WO2021179906A1 - 订单状态提示 - Google Patents

订单状态提示 Download PDF

Info

Publication number
WO2021179906A1
WO2021179906A1 PCT/CN2021/077486 CN2021077486W WO2021179906A1 WO 2021179906 A1 WO2021179906 A1 WO 2021179906A1 CN 2021077486 W CN2021077486 W CN 2021077486W WO 2021179906 A1 WO2021179906 A1 WO 2021179906A1
Authority
WO
WIPO (PCT)
Prior art keywords
order
target
probability
time period
estimation model
Prior art date
Application number
PCT/CN2021/077486
Other languages
English (en)
French (fr)
Inventor
杨迪昇
Original Assignee
北京三快在线科技有限公司
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 北京三快在线科技有限公司 filed Critical 北京三快在线科技有限公司
Publication of WO2021179906A1 publication Critical patent/WO2021179906A1/zh

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q30/00—Commerce
    • G06Q30/06—Buying, selling or leasing transactions
    • G06Q30/0601—Electronic shopping [e-shopping]
    • G06Q30/0633—Lists, e.g. purchase orders, compilation or processing
    • G06Q30/0635—Processing of requisition or of purchase orders
    • G06Q30/0637—Approvals
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q10/00—Administration; Management
    • G06Q10/04—Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q30/00—Commerce
    • G06Q30/06—Buying, selling or leasing transactions

Definitions

  • the present disclosure relates to the field of data processing, and in particular, to an order status prompt.
  • the purpose of the present disclosure is to provide an order status prompt, and the technical solution is as follows.
  • an order status prompt method which includes: inputting order feature information of a target order into a pre-trained target probability estimation model, and the target probability estimation model is used to output the order status from the current moment.
  • the target probability estimation model is used to output the order status from the current moment.
  • the probability of a delivery person taking orders there is the probability of a delivery person taking orders; the probability of taking orders for the target order is determined based on the output of the target probability estimation model; when the probability of taking orders is less than the reference probability threshold, generating In order to present the user's order operation recommendations.
  • the method further includes: determining the target time period elapsed by the target order from the moment the order is placed to the current moment; inputting the order feature information of the target order into a pre-trained target probability estimation model, It includes: determining the target probability prediction model from a plurality of pre-trained probability prediction models, the sample time period of the target probability model is the same as the target time period, wherein each of the probability prediction models It is used to output the probability that the order will be accepted by the dispatcher in the second reference time period from the moment of order placement, where the second reference time period is the sum of the target time period and the first reference time period; The order characteristic information of the target order is input into the target probability estimation model.
  • each of the probability estimation models is obtained by training from a sample order set, and the sample order sets include positive sample orders for which a delivery person takes orders during the sample time period corresponding to the probability estimation model , And a negative sample order without a delivery person taking an order within the sample time period, wherein the training weight of the negative sample order is higher than the positive sample order.
  • each of the probability estimation models is obtained through training of a sample training set, and the sample training set includes a positive sample order that has a delivery person taking an order within the first reference time period from the current moment, And a negative sample order without a delivery person taking an order in the first reference time period from the current moment, wherein the training weight of the negative sample order is higher than the positive sample order.
  • the generating an order operation suggestion for presentation to the user when the probability of receiving an order is less than a reference probability threshold includes: generating a service for canceling an order when the probability of receiving an order is less than the reference probability threshold Channels and recommendations for order cancellation actions presented to users.
  • the method further includes:
  • the target time period includes the target order The period of time elapsed from the time the order was placed to the current position.
  • the method further includes: determining the reference probability threshold value according to a loss budget value, wherein the sum of the loss values caused by the cancelled orders does not exceed the loss budget value.
  • said inputting the order characteristic information of the target order into a pre-trained target probability estimation model includes: determining whether the target order meets the condition of insufficient capacity according to the order characteristic information of the target order; When the order satisfies the insufficient capacity condition, enter the order feature information into the probability estimation model for the insufficient capacity situation and the probability estimation model for the ordinary situation; in the case that the target order does not meet the insufficient capacity condition , Inputting the order characteristic information into a probability estimation model of a common situation; the determining the probability of receiving the target order based on the output of the target probability estimation model includes: when the target order satisfies the insufficient capacity In the case of conditions, the first probability output by the probability prediction model of the insufficient capacity situation and the second probability output by the probability prediction model of the ordinary situation are obtained, and the first probability and the total probability are combined based on the reference combination rule.
  • the second probability combination is the order acceptance probability; in the case that the target order does not meet the insufficient capacity condition, the order acceptance probability output by the probability estimation model of
  • the method before determining the target probability estimation model from the plurality of pre-trained probability estimation models, the method further includes: determining that the target time period is greater than or equal to a first reference target time period; The method further includes: in the case that the target time period is greater than the second reference target time period, generating the order operation suggestion, where the second reference target time period is greater than the first reference target time period .
  • the method further includes: determining that the target order satisfies a reference prompt condition; wherein the reference prompt condition includes at least one of the following: said The estimated delivery time of the target order is less than the reference delivery time threshold; the ratio of the number of overtime orders in the delivery area of the target order to the number of dispatchers in the delivery area is less than the reference ratio threshold; the number of dispatchers who can deliver the target order Greater than the reference number threshold.
  • an order status prompting device which includes: a feature input module for inputting order feature information of a target order into a pre-trained target probability estimation model, and the target probability estimation model is used to output The probability that the order is accepted by the delivery person in the first reference time period from the current moment; the probability determination module is used to determine the probability of the target order based on the output of the target probability estimation model; the suggestion generation module , For generating an order operation suggestion for presenting to the user when the order acceptance probability is less than the reference probability threshold.
  • the device further includes: a time determination module, configured to determine the target time period that has passed from the moment the order is placed to the current moment; the feature input module, configured to obtain data from the pre-trained In one probability prediction model, the target probability prediction model is determined, and the sample time period of the target probability model is the same as the target time period, wherein each of the probability prediction models is used to output the order at the time of placing the order There is a probability that the delivery person will take the order in the second reference time period from the beginning, the second reference time period is the sum of the target time period and the first reference time period, and the order feature information of the target order Input the target probability estimation model.
  • a time determination module configured to determine the target time period that has passed from the moment the order is placed to the current moment
  • the feature input module configured to obtain data from the pre-trained In one probability prediction model, the target probability prediction model is determined, and the sample time period of the target probability model is the same as the target time period, wherein each of the probability prediction models is used to output the order at
  • each of the probability estimation models is obtained by training from a sample order set, and the sample order sets include positive sample orders for which a delivery person takes orders during the sample time period corresponding to the probability estimation model , And a negative sample order without a delivery person taking an order within the sample time period, wherein the training weight of the negative sample order is higher than the positive sample order.
  • each of the probability estimation models is obtained through training of a sample training set, and the sample training set includes a positive sample order that has a delivery person taking an order within the first reference time period from the current moment, And a negative sample order without a delivery person taking an order in the first reference time period from the current moment, wherein the training weight of the negative sample order is higher than the positive sample order.
  • the suggestion generation module is configured to generate a service channel for order cancellation and an order cancellation operation suggestion for presenting to the user when the order acceptance probability is less than a reference probability threshold.
  • the device further includes a prompt module, configured to determine the suggested prompt mode of the order operation according to the target time period, and the prompt mode includes prompt bar display, order cancellation entry display, pop-up prompt, and SMS prompt.
  • the target time period includes the time period that elapses from the time of placing the order to the current time of the target order.
  • the device further includes a threshold value determining module, configured to determine the reference probability threshold value according to the loss budget value, wherein the sum of the loss values caused by the cancelled orders does not exceed the loss budget value.
  • a threshold value determining module configured to determine the reference probability threshold value according to the loss budget value, wherein the sum of the loss values caused by the cancelled orders does not exceed the loss budget value.
  • the feature input module is configured to determine whether the target order meets the insufficient capacity condition according to the order feature information of the target order; when the target order meets the insufficient capacity condition, the The order feature information is entered into the probability prediction model of the case of insufficient capacity and the probability prediction model of the ordinary case; in the case that the target order does not meet the condition of the insufficient capacity, the probability of inputting the order feature information into the ordinary case is predicted Estimation model; the probability determination module is configured to obtain the first probability output by the probability estimation model of the insufficient capacity situation and the probability prediction of the ordinary situation when the target order meets the condition of insufficient capacity Estimate the second probability output by the model, and combine the first probability and the second probability into the order probability based on the reference combination rule; in the case that the target order does not meet the insufficient capacity condition, obtain The probability of receiving the order output by the probability estimation model of the common situation.
  • the device further includes a target determining module, configured to determine that the target time period is greater than or equal to a first reference target time period; the condition determining module is further configured to determine that the target time period is greater than the first reference target time period. 2. In the case of referring to the target time period, the order operation recommendation is generated, and the second reference target time period is greater than the first reference target time period.
  • the device further includes a condition determination module for determining that the target order meets a reference prompt condition; wherein the reference prompt condition includes at least one of the following: the estimated delivery time of the target order is less than the reference delivery time Threshold; the ratio of the number of overtime orders in the delivery area of the target order to the number of dispatchers in the delivery area is less than a reference ratio threshold; the number of dispatchers who can deliver the target order is greater than the reference number threshold.
