing an order by the respective supplier; 
 obtaining second training information indicative of active requesters that have launched a service application for a delivery service and conversion events indicating events where an order request was placed; 
 training, based on the second training information, a second predictive model to predict a number of order requests for a given time interval based on a number of active requesters that have launched the service application for the delivery service on an associated computing device and have yet not placed an order request; and 
 maintaining, in a database, the first and second predictive models. 
   
     
     
         22 . The computing system of  claim 21 , wherein the second predictive model is trained to predict at least one of: (i) the number of active requesters that may launch the service application in a given time interval, (ii) a conversion rate of active requesters, or (iii) a number of order requests which can potentially be generated from the number of active requesters. 
     
     
         23 . The computing system of  claim 21 , wherein the machine learning process further comprises:
 obtaining third training information indicative of a historical number of available service providers; and   training a third predictive model to predict a provisioning level determination, wherein the provisioning level determination reflects an adequacy of one or more available service providers for handling the number of order requests.   
     
     
         24 . The computing system of  claim 23 , wherein training the third predictive model further comprises:
 monitoring a requester status store;   comparing an output of the first predictive model with an actual outcome; and   tuning the first predictive model, based on comparing the output to the actual outcome.   
     
     
         25 . The computing system of  claim 24 , wherein the output of the third predictive model comprises at least one of: (i) a predicted conversion increase, (ii) a predicted number of order requests received, or (iii) a predicted number of available service providers. 
     
     
         26 . The computing system of  claim 21 , wherein maintaining, in the database, the first and second predictive models further comprises:
 obtaining historical data comprising one or more features related to a prior user session;   obtaining data associated with one or more current user sessions; and   detecting, based on comparing the data associated with the one or more current user sessions to the historical data at least one of (i) an active requester session event or (ii) a conversion event.   
     
     
         27 . The computing system of  claim 26 , wherein the active requester session event is indicative of a user initiating a session with the computing system by launching a service application on a requester device. 
     
     
         28 . The computing system of  claim 26 , wherein the conversion event is indicative of an active requester making an order request. 
     
     
         29 . A computer implemented method comprising:
 receiving, over one or more networks, a plurality of order requests, each order request originating from a corresponding requester device and specifying one or more items requested to be transported from a first supplier to a corresponding requester; and   for each of the plurality of order requests:
 computing, using a first machine learned model, a predicted order preparation time for the order request based on information for the corresponding supplier, the information being collected during a current order request session; 
 computing, using a second machine learned model, a predicted provisioning level determination indicative of an adequacy of one or more available service providers to service the order request; 
 determining, based on the predicted order preparation time and predicted provisioning level determination, a time to perform a matching process such that a service provider is estimated to arrive at the first supplier within a designated threshold of the predicted order preparation time for the order request; 
 communicating, based on the determined time, data indicative of the order request to the service provider; 
 tracking a location of the service provider; and 
 updating, based on tracked location of the service provider, a requester status store to include data indicative of one or more events. 
   
     
     
         30 . The method of  claim 29 , wherein computing, using the first machine learned model, the predicted order preparation time for the order request based on information for the first supplier comprises:
 obtaining real-time information data; and   predicting, based on the real-time information data, using the first machine learned model, an expected number of order requests for a first time period.   
     
     
         31 . The method of  claim 30 , wherein the real-time information data comprises at least one of a number of active requesters that have not made an order request, an amount of time each active requester has spent viewing a menu content for a respective supplier, or menus associated with respective suppliers that one or more users has viewed. 
     
     
         32 . The method of  claim 29 , wherein the data indicative of one or more events comprises data indicative of at least one of: (i) the order request being communicated to the corresponding supplier; (ii) the corresponding supplier acknowledging the order request; (iii) one or more progress indicators for the order request being prepared; (iv) a service provider picking up the corresponding order; or (v) arrival of the service provider at service location of a requester. 
     
     
         33 . The method of  claim 32 , wherein the one or more progress indicators for the order request being prepared comprises an indicator that a corresponding delivery order is ready. 
     
     
         34 . The method of  claim 32 , wherein the service location of the requester comprises at least one of: (i) a current location of the requester or (ii) a specified location of the requester. 
     
     
         35 . The method of  claim 29 , wherein determining the time to perform the matching process further comprises:
 obtaining historical information data;   associating the historical information data with one or more categories;   storing data indicative of the historical information data associated with the one or more categories in a data store; and   comparing the historical information data associated with the one or more categories in the data store with data indicative of a category associated with the order request.   
     
     
         36 . A computer implemented method comprising:
 receiving, over one or more networks, a plurality of order requests, each order request originating from a corresponding requester device and specifying one or more items requested to be transported from a first supplier to a corresponding requester; and   for each of the plurality of order requests:
 computing, using a first machine learned model, a predicted number of order requests for a given time interval based on a number of active requesters that have launched a service application for a delivery service on an associated computing device and have yet not placed an order request; 
 computing, using a second machine learned model, a predicted provisioning level determination indicative of an adequacy of one or more available service providers to service the order request; 
 determining, based on the predicted number of order requests for a given time interval and predicted provisioning level determination, a time to perform a matching process such that a service provider is estimated to arrive at the first supplier within a designated threshold of a predicted order preparation time for the order request; 
 communicating, at the determined time, data indicative of the order request to the service provider; 
 tracking a location of the service provider; and 
 updating, based on tracked location of the service provider, a requester status store to include data indicative of one or more events. 
   
     
     
         37 . The computer implemented method of  claim 36 , wherein the predicted provisioning level corresponds to a metric based on a comparison of: (i) an estimated number of order requests received for a first time period for a first region and (ii) a number of service providers in the first region that are likely to be available to provide delivery service for the number of order requests. 
     
     
         38 . The computer implemented method of  claim 36 , wherein the second machine learned model estimates a predicted provisioning level for at least one of: (i) a current time interval or (ii) one or more future time intervals. 
     
     
         39 . The computer implemented method of  claim 36 , wherein computing the predicted provisioning level determination comprises:
 obtaining, from the requester status store, real-time data comprising a number of active requesters who have not yet submitted an order request; and   determining, based on obtaining the real-time data, a predicted provisioning level determination.   
     
     
         40 . The method of  claim 36 , wherein determining the predicted provisioning level determination comprises:
 obtaining input parameters indicative of a velocity value corresponding to at least one of a number of order requests, a number of active order requests, or a number of converted requesters; and   determining based at least in part on the velocity value the predicted provisioning level.