  • the reference prompt condition includes at least one of the following: the estimated delivery time of the target order is less than the reference delivery time Threshold; the ratio of the number of overtime orders in the delivery area of the target order to the number of dispatchers in the delivery area is less than a reference ratio threshold; the number of dispatchers who can deliver the target order is greater than the reference number threshold.
  • a computer-readable storage medium on which a computer program is stored, and when the program is executed by a processor, the steps of any optional order status prompt method of the present disclosure are realized.
  • an electronic device including: a memory on which a computer program is stored; a processor, configured to execute the computer program in the memory, so as to implement any optional order of the present disclosure The steps of the status notification method.
  • the order feature information of the order can be input into the pre-trained probability estimation model. Based on the output of the probability estimation model, the probability that the order will be accepted by the dispatcher is determined, and the order is received When the single probability is less than the reference probability threshold, an order operation suggestion is generated. In this way, it can be judged by the characteristics of the order whether there is a high probability that the order is not taken, and the judgment result and the corresponding operation suggestion are presented to the user, so that the user The status of the order is more clear, so that the order can be handled reasonably.
  • the user can cancel the order with a lower probability of receiving an order, continue to wait for the order with a higher probability of receiving an order, and balance the dispatching resources of the takeaway platform for orders, and reduce Loss of merchants and platforms, and can improve user experience.
  • Fig. 1 is a flowchart showing a method for prompting an order status according to an exemplary disclosed embodiment.
  • Fig. 2 is a flow chart showing a method for prompting an order status according to an exemplary disclosed embodiment.
  • Fig. 3 is a block diagram showing a device for prompting an order status according to an exemplary disclosed embodiment.
  • Fig. 4 is a block diagram showing an electronic device according to an exemplary disclosed embodiment.
  • the takeaway platform publishes the order-related delivery requirements to multiple delivery staff in the same area.
  • the delivery staff who receives the delivery demand chooses whether to accept the delivery of the order according to their own conditions.
  • the user obtains the product from the merchant and delivers it to the user's designated address.
  • not all orders will be received by a delivery person.
  • Some orders face the situation that no delivery person can take the order, or some orders face the situation that there is no delivery person to take the order in a short time. For example, in some special circumstances that are not conducive to delivery, such as late hours, long journeys, poor weather, etc., after the user places an order, there may be no delivery staff to take the order.
  • the platform also needs to provide comfort compensation for the loss of the merchant or user; and if the order is received by the delivery staff at any time, if the user chooses to continue to wait at 20 minutes, it will be There is still no one to take orders within the waiting time, which will affect the user experience, and long-term orders will also lead to the problem of reduced platform scheduling efficiency.
  • the present disclosure can estimate the probability of receiving the order by the delivery staff for the user, and generate operation suggestions for the user according to the probability of receiving the order, thereby solving the above-mentioned problems.
  • Fig. 1 is a flowchart showing a method for prompting an order status according to an exemplary disclosed embodiment. As shown in Figure 1, the method may include the following steps:
  • Order characteristic information refers to the information that can reflect the objective characteristics of the order. These objective characteristics will affect the delivery staff's subjective willingness to accept the order to a certain extent, thereby affecting the probability of the target order being accepted. For example, when the location of the merchant is far away from the destination of the target order, the delivery staff may be unwilling to take the order, resulting in a lower probability of receiving orders with the location of the merchant far away from the destination; when the target order is placed on the weekend When the number of orders on weekends is relatively large, and the number of delivery personnel has decreased, the average waiting time for receiving orders on weekends will increase, thereby reducing the possibility of the target order being received within a certain time; when the target order corresponds When the meal time of the merchants is longer, the waiting time of the delivery staff in the store will also increase, which causes the delivery staff to be reluctant to accept orders from the merchants, thereby reducing the probability of the target order being received.
  • Order feature information can include one or more of the following information: business feature information, statistical feature information, real-time feature information, and merchant feature information.
  • business feature information can include: delivery distance, delivery amount, estimated delivery time, delivery Information such as fees, order time, etc., this type of feature information varies greatly for different orders;
  • statistical feature information can include the average pick-up time in the region, the average delivery time in the region, etc.
  • the feature information of the dimension is the same and unchanged;
  • the real-time feature information can include: the number of uncollected orders in the region, the number of undelivered orders in the region, the number of unaccepted orders in the region, the number of orders placed within 10 minutes in the region, etc., these feature information It will change in real time over time, but for orders in the same area, the characteristic information of this dimension is the same; the characteristic information of merchants can include: average merchant picking speed, average merchant waiting time at the store, etc.
  • the order characteristic information of the target order can be collected and input into the pre-trained target probability estimation model.
  • the target probability estimation model is obtained by training on a sample training set.
  • the sample training set includes positive sample orders that have been received by a delivery person in the first reference time period from the current moment, and the first reference from the current moment Negative sample orders without a delivery person taking orders during the time period, where the training weight of the negative sample orders is higher than that of the positive sample orders.
  • the sample training set is also called the sample order set.
  • the target probability estimation model is obtained through the training of the sample training set.
  • the sample training set includes the positive sample orders received by the delivery person in the first reference time period from the statistical time, and the first reference time from the statistical time Negative sample orders without a delivery person (including orders received after the first reference time period, cancellation within the first reference time period, and cancellation after the first reference time period), each sample order includes a sample The order feature information and sample order time of the order, where the statistical time corresponds to the "current time" of the target order, that is, the length of time between the order time of the target order and the current time is the same as the time of the sample order. The duration from a single time to the statistical time is the same.
  • the target probability estimation model used to output the probability of the order is started from the statistical time 15 minutes after the time of placing the order of the sample order, and there is delivery within the first reference time period. “Positive sample orders received by the clerk” and “Negative sample orders not received by the delivery clerk within the first reference time period starting from the statistical time 15 minutes after the time of placing the sample order”, if the first reference The time period is 10 minutes, then the sample training set is a positive sample order with a delivery person taking the order within 15 to 25 minutes from the order placed, and a negative sample order without a delivery person taking the order within 15 to 25 minutes from the order placement , Orders received by a delivery person within 15 minutes or cancelled within 15 minutes are not included in the sample training set of the target delivery model.
  • the target probability estimation model needs to be updated regularly, and time-sensitive historical orders are used as the sample training set. Moreover, because in the actual delivery business, the number of orders received by a delivery person within a short period of time (for example, within 20 minutes or 30 minutes) is much higher than the number of orders received by a delivery person within a short period of time. Therefore, when collecting the sample training set, the number of positive sample orders is also higher than the number of negative sample orders, and during model training, when the samples are not balanced, the loss function will give priority to improving the accuracy of the larger number. , So that the accuracy of the smaller number decreases.
  • the present disclosure uses the weighted cross-entropy loss function for training during model training, so as to achieve the purpose of weighted training for fewer categories, and balance positive samples and negative samples The accuracy rate.
  • Each order feature information is output by the target probability estimation model as a label value, and the label value represents the degree of influence of the order feature information on the order probability.
  • the target probability estimation model can count all the label values and perform weighting processing to obtain the final label value.
  • the final label value indicates that the target order is received by the dispatcher in the first reference time period from the current moment. Probability.
  • the target probability estimation model under the condition of insufficient capacity is different from the target probability estimation model under other conditions. It is trained from the sample training set under the condition of insufficient capacity. All sample orders are historical orders under the condition of insufficient capacity. Before inputting the order characteristics into the target probability estimation model, according to the order characteristic information of the target order, it is determined whether the target order meets the insufficient capacity condition. If the target order meets the insufficient capacity condition, the The order feature information is entered into the probability prediction model of the case of insufficient capacity and the probability prediction model of the ordinary case. In the case that the target order does not meet the condition of the insufficient capacity, the probability of inputting the order feature information into the ordinary case is predicted Estimate the model.
  • the target order meets the insufficient capacity condition, obtain the first probability and the normal situation output by the probability estimation model of the insufficient capacity situation
  • the probability predicts the second probability output by the model, and combines the first probability and the second probability into the order probability based on a reference combination rule.
  • condition of insufficient capacity may include: the ten-minute order acceptance rate in the delivery area at the current moment is less than the reference order acceptance rate threshold, and the order acceptance rate is the total number of orders that have been received by a delivery person among orders created within 10 minutes in the area , The ratio of the difference between the total number of orders created in 10 minutes in this area and the total number of cancelled orders in this area.
  • S12 Determine the order probability of the target order based on the output of the target probability estimation model.
  • the order probability of the target order can be determined according to the label value. For example, if the tag value is 0.78, it can be determined that the order probability of the target order is 78%.
  • the probability of receiving an order refers to the "probability of a delivery person receiving an order within the first reference time period from the current time of the target order".
  • the order probability can be presented to the user in the form of a reminder message, and the user can determine what action should be taken on the order; it can also directly generate recommended order operation suggestions based on the order probability. And present it to the user.
  • the order probability When the order probability is less than the reference probability threshold, it indicates that the target order has a low probability of being received by the delivery person in the first reference time period from the current moment, and an order operation suggestion that recommends the user to cancel the order can be generated, for example, , You can generate a prompt message "The current order has not been received by the rider, you can choose to cancel the order" and send it to the user in the form of a pop-up window.
  • the prompt manner of the prompt information can be determined according to the elapsed time after the user places the order, and the prompt strength of the prompt manner is positively correlated with the elapsed time after the user places the order. For example, when a user places an order for 10 minutes, the prompt message can be displayed in the form of a permanent reminder in the app. When the user places an order for 20 minutes, it can be displayed in the app by opening the user’s disclaimer and cancellation portal. The prompt information, when the user places an order for 30 minutes, the prompt information can be displayed in a combination of a notification pop-up window and a short message.
  • the target characteristic information that has the highest degree of influence on the order acceptance probability in the order characteristic information may be determined, and prompt information may be generated according to the target characteristic information.
  • prompt information may be generated according to the target characteristic information.
  • the target order meets the reference prompt condition: the estimated delivery time of the target order is less than the reference delivery time threshold, or the delivery of the target order
  • the ratio of the number of overtime orders in the area to the number of dispatchers in the delivery area is less than the reference ratio threshold, or the number of dispatchers who can deliver the target order is greater than the reference number threshold.
  • the reference probability threshold can be determined in conjunction with the loss budget value, and the sum of the loss value caused by the cancelled order Do not exceed the loss budget value.
  • the order feature information of the order can be input into the pre-trained probability estimation model. Based on the output of the probability estimation model, the probability that the order will be accepted by the dispatcher is determined, and the order is received When the single probability is less than the reference probability threshold, an order operation suggestion is generated. In this way, it can be judged by the characteristics of the order whether there is a high probability that the order is not taken, and the judgment result and the corresponding operation suggestion are presented to the user, so that the user The status of the order is more clear, so that the order can be handled reasonably.
  • the user can cancel the order with a lower probability of receiving an order, continue to wait for the order with a higher probability of receiving an order, and balance the dispatching resources of the takeaway platform for orders, and reduce Loss of merchants and platforms, and can improve user experience.
  • Fig. 2 is a flow chart showing a method for prompting an order status according to an exemplary disclosed embodiment. As shown in Figure 2, the method may include the following steps:
  • S21 Determine the target time period that elapses from the time of placing the order to the current time of the target order.
  • the target time period is 16 minutes.
  • S22 Determine the relationship between the target time period and the first reference target time period and the second reference target time period, where the second reference target time period is greater than the first reference target time period.
  • step S21 If the target time period is less than the first reference target time period, return to step S21; if the target time period is greater than or equal to the first reference target time period, and is less than or equal to the second reference target time period, then perform step S23: In the case that the target time period is greater than the second reference target time period, an order operation recommendation is directly generated.
  • the target probability estimation model is used to judge the order probability of the target order, and it is less than the first reference target time period. 1.
  • the order placement time is relatively short, and the possibility of receiving the order is high. There is no need to calculate the probability of receiving orders for this type of order; when the target time period exceeds the second reference target time period, The waiting time after the user places an order has been too long, and most users have no patience to wait for the delivery person to take the order. At this time, they can directly put forward order operation suggestions to the user and guide the user to perform appropriate operations on the order.
  • each of the probability estimation models is used to output the probability that the order will be received by the delivery person in a second reference time period from the moment of placing the order, and the second reference time period is the target time period and the first reference time period. 1. The sum of reference time periods.
  • the accuracy of the target time period is one minute, that is to say, when the target time period is between the first reference target time period and the second reference target time period, a new target probability estimate needs to be determined after each minute has passed Model, if the first reference target time period is 10 minutes, and the second reference target time period is 20 minutes, then the 10th, 11th, 12th, 13th, 14th, and 14th minutes from the moment of order There is a reference target probability estimation model for 15 minutes, 16 minutes, 17 minutes, 18 minutes, 19 minutes, and 20 minutes.
  • the probability estimation model is used to output the 20th minute (the sum of the first reference time period 10 minutes and the target time period 10 minutes), the 21st minute, the 22nd minute, the 23rd minute, the 24th minute, and the 20th minute from the time the order was placed. There is a probability that the delivery person will take the order in the 25th, 26th, 27th, 28th, 29th, and 30th minutes.
  • the target probability estimation model is obtained through the training of the sample training set.
  • the sample training set includes the positive sample orders received by the delivery person in the first reference time period from the statistical time, and the first reference time from the statistical time Negative sample orders without a delivery person (including orders received after the first reference time period, cancellation within the first reference time period, and cancellation after the first reference time period), each sample order includes a sample The order feature information and sample order time of the order, where the statistical time corresponds to the "current time" of the target order, that is, the length of time between the order time of the target order and the current time is the same as the time of the sample order. The duration from a single time to the statistical time is the same.
  • the target probability estimation model used to output the probability of the order is started from the statistical time 15 minutes after the time of placing the order of the sample order, and there is delivery within the first reference time period. “Positive sample orders received by the clerk” and “Negative sample orders not received by the delivery clerk within the first reference time period starting from the statistical time 15 minutes after the time of placing the sample order”, if the first reference The time period is 10 minutes, then the sample training set is a positive sample order with a delivery person taking the order within 15 to 25 minutes from the order placed, and a negative sample order without a delivery person taking the order within 15 to 25 minutes from the order placement , Orders received by a delivery person within 15 minutes or cancelled within 15 minutes are not included in the sample training set of the target delivery model.
  • the target probability estimation model needs to be updated regularly, and time-sensitive historical orders are used as the sample training set. Moreover, because in the actual delivery business, the number of orders received by a delivery person within a short period of time (for example, within 20 minutes or 30 minutes) is much higher than the number of orders received by a delivery person within a short period of time. Therefore, when collecting the sample training set, the number of positive sample orders is also higher than the number of negative sample orders, and during model training, when the samples are not balanced, the loss function will give priority to improving the accuracy of the larger number. , So that the accuracy of the smaller number decreases.
  • the present disclosure uses the weighted cross-entropy loss function for training during model training, so as to achieve the purpose of weighted training for fewer categories, and balance positive samples and negative samples The accuracy rate.
  • Order characteristic information refers to the information that can reflect the objective characteristics of the order. These objective characteristics will affect the delivery staff's subjective willingness to accept the order to a certain extent, thereby affecting the probability of the target order being accepted. For example, when the location of the merchant is far away from the destination of the target order, the delivery staff may be unwilling to take the order, resulting in a lower probability of receiving orders with the location of the merchant far away from the destination; when the target order is placed on the weekend When the number of orders on weekends is relatively large, and the number of delivery personnel has decreased, the average waiting time for receiving orders on weekends will increase, thereby reducing the possibility of the target order being received within a certain time; when the target order corresponds When the meal time of the merchants is longer, the waiting time of the delivery staff in the store will also increase, which causes the delivery staff to be reluctant to accept orders from the merchants, thereby reducing the probability of the target order being received.
  • Order feature information can include one or more of the following information: business feature information, statistical feature information, real-time feature information, and merchant feature information.
  • business feature information can include: delivery distance, delivery amount, estimated delivery time, delivery Information such as fees, order time, etc., this type of feature information varies greatly for different orders;
  • statistical feature information can include the average pick-up time in the region, the average delivery time in the region, etc.
  • the feature information of the dimension is the same and unchanged;
  • the real-time feature information can include: the number of uncollected orders in the region, the number of undelivered orders in the region, the number of unaccepted orders in the region, the number of orders placed within 10 minutes in the region, etc., these feature information It will change in real time over time, but for orders in the same area, the characteristic information of this dimension is the same; the characteristic information of merchants can include: average merchant picking speed, average merchant waiting time at the store, etc.
  • the order characteristic information of the target order can be collected and input into the pre-trained target probability estimation model.
  • Each order feature information is output by the target probability estimation model as a label value, and the label value represents the degree of influence of the order feature information on the order probability.
  • the target probability estimation model can count all the label values and perform weighting processing to obtain the final label value.
  • the final label value indicates that the target order is received by the dispatcher in the first reference time period from the current moment. Probability.
  • the target probability estimation model under the condition of insufficient capacity is different from the target probability estimation model under other conditions. It is trained from the sample training set under the condition of insufficient capacity. All sample orders are historical orders under the condition of insufficient capacity. Before inputting the order characteristics into the target probability estimation model, according to the order characteristic information of the target order, it is determined whether the target order meets the insufficient capacity condition. If the target order meets the insufficient capacity condition, the The order feature information is entered into the probability prediction model of the case of insufficient capacity and the probability prediction model of the ordinary case. In the case that the target order does not meet the condition of the insufficient capacity, the probability of inputting the order feature information into the ordinary case is predicted Estimate the model.
  • the target order meets the insufficient capacity condition, obtain the first probability and the normal situation output by the probability estimation model of the insufficient capacity situation
  • the probability predicts the second probability output by the model, and combines the first probability and the second probability into the order probability based on a reference combination rule.
  • condition of insufficient capacity may include: the ten-minute order acceptance rate in the delivery area at the current moment is less than the reference order acceptance rate threshold, and the order acceptance rate is the total number of orders that have been received by a delivery person among orders created within 10 minutes in the area , The ratio of the difference between the total number of orders created in 10 minutes in this area and the total number of cancelled orders in this area.
  • the order probability of the target order can be determined according to the label value. For example, if the tag value is 0.78, it can be determined that the order probability of the target order is 78%.
  • the probability of receiving an order refers to the "probability of a delivery person receiving an order within the first reference time period from the current time of the target order".
  • the order probability can be presented to the user in the form of a reminder message, and the user can determine what action should be taken on the order; it can also directly generate recommended order operation suggestions based on the order probability. And present it to the user.
  • the order probability When the order probability is less than the reference probability threshold, it indicates that the target order has a low probability of being received by the delivery person in the first reference time period from the current moment, and an order operation suggestion that recommends the user to cancel the order can be generated, for example, , You can generate a prompt message "The current order has not been received by the rider, you can choose to cancel the order" and send it to the user in the form of a pop-up window.
  • the target characteristic information that has the highest degree of influence on the order acceptance probability in the order characteristic information may be determined, and prompt information may be generated according to the target characteristic information.
  • prompt information may be generated according to the target characteristic information.
  • the prompt manner of the prompt information may be determined according to the target time period, and the prompt strength of the prompt manner is positively correlated with the target time period. For example, when the user places an order for 10 minutes (that is, when the target time period is 10 minutes), the prompt message can be displayed in the form of a resident prompt bar in the application. When the user places an order for 20 minutes, the prompt information can be displayed in the application. The prompt information is displayed in the form of opening the user exemption cancellation portal. When the user places an order for 30 minutes, the prompt information can be displayed in a combination of a notification pop-up window and a short message.
  • the target order meets the reference prompt condition: the estimated delivery time of the target order is less than the reference delivery time threshold, or the delivery of the target order
  • the ratio of the number of overtime orders in the area to the number of dispatchers in the delivery area is less than the reference ratio threshold, or the number of dispatchers who can deliver the target order is greater than the reference number threshold.
  • the reference probability threshold can be determined by the following formula combined with the loss budget value: Among them, St.budget is the total loss budget value, and cost(x i ) is the loss budget value of the i-th model.
  • the order characteristic information of the order whose target time period elapsed from the time of the order to the current time is located between the first reference target time period and the second reference target time period Input the pre-trained probability estimation model, determine the order probability that the delivery person takes the order based on the output of the probability estimation model, and generate order operation suggestions when the order probability is less than the reference probability threshold, so that you can save calculations
  • the characteristics of the order are used to determine whether the possibility of unmanned orders is high, and the judgment results and corresponding operation suggestions are presented to the user, so that the user is more clear about the status of the order , So as to make reasonable operation and processing of orders, so that users can cancel orders with a low probability of taking orders, continue to wait for orders with a high probability of taking orders, balance the dispatching resources of the takeaway platform for orders, and reduce the loss of merchants and platforms. And can improve the user experience.
  • Fig. 3 is a block diagram showing a device for prompting an order status according to an exemplary disclosed embodiment.
  • the order status prompt device 300 includes:
  • the feature input module 301 is used to input the order feature information of the target order into a pre-trained target probability estimation model, and the target probability estimation model is used to output that the order has a delivery person in the first reference time period from the current moment. Probability of taking orders.
  • the probability determination module 302 is configured to determine the order probability of the target order based on the output of the target probability estimation model.
  • the suggestion generation module 303 is configured to generate an order operation suggestion for presenting to the user when the order acceptance probability is less than the reference probability threshold.
  • the device further includes: a time determination module, configured to determine the target time period that has passed from the moment the order is placed to the current moment; the feature input module, configured to obtain data from the pre-trained In one probability prediction model, the target probability prediction model is determined, and the sample time period of the target probability model is the same as the target time period, wherein each of the probability prediction models is used to output the order at the time of placing the order There is a probability that the delivery person will take the order in the second reference time period from the beginning, the second reference time period is the sum of the target time period and the first reference time period, and the order feature information of the target order Input the target probability estimation model.
  • a time determination module configured to determine the target time period that has passed from the moment the order is placed to the current moment
  • the feature input module configured to obtain data from the pre-trained In one probability prediction model, the target probability prediction model is determined, and the sample time period of the target probability model is the same as the target time period, wherein each of the probability prediction models is used to output the order at
  • each of the probability estimation models is obtained by training from a sample order set, and the sample order sets include positive sample orders for which a delivery person takes orders during the sample time period corresponding to the probability estimation model , And a negative sample order without a delivery person taking an order within the sample time period, wherein the training weight of the negative sample order is higher than the positive sample order.
  • each of the probability estimation models is obtained through training of a sample training set, and the sample training set includes a positive sample order that has a delivery person taking an order within the first reference time period from the current moment, And a negative sample order without a delivery person taking an order in the first reference time period from the current moment, wherein the training weight of the negative sample order is higher than the positive sample order.
  • the suggestion generation module is configured to generate a service channel for order cancellation and an order cancellation operation suggestion for presenting to the user when the order acceptance probability is less than a reference probability threshold.
  • the device further includes a prompt module, configured to determine the suggested prompt mode of the order operation according to the target time period, and the prompt mode includes prompt bar display, order cancellation entry display, pop-up prompt, and SMS prompt.
  • the target time period includes the time period that elapses from the time of placing the order to the current time of the target order.
  • the device further includes a threshold value determining module, configured to determine the reference probability threshold value according to the loss budget value, wherein the sum of the loss values caused by the cancelled orders does not exceed the loss budget value.
  • a threshold value determining module configured to determine the reference probability threshold value according to the loss budget value, wherein the sum of the loss values caused by the cancelled orders does not exceed the loss budget value.
  • the feature input module is configured to determine whether the target order meets the insufficient capacity condition according to the order feature information of the target order; when the target order meets the insufficient capacity condition, the The order feature information is entered into the probability prediction model of the case of insufficient capacity and the probability prediction model of the ordinary case; in the case that the target order does not meet the condition of the insufficient capacity, the probability of inputting the order feature information into the ordinary case is predicted Estimation model; the probability determination module is configured to obtain the first probability output by the probability estimation model of the insufficient capacity situation and the probability prediction of the ordinary situation when the target order meets the condition of insufficient capacity Estimate the second probability output by the model, and combine the first probability and the second probability into the order probability based on the reference combination rule; in the case that the target order does not meet the insufficient capacity condition, obtain The probability of receiving the order output by the probability estimation model of the common situation.
  • the device further includes a target determining module, configured to determine that the target time period is greater than or equal to a first reference target time period; the condition determining module is further configured to determine that the target time period is greater than the first reference target time period. 2. In the case of referring to the target time period, the order operation suggestion is generated, and the second reference target time period is greater than the first reference target time period.
  • the device further includes a condition determination module for determining that the target order meets a reference prompt condition; wherein the reference prompt condition includes at least one of the following: the estimated delivery time of the target order is less than the reference delivery time Threshold; the ratio of the number of overtime orders in the delivery area of the target order to the number of dispatchers in the delivery area is less than a reference ratio threshold; the number of dispatchers who can deliver the target order is greater than the reference number threshold.
  • the reference prompt condition includes at least one of the following: the estimated delivery time of the target order is less than the reference delivery time Threshold; the ratio of the number of overtime orders in the delivery area of the target order to the number of dispatchers in the delivery area is less than a reference ratio threshold; the number of dispatchers who can deliver the target order is greater than the reference number threshold.
  • the order feature information of the order can be input into the pre-trained probability estimation model, and based on the output of the probability estimation model, the probability of receiving the order with a delivery person is determined, and the probability of receiving the order is less than the reference probability Order operation suggestions are generated when the threshold is set. In this way, it is possible to judge whether the order is unacceptable by the characteristics of the order, and the judgment result and corresponding operation suggestions are presented to the user, so that the user can better understand the status of the order. Clearly, so as to make reasonable handling of orders. In this way, users can cancel orders with low probability of receiving orders, continue to wait for orders with high probability of receiving orders, balance the dispatching resources of the takeaway platform for orders, and reduce the loss of merchants and platforms , And can improve the user experience.
  • Fig. 4 is a block diagram showing an electronic device 400 according to an exemplary embodiment.
  • the electronic device 400 can be provided as a server.
  • the electronic device 400 may include: a processor 401 and a memory 402.
  • the electronic device 400 may further include one or more of a multimedia component 403, an input/output (I/O) interface 404, and a communication component 405.
  • the processor 401 is used to control the overall operation of the electronic device 400 to complete all or part of the steps in the above-mentioned order status prompt method.
  • the memory 402 is used to store various types of data to support operations on the electronic device 400.
  • the data may include, for example, instructions for any application or method to operate on the electronic device 400, and application-related data.
  • the memory 402 can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (Static Random Access Memory, SRAM for short), electrically erasable programmable read-only memory ( Electrically Erasable Programmable Read-Only Memory (EEPROM), Erasable Programmable Read-Only Memory (EPROM), Programmable Read-Only Memory (PROM), Read-only Memory (Read-Only Memory, ROM for short), magnetic memory, flash memory, magnetic disk or optical disk.
  • the multimedia component 403 may include a screen and an audio component.
  • the screen may be a touch screen, for example, and the audio component is used to output and/or input audio signals.
  • the audio component may include a microphone, which is used to receive external audio signals.
  • the received audio signal can be further stored in the memory 402 or sent through the communication component 405.
  • the audio component also includes at least one speaker for outputting audio signals.
  • the I/O interface 404 provides an interface between the processor 401 and other interface modules.
  • the above-mentioned other interface modules may be a keyboard, a mouse, a button, and the like. These buttons can be virtual buttons or physical buttons.
  • the communication component 405 is used for wired or wireless communication between the electronic device 400 and other devices.
  • Wireless communication such as Wi-Fi, Bluetooth, Near Field Communication (NFC), 2G, 3G, 4G, NB-IOT, eMTC, or other 5G, etc., or one or more of them
  • the corresponding communication component 405 may include: a Wi-Fi module, a Bluetooth module, an NFC module, and so on.
  • the electronic device 400 may be used by one or more application specific integrated circuits (Application Specific Integrated Circuit, ASIC for short), digital signal processor (Digital Signal Processor, DSP for short), and digital signal processing equipment (Digital Signal Processor for short).
  • ASIC Application Specific Integrated Circuit
  • DSP Digital Signal Processor
  • DSP Digital Signal Processor
  • Digital Signal Processor Digital Signal Processor for short
  • Signal Processing Device DSPD for short
  • PLD Programmable Logic Device
  • FPGA Field Programmable Gate Array
  • controller microcontroller, microprocessor or other electronic components Realization, used to implement the above-mentioned order status prompt method.
  • a computer-readable storage medium including program instructions that, when executed by a processor, implement the steps of the above-mentioned order status prompt method.
  • the computer-readable storage medium may be the foregoing memory 402 including program instructions, and the foregoing program instructions may be executed by the processor 401 of the electronic device 400 to complete the foregoing order status prompt method.
  • a computer program product is further provided.
  • the computer program product includes a computer program that can be executed by a programmable device.
  • the code part of the order status notification method.

Landscapes

  • Business, Economics & Management (AREA)
  • Engineering & Computer Science (AREA)
  • Strategic Management (AREA)
  • Accounting & Taxation (AREA)
  • Finance (AREA)
  • Economics (AREA)
  • General Business, Economics & Management (AREA)
  • Physics & Mathematics (AREA)
  • Marketing (AREA)
  • General Physics & Mathematics (AREA)
  • Development Economics (AREA)
  • Theoretical Computer Science (AREA)
  • Human Resources & Organizations (AREA)
  • Game Theory and Decision Science (AREA)
  • Entrepreneurship & Innovation (AREA)
  • Operations Research (AREA)
  • Quality & Reliability (AREA)
  • Tourism & Hospitality (AREA)
  • Management, Administration, Business Operations System, And Electronic Commerce (AREA)

Abstract

本公开涉及一种订单状态提示，包括：将目标订单的订单特征信息输入预训练的目标概率预估模型，所述目标概率预估模型用于输出订单在从当前时刻起的第一参考时间段内有配送员接单的概率；基于所述目标概率预估模型的输出确定所述目标订单的接单概率；在所述接单概率小于参考概率阈值的情况下，生成用于呈现给用户的订单操作建议。

Description

订单状态提示
本公开要求于2020年03月13日提交的申请号为202010176669.1、申请名称为“订单状态提示方法和装置、存储介质和电子设备”的中国专利申请的优先权，其全部内容通过引用结合在本公开中。
技术领域
本公开涉及数据处理领域，具体地，涉及一种订单状态提示。
背景技术
随着网络技术的发展，外卖行业也逐渐发展壮大，用户可以从网上以外卖的方式购买各种商品，如餐品、百货、药品、鲜花等，通过外卖平台购买商品的购物形式也深入了当代人的生活之中。网上订餐让用户可以足不出户购买到各种商品，极大地增加了人们的生活便利度。
发明内容
本公开的目的是提供一种订单状态提示，技术方案如下。
本公开的一方面，提供一种订单状态提示方法，包括：将目标订单的订单特征信息输入预训练的目标概率预估模型，所述目标概率预估模型用于输出订单在从当前时刻起的第一参考时间段内有配送员接单的概率；基于所述目标概率预估模型的输出确定所述目标订单的接单概率；在所述接单概率小于参考概率阈值的情况下，生成用于呈现给用户的订单操作建议。
可选地，所述方法还包括：确定所述目标订单从下单时刻起至当前时刻为止所经过的目标时间段；所述将目标订单的订单特征信息输入预训练的目标概率预估模型，包括：从预训练的多个概率预估模型中，确定所述目标概率预估模型，所述目标概率模型的样本时间段与所述目标时间段相同，其中，每一所述概率预估模型用于输出订单在下单时刻起的第二参考时间段内有配送员接单的概率，所述第二参考时间段为所述目标时间段与所述第一参考时间段之和；将所述目标订单的订单特征信息输入所述目标概率预估模型。
可选地，每一所述概率预估模型是由样本订单集合训练得到，所述样本订单集合包括在所述概率预估模型对应的所述样本时间段内有配送员接单的正样本订单，及在所述样本时间段内无配送员接单的负样本订单，其中，所述负样本订单的训练权重高于所述正样本订单。
可选地，每一所述概率预估模型由样本训练集训练得到，所述样本训练集包括从所述当前时刻起的所述第一参考时间段内有配送员接单的正样本订单，及从所述当前时刻起的所述第一参考时间段内无配送员接单的负样本订单，其中，所述负样本订单的训练权重高于所述正样本订单。
可选地，所述在所述接单概率小于参考概率阈值，生成用于呈现给用户的订单操作建议包括：在所述接单概率小于参考概率阈值的情况下，生成用于取消订单的服务通道以及用于呈现给用户的订单取消操作建议。
可选地，在所述生成用于呈现给用户的订单操作建议之后，所述方法还包括：
根据目标时间段确定所述订单操作建议的提示方式，所述提示方式包括提示条展示、订单取消入口展示、弹窗提示、短信提示中的至少一种，所述目标时间段包括所述目标订单从下单时刻至当前时刻位置所经过的时间段。
可选地，所述方法还包括：根据损失预算值确定所述参考概率阈值，其中，被取消的订单造成的损失值之和不超过所述损失预算值。
可选地，所述将目标订单的订单特征信息输入预训练的目标概率预估模型，包括：根据所述目标订单的订单特征信息，确定所述目标订单是否满足运力不足条件；在所述目标订单满足所述运力不足条件的情况下，将所述订单特征信息输入运力不足情形的概率预估模型和普通情形的概率预估模型；在所述目标订单不满足所述运力不足条件的情况下，将所述订单特征信息输入普通情形的概率预估模型；所述基于所述目标概率预估模型的输出确定所述目标订单的接单概率，包括：在所述目标订单满足所述运力不足条件的情况下，获取所述运力不足情形的概率预估模型输出的第一概率和所述普通情形的概率预估模型输出的第二概率，并基于参考结合规则将所述第一概率和所述第二概率结合为所述接单概率；在所述目标订单不满足所述运力不足条件的情况下，获取所述普通情形的概率预估模型输出的所述接单概率。
可选地，所述从预训练的多个概率预估模型中，确定所述目标概率预估模型之前，所述方法还包括：确定所述目标时间段大于或等于第一参考目标时间 段；所述方法还包括：在所述目标时间段大于所述第二参考目标时间段的情况下，则生成所述订单操作建议，所述第二参考目标时间段大于所述第一参考目标时间段。
可选地，在所述生成用于呈现给用户的订单操作建议之前，所述方法还包括：确定所述目标订单满足参考提示条件；其中，所述参考提示条件包括以下至少一种：所述目标订单的预计配送时间小于参考配送时间阈值；所述目标订单的配送区域中的超时订单数目与所述配送区域中的配送员数目比值小于参考比例阈值；可配送所述目标订单的配送员数目大于参考数目阈值。
本公开的一方面，公开了一种订单状态提示装置，包括：特征输入模块，用于将目标订单的订单特征信息输入预训练的目标概率预估模型，所述目标概率预估模型用于输出订单在从当前时刻起的第一参考时间段内有配送员接单的概率；概率确定模块，用于基于所述目标概率预估模型的输出确定所述目标订单的接单概率；建议生成模块，用于在所述接单概率小于参考概率阈值的情况下，生成用于呈现给用户的订单操作建议。
可选地，所述装置还包括：时间确定模块，用于确定所述目标订单从下单时刻起至当前时刻为止所经过的目标时间段；所述特征输入模块，用于从预训练的多个概率预估模型中，确定所述目标概率预估模型，所述目标概率模型的样本时间段与所述目标时间段相同，其中，每一所述概率预估模型用于输出订单在下单时刻起的第二参考时间段内有配送员接单的概率，所述第二参考时间段为所述目标时间段与所述第一参考时间段之和，并将所述目标订单的订单特征信息输入所述目标概率预估模型。
可选地，每一所述概率预估模型是由样本订单集合训练得到，所述样本订单集合包括在所述概率预估模型对应的所述样本时间段内有配送员接单的正样本订单，及在所述样本时间段内无配送员接单的负样本订单，其中，所述负样本订单的训练权重高于所述正样本订单。
可选地，每一所述概率预估模型由样本训练集训练得到，所述样本训练集包括从所述当前时刻起的所述第一参考时间段内有配送员接单的正样本订单，及从所述当前时刻起的所述第一参考时间段内无配送员接单的负样本订单，其中，所述负样本订单的训练权重高于所述正样本订单。
可选地，所述建议生成模块，用于在所述接单概率小于参考概率阈值的情况下，生成用于取消订单的服务通道以及用于呈现给用户的订单取消操作建议。
可选地，所述装置还包括提示模块，用于根据目标时间段确定所述订单操作建议的提示方式，所述提示方式包括提示条展示、订单取消入口展示、弹窗提示、短信提示中的至少一种，所述目标时间段包括所述目标订单从下单时刻起至当前时刻为止所经过的时间段。
可选地，所述装置还包括，阈值确定模块，用于根据损失预算值确定所述参考概率阈值，其中，被取消的订单造成的损失值之和不超过所述损失预算值。
可选地，所述特征输入模块，用于根据所述目标订单的订单特征信息，确定所述目标订单是否满足运力不足条件；在所述目标订单满足所述运力不足条件的情况下，将所述订单特征信息输入运力不足情形的概率预估模型和普通情形的概率预估模型；在所述目标订单不满足所述运力不足条件的情况下，将所述订单特征信息输入普通情形的概率预估模型；所述概率确定模块，用于在所述目标订单满足所述运力不足条件的情况下，获取所述运力不足情形的概率预估模型输出的第一概率和所述普通情形的概率预估模型输出的第二概率，并基于参考结合规则将所述第一概率和所述第二概率结合为所述接单概率；在所述目标订单不满足所述运力不足条件的情况下，获取所述普通情形的概率预估模型输出的所述接单概率。
可选地，所述装置还包括目标确定模块，用于确定所述目标时间段大于或等于第一参考目标时间段；所述条件确定模块，还用于在所述目标时间段大于所述第二参考目标时间段的情况下，则生成所述订单操作建议，所述第二参考目标时间段大于所述第一参考目标时间段。
可选地，所述装置还包括条件确定模块，用于确定所述目标订单满足参考提示条件；其中，所述参考提示条件包括以下至少一种：所述目标订单的预计配送时间小于参考配送时间阈值；所述目标订单的配送区域中的超时订单数目与所述配送区域中的配送员数目比值小于参考比例阈值；可配送所述目标订单的配送员数目大于参考数目阈值。
本公开的一方面，提供一种计算机可读存储介质，其上存储有计算机程序，该程序被处理器执行时实现本公开任一可选地订单状态提示方法的步骤。
本公开的另一方面，提供一种电子设备，包括：存储器，其上存储有计算机程序；处理器，用于执行所述存储器中的所述计算机程序，以实现本公开任一可选地订单状态提示方法的步骤。
通过上述技术方案，在用户下单后，可以将订单的订单特征信息输入预训 练的概率预估模型，基于概率预估模型的输出确定订单有配送员接单的接单概率，并在该接单概率小于参考概率阈值时生成订单操作建议，这样，可以通过订单的特征来判断订单无人接单的可能性是否较高，并将判断结果和与之对应的操作建议呈现给用户，使用户对订单的状态更明确，从而对订单做出合理的操作处理，这样，用户可以取消接单概率较低的订单，继续等待接单概率较高的订单，平衡外卖平台对订单的调度资源，减少商家及平台的损失，并且可以提升用户的使用体验。
本公开的其他特征和优点将在随后的具体实施方式部分予以详细说明。
附图说明
附图是用来提供对本公开的进一步理解，并且构成说明书的一部分，与下面的具体实施方式一起用于解释本公开，但并不构成对本公开的限制。在附图中：
图1是根据一示例性公开实施例示出的一种订单状态提示方法的流程图。
图2是根据一示例性公开实施例示出的一种订单状态提示方法的流程图。
图3是根据一示例性公开实施例示出的一种订单状态提示装置的框图。
图4是根据一示例性公开实施例示出的一种电子设备的框图。
具体实施方式
以下结合附图对本公开的具体实施方式进行详细说明。应当理解的是，此处所描述的具体实施方式仅用于说明和解释本公开，并不用于限制本公开。
通常的情况下，用户通过外卖平台下单，外卖平台将订单相关的配送需求发布至同区域的多个配送员，收到该配送需求的配送员根据自身的状况选择是否接受该订单的配送，并在接受该订单之后为用户从商家处获取商品并配送至用户的指定地址。但是，配送需求发布以后，并不是所有的订单都会有配送员接单，一些订单面临没有配送员接单的情况，或者，一些订单面临在短时间内没有配送员接单的情况。比如，在一些不利于配送的特殊的情况下，例如在时间较晚、路途较长、天气较差等情况下，用户下单之后可能迟迟没有配送员接单。
在现有技术中，用户下单一段时间之后(例如20分钟之后)没有配送员接单，平台会为用户开放无责任取消窗口，用户可以选择取消订单或继续等待。 但是，用户只能查看订单是否有配送员接单，无法准确地知晓订单的状态，从而难以抉择应该取消订单还是继续等待。
例如，订单在下单后，商家已经开始准备商品，而一个在20分钟之后有配送员接单的概率很高的订单，在用户在下单20分钟时被取消，此时商家会面对商品已经准备完毕却无法卖出的局面，平台也需对商家或用户的损失进行安抚性赔偿；而如果任何时间被配送员接单概率都较低的订单，如果用户在20分钟时选择继续等待，则在等待时间内仍旧无人接单，从而会影响用户的使用体验，订单长时间滞留也会导致平台的调度效率降低的问题。
本公开可以为用户预估出订单被配送员接单的接单概率，并根据接单概率为用户生成操作建议，从而解决上述的问题。
图1是根据一示例性公开实施例示出的一种订单状态提示方法的流程图。如图1所示，该方法可以包括以下步骤：
S11、将目标订单的订单特征信息输入预训练的目标概率预估模型，所述目标概率预估模型用于输出订单在从当前时刻起的第一参考时间段内有配送员接单的概率。
订单特征信息指可以体现订单的客观特性的信息，这些客观特性会从一定程度上影响配送员接单的主观意愿，从而影响到目标订单被接单的概率大小。例如，当商家位置与目标订单的目的地相隔较远时，配送员可能不愿意接单，导致商家位置与目的地相隔较远的订单的接单概率降低；当目标订单的下单时间为周末时，周末订单量较大，而配送员数量有所减少，因此周末的订单的平均接单等待时长会增长，从而使目标订单在一定时间以内被接单的可能性降低；当目标订单所对应的商家的出餐时间较长时，配送员在店内的等候时间也会增长，从而导致配送员不愿接受该商家的订单，从而使目标订单被接单的概率变小。
订单特征信息可以包括以下信息中的一个或多个：业务特征信息、统计特征信息、实时特征信息及商户特征信息，其中，业务特征信息可以包括：配送距离、配送金额、预计送达时长、配送费、下单时间等信息，这类特征信息对不同的订单而言差异很大；统计特征信息可以包括区域平均取餐时间、区域平均送达时间等，对处于同一区域的订单而言，该维度的特征信息是相同且不变的；实时特征信息可以包括：区域未取餐订单数目、区域未送达订单数目、区域未接单订单数目、区域10分钟内下单数目等，这些特征信息会随时间实时变 化，但对处于同一区域内的订单而言，该维度的特征信息是相同的；商户特征信息可以包括：商户平均取餐速度、商户平均到店等待时长等。通过上述的特征信息，可以对目标订单可能对接单概率造成影响的因素进行归类。
在接受到用户提交的订单后，可以对目标订单的订单特征信息进行收集，并将其输入至预训练的目标概率预估模型。
可选地，目标概率预估模型由样本训练集训练得到，样本训练集包括从当前时刻起的第一参考时间段内有配送员接单的正样本订单，及从当前时刻起的第一参考时间段内无配送员接单的负样本订单，其中，负样本订单的训练权重高于正样本订单。其中，样本训练集也称为样本订单集合。
其中，目标概率预估模型是通过样本训练集训练得到的，样本训练集中包括从统计时刻起第一参考时间段内有配送员接单的正样本订单，以及从统计时刻起第一参考时间内无配送员接单的负样本订单(包括在第一参考时间段之后接单、在第一参考时间段之内取消、在第一参考时间段之后取消的情况)，每一样本订单包括了样本订单的订单特征信息和样本接单时间，其中，统计时刻与所述目标订单的“当前时刻”相对应，即，目标订单的下单时刻至所述当前时刻之间的时长与样本订单的下单时刻至统计时刻的时长相同。例如，在下单后第15分钟时，用于输出接单概率的目标概率预估模型是由“从样本订单的下单时刻起15分钟之后的统计时刻开始，在第一参考时间段内有配送员接单的正样本订单”以及“从样本订单的下单时刻起15分钟之后的统计时刻开始，在第一参考时间段内没有配送员接单的负样本订单”组成，如果该第一参考时间段为10分钟，则样本训练集为从下单起15分钟至25分钟内有配送员接单的正样本订单，及从下单起15至25分钟内无配送员接单的负样本订单，15分钟以内有配送员接单或15分钟以内被取消的订单不纳入该目标配送模型的样本训练集中。
由于订单特征信息与接单概率的对应关系可能发生变化，因此，目标概率预估模型需要定期更新，且使用具有时效性的历史订单作为样本训练集。并且，由于在实际的配送业务中，在短时间内(例如20分钟或30分钟内)有配送员接单的订单的数目是远高于在短时间内无配送员接单的订单数目的，因此，在收集样本训练集时，该正样本订单的数目也高于负样本订单的数目，而模型训练时，在样本不均衡的情况下，损失函数会优先使数目较多一方的准确率提升，而使数目较少一方的准确率下降，因此，本公开在进行模型训练时，利用加权交叉熵损失函数进行训练，从而达到对较少的类别进行加权训练的目的，均衡 正样本与负样本的准确率。在具体的实现过程中，可以使用xgboost模型，并且将模型训练的迭代轮次设置为200，训练深度设置为6，样本的采样比例设置为0.4，学习率设置为0.1，为负样本的加权值设置为5。
每一订单特征信息由目标概率预估模型输出为一标签值，该标签值表征该订单特征信息对接单概率的影响程度。目标概率预估模型可以对所有标签值进行统计并进行加权处理，得到最终的标签值，该最终的标签值表征该目标订单在从当前时刻起的第一参考时间段内有配送员接单的概率。
在一种可能的实施方式中，运力不足条件下的目标概率预估模型与其他条件下的目标概率预估模型不同，是由运力不足条件下的样本训练集训练而成的，样本训练集中的所有样本订单都是运力不足条件下的历史订单。在将订单特征输入目标概率预估模型之前，根据所述目标订单的订单特征信息，确定所述目标订单是否满足运力不足条件，在所述目标订单满足所述运力不足条件的情况下，将所述订单特征信息输入运力不足情形的概率预估模型和普通情形的概率预估模型，在所述目标订单不满足所述运力不足条件的情况下，将所述订单特征信息输入普通情形的概率预估模型。在后续步骤中获取目标概率预估模型的输出时，在所述目标订单满足所述运力不足条件的情况下，获取所述运力不足情形的概率预估模型输出的第一概率和所述普通情形的概率预估模型输出的第二概率，并基于参考结合规则将所述第一概率和所述第二概率结合为所述接单概率。
其中，所述运力不足条件可以包括：当前时刻配送区域的十分钟接单率小于参考接单率阈值，该接单率为本区域内10分钟创建的订单中有配送员接单的订单总量，与本区域内10分钟创建的订单总量和本区域内被取消的订单总量之差的比值。
S12、基于所述目标概率预估模型的输出确定所述目标订单的接单概率。
在获得目标概率预估模型输出的最终的标签值之后，可以根据该标签值确定目标订单的接单概率。例如，如果该标签值为0.78，则可以确定该目标订单的接单概率为78％。其中，该接单概率是指“目标订单从当前时刻起的第一参考时间段内有配送员接单的概率”。
S13、在所述接单概率小于参考概率阈值的情况下，生成用于呈现给用户的订单操作建议。
在确定了该接单概率之后，可以以提醒信息的形式将该接单概率呈现给用 户，由用户判断应该对订单采取何种操作；也可以直接根据该接单概率生成推荐的订单操作建议，并将其呈现给用户。
当所述接单概率小于参考概率阈值时，表明该目标订单从当前时刻起的第一参考时间段内有配送员接单的概率较低，则可以生成建议用户取消订单的订单操作建议，例如，可以生成“当前订单还没有骑手接单，您可以选择取消订单”的提示信息，并以弹窗的形式发送给用户。
在一种可能的方式中，可以根据用户下单后经过的时长确定该提示信息的提示方式，该提示方式的提示强度与用户下单后经过的时长成正相关关系。例如，当用户下单10分钟时，可以在应用程序内以常驻提示条的形式展示该提示信息，当用户下单20分钟时，可以在应用程序内通过开放用户免责取消的入口的形式展示该提示信息，当用户下单30分钟时，可以在以通知弹窗与短信相结合的方式展示该提示信息。
在一种可能的实施方式中，可以确定所述订单特征信息中对所述接单概率的影响程度最高的目标特征信息，并根据该目标特征信息生成提示信息。例如。当该目标特征信息为“天气：下雨”时，则可以生成“雨天送餐困难，目前还没有骑手接单，您可以选择取消订单”，以便用户进一步了解订单的状态，并结合订单操作建议进行恰当的订单操作。
在一种可能的实施方式中，在生成订单操作建议之前，还可以确定所述目标订单满足参考提示条件：所述目标订单的预计配送时间小于参考配送时间阈值，或者，所述目标订单的配送区域中的超时订单数目与所述配送区域中的配送员数目比值小于参考比例阈值，或者，可配送所述目标订单的配送员数目大于参考数目阈值。在上述条件下，订单有配送员接单的概率会大大提升，因此可以根据上述条件对由模型确定为大概率无人接单的订单进行过滤。
值得说明的是，由于订单取消会对商家、用户造成损失，一些外卖平台会对此损失进行补偿，因此，该参考概率阈值可以结合损失预算值进行确定，被取消的订单造成的损失值之和不超过所述损失预算值。
通过上述技术方案，在用户下单后，可以将订单的订单特征信息输入预训练的概率预估模型，基于概率预估模型的输出确定订单有配送员接单的接单概率，并在该接单概率小于参考概率阈值时生成订单操作建议，这样，可以通过订单的特征来判断订单无人接单的可能性是否较高，并将判断结果和与之对应的操作建议呈现给用户，使用户对订单的状态更明确，从而对订单做出合理的 操作处理，这样，用户可以取消接单概率较低的订单，继续等待接单概率较高的订单，平衡外卖平台对订单的调度资源，减少商家及平台的损失，并且可以提升用户的使用体验。
图2是根据一示例性公开实施例示出的一种订单状态提示方法的流程图。如图2所示，该方法可以包括以下步骤：
S21、确定所述目标订单从下单时刻起至当前时刻为止所经过的目标时间段。
例如，用户在10:00下单，且当前时刻为10:16时，该目标时间段为16分钟。
S22、判断所述目标时间段与第一参考目标时间段及第二参考目标时间段的关系，其中，第二参考目标时间段大于第一参考目标时间段。
在目标时间段小于第一参考目标时间段的情况下，则返回步骤S21；在目标时间段大于或等于第一参考目标时间段的情况下，并小于等于第二参考目标时间段，则执行步骤S23；在目标时间段大于第二参考目标时间段的情况下，则直接生成订单操作建议。
也就是说，当从下单时刻起至当前时刻为止所经过的目标时间段超过第一参考目标时间段时，才开始使用目标概率预估模型对目标订单的接单概率进行判断，在小于第一参考目标时间段时，订单下单时长还较短，被接单的可能性较高，尚不需对此类订单进行接单概率计算；当目标时间段超过第二参考目标时间段后，用户下单之后的等待时间已经过长，多数用户已经没有耐心等待配送员接单，此时可以直接向用户提出订单操作建议，指导用户对订单进行适当的操作。
S23、从预训练的多个概率预估模型中，确定所述目标概率预估模型，所述目标概率模型的样本时间段与所述目标时间段相同。
其中，每一所述概率预估模型用于输出订单在下单时刻起的第二参考时间段内有配送员接单的概率，所述第二参考时间段为所述目标时间段与所述第一参考时间段之和。
其中，目标时间段的精度为一分钟，也就是说，目标时间段在第一参考目标时间段至第二参考目标时间段之间时，每经过一分钟，就需要确定新的目标概率预估模型，如果第一参考目标时间段为10分钟，第二参考目标时间段为20分钟，则在下单时刻起的第10分钟、第11分钟、第12分钟、第13分钟、第 14分钟、第15分钟、第16分钟、第17分钟、第18分钟、第19分钟、第20分钟，分别都有一个参考的目标概率预估模型，当第一参考时间段为10分钟时，这11个目标概率预估模型分别用于输出从下单时刻起第20分钟(第一参考时间段10分钟与目标时间段10分钟之和)、第21分钟、第22分钟、第23分钟、第24分钟、第25分钟、第26分钟、第27分钟、第28分钟、第29分钟、第30分钟内有配送员接单的概率。
其中，目标概率预估模型是通过样本训练集训练得到的，样本训练集中包括从统计时刻起第一参考时间段内有配送员接单的正样本订单，以及从统计时刻起第一参考时间内无配送员接单的负样本订单(包括在第一参考时间段之后接单、在第一参考时间段之内取消、在第一参考时间段之后取消的情况)，每一样本订单包括了样本订单的订单特征信息和样本接单时间，其中，统计时刻与所述目标订单的“当前时刻”相对应，即，目标订单的下单时刻至所述当前时刻之间的时长与样本订单的下单时刻至统计时刻的时长相同。例如，在下单后第15分钟时，用于输出接单概率的目标概率预估模型是由“从样本订单的下单时刻起15分钟之后的统计时刻开始，在第一参考时间段内有配送员接单的正样本订单”以及“从样本订单的下单时刻起15分钟之后的统计时刻开始，在第一参考时间段内没有配送员接单的负样本订单”组成，如果该第一参考时间段为10分钟，则样本训练集为从下单起15分钟至25分钟内有配送员接单的正样本订单，及从下单起15至25分钟内无配送员接单的负样本订单，15分钟以内有配送员接单或15分钟以内被取消的订单不纳入该目标配送模型的样本训练集中。
由于订单特征信息与接单概率的对应关系可能发生变化，因此，目标概率预估模型需要定期更新，且使用具有时效性的历史订单作为样本训练集。并且，由于在实际的配送业务中，在短时间内(例如20分钟或30分钟内)有配送员接单的订单的数目是远高于在短时间内无配送员接单的订单数目的，因此，在收集样本训练集时，该正样本订单的数目也高于负样本订单的数目，而模型训练时，在样本不均衡的情况下，损失函数会优先使数目较多一方的准确率提升，而使数目较少一方的准确率下降，因此，本公开在进行模型训练时，利用加权交叉熵损失函数进行训练，从而达到对较少的类别进行加权训练的目的，均衡正样本与负样本的准确率。在具体的实现过程中，可以使用xgboost模型，并且将模型训练的迭代轮次设置为200，训练深度设置为6，样本的采样比例设置为0.4，学习率设置为0.1，为负样本的加权值设置为5。
S24、将所述目标订单的订单特征信息输入所述目标概率预估模型。
订单特征信息指可以体现订单的客观特性的信息，这些客观特性会从一定程度上影响配送员接单的主观意愿，从而影响到目标订单被接单的概率大小。例如，当商家位置与目标订单的目的地相隔较远时，配送员可能不愿意接单，导致商家位置与目的地相隔较远的订单的接单概率降低；当目标订单的下单时间为周末时，周末订单量较大，而配送员数量有所减少，因此周末的订单的平均接单等待时长会增长，从而使目标订单在一定时间以内被接单的可能性降低；当目标订单所对应的商家的出餐时间较长时，配送员在店内的等候时间也会增长，从而导致配送员不愿接受该商家的订单，从而使目标订单被接单的概率变小。
订单特征信息可以包括以下信息中的一个或多个：业务特征信息、统计特征信息、实时特征信息及商户特征信息，其中，业务特征信息可以包括：配送距离、配送金额、预计送达时长、配送费、下单时间等信息，这类特征信息对不同的订单而言差异很大；统计特征信息可以包括区域平均取餐时间、区域平均送达时间等，对处于同一区域的订单而言，该维度的特征信息是相同且不变的；实时特征信息可以包括：区域未取餐订单数目、区域未送达订单数目、区域未接单订单数目、区域10分钟内下单数目等，这些特征信息会随时间实时变化，但对处于同一区域内的订单而言，该维度的特征信息是相同的；商户特征信息可以包括：商户平均取餐速度、商户平均到店等待时长等。通过上述的特征信息，可以对目标订单可能对接单概率造成影响的因素进行归类。
在接受到用户提交的订单后，可以对目标订单的订单特征信息进行收集，并将其输入至预训练的目标概率预估模型。
每一订单特征信息由目标概率预估模型输出为一标签值，该标签值表征该订单特征信息对接单概率的影响程度。目标概率预估模型可以对所有标签值进行统计并进行加权处理，得到最终的标签值，该最终的标签值表征该目标订单在从当前时刻起的第一参考时间段内有配送员接单的概率。
在一种可能的实施方式中，运力不足条件下的目标概率预估模型与其他条件下的目标概率预估模型不同，是由运力不足条件下的样本训练集训练而成的，样本训练集中的所有样本订单都是运力不足条件下的历史订单。在将订单特征输入目标概率预估模型之前，根据所述目标订单的订单特征信息，确定所述目标订单是否满足运力不足条件，在所述目标订单满足所述运力不足条件的情况 下，将所述订单特征信息输入运力不足情形的概率预估模型和普通情形的概率预估模型，在所述目标订单不满足所述运力不足条件的情况下，将所述订单特征信息输入普通情形的概率预估模型。在后续步骤中获取目标概率预估模型的输出时，在所述目标订单满足所述运力不足条件的情况下，获取所述运力不足情形的概率预估模型输出的第一概率和所述普通情形的概率预估模型输出的第二概率，并基于参考结合规则将所述第一概率和所述第二概率结合为所述接单概率。
其中，所述运力不足条件可以包括：当前时刻配送区域的十分钟接单率小于参考接单率阈值，该接单率为本区域内10分钟创建的订单中有配送员接单的订单总量，与本区域内10分钟创建的订单总量和本区域内被取消的订单总量之差的比值。
S25、基于所述目标概率预估模型的输出确定所述目标订单的接单概率。
在获得目标概率预估模型输出的最终的标签值之后，可以根据该标签值确定目标订单的接单概率。例如，如果该标签值为0.78，则可以确定该目标订单的接单概率为78％。其中，该接单概率是指“目标订单从当前时刻起的第一参考时间段内有配送员接单的概率”。
S26、在所述接单概率小于参考概率阈值的情况下，生成用于呈现给用户的订单操作建议。
在确定了该接单概率之后，可以以提醒信息的形式将该接单概率呈现给用户，由用户判断应该对订单采取何种操作；也可以直接根据该接单概率生成推荐的订单操作建议，并将其呈现给用户。
当所述接单概率小于参考概率阈值时，表明该目标订单从当前时刻起的第一参考时间段内有配送员接单的概率较低，则可以生成建议用户取消订单的订单操作建议，例如，可以生成“当前订单还没有骑手接单，您可以选择取消订单”的提示信息，并以弹窗的形式发送给用户。
在一种可能的实施方式中，可以确定所述订单特征信息中对所述接单概率的影响程度最高的目标特征信息，并根据该目标特征信息生成提示信息。例如。当该目标特征信息为“天气：下雨”时，则可以生成“雨天送餐困难，目前还没有骑手接单，您可以选择取消订单”，以便用户进一步了解订单的状态，并结合订单操作建议进行恰当的订单操作。
在一种可能的方式中，可以根据目标时间段确定该提示信息的提示方式， 该提示方式的提示强度与目标时间段成正相关关系。例如，当用户下单10分钟时(即目标时间段为10分钟时)，可以在应用程序内以常驻提示条的形式展示该提示信息，当用户下单20分钟时，可以在应用程序内通过开放用户免责取消的入口的形式展示该提示信息，当用户下单30分钟时，可以在以通知弹窗与短信相结合的方式展示该提示信息。
在一种可能的实施方式中，在生成订单操作建议之前，还可以确定所述目标订单满足参考提示条件：所述目标订单的预计配送时间小于参考配送时间阈值，或者，所述目标订单的配送区域中的超时订单数目与所述配送区域中的配送员数目比值小于参考比例阈值，或者，可配送所述目标订单的配送员数目大于参考数目阈值。在上述条件下，订单有配送员接单的概率会大大提升，因此可以根据上述条件对由模型确定为大概率无人接单的订单进行过滤。
值得说明的是，由于订单取消会对商家、用户造成损失，一些外卖平台会对此损失进行补偿，被取消的订单造成的损失值之和不超过所述损失预算值，且所有的概率预估模型的损失预算值之和不超过总的损失预算值。因此，该参考概率阈值可以通过下述公式结合损失预算值进行确定：
Figure PCTCN2021077486-appb-000001
Figure PCTCN2021077486-appb-000002
其中，St.budget为总损失预算值，cost(x i)为第i个模型的损失预算值。
通过上述技术方案，在用户下单后，可以将从下单时刻起至当前时刻为止所经过的目标时间段位于第一参考目标时间段和第二参考目标时间段之间的订单的订单特征信息输入预训练的概率预估模型，基于概率预估模型的输出确定订单有配送员接单的接单概率，并在该接单概率小于参考概率阈值时生成订单操作建议，这样，可以在节省计算资源并符合使用实际的前提下，通过订单的特征来判断订单无人接单的可能性是否较高，并将判断结果和与之对应的操作建议呈现给用户，使用户对订单的状态更明确，从而对订单做出合理的操作处理，从而，用户可以取消接单概率较低的订单，继续等待接单概率较高的订单，平衡外卖平台对订单的调度资源，减少商家及平台的损失，并且可以提升用户的使用体验。
图3是根据一示例性公开实施例示出的一种订单状态提示装置的框图。如图3所示，所述订单状态提示装置300包括：
特征输入模块301，用于将目标订单的订单特征信息输入预训练的目标概率 预估模型，所述目标概率预估模型用于输出订单在从当前时刻起的第一参考时间段内有配送员接单的概率。
概率确定模块302，用于基于所述目标概率预估模型的输出确定所述目标订单的接单概率。
建议生成模块303，用于在所述接单概率小于参考概率阈值的情况下，生成用于呈现给用户的订单操作建议。
可选地，所述装置还包括：时间确定模块，用于确定所述目标订单从下单时刻起至当前时刻为止所经过的目标时间段；所述特征输入模块，用于从预训练的多个概率预估模型中，确定所述目标概率预估模型，所述目标概率模型的样本时间段与所述目标时间段相同，其中，每一所述概率预估模型用于输出订单在下单时刻起的第二参考时间段内有配送员接单的概率，所述第二参考时间段为所述目标时间段与所述第一参考时间段之和，并将所述目标订单的订单特征信息输入所述目标概率预估模型。
可选地，每一所述概率预估模型是由样本订单集合训练得到，所述样本订单集合包括在所述概率预估模型对应的所述样本时间段内有配送员接单的正样本订单，及在所述样本时间段内无配送员接单的负样本订单，其中，所述负样本订单的训练权重高于所述正样本订单。
可选地，每一所述概率预估模型由样本训练集训练得到，所述样本训练集包括从所述当前时刻起的所述第一参考时间段内有配送员接单的正样本订单，及从所述当前时刻起的所述第一参考时间段内无配送员接单的负样本订单，其中，所述负样本订单的训练权重高于所述正样本订单。
可选地，所述建议生成模块，用于在所述接单概率小于参考概率阈值的情况下，生成用于取消订单的服务通道以及用于呈现给用户的订单取消操作建议。
可选地，所述装置还包括提示模块，用于根据目标时间段确定所述订单操作建议的提示方式，所述提示方式包括提示条展示、订单取消入口展示、弹窗提示、短信提示中的至少一种，所述目标时间段包括所述目标订单从下单时刻起至当前时刻为止所经过的时间段。
可选地，所述装置还包括，阈值确定模块，用于根据损失预算值确定所述参考概率阈值，其中，被取消的订单造成的损失值之和不超过所述损失预算值。
可选地，所述特征输入模块，用于根据所述目标订单的订单特征信息，确定所述目标订单是否满足运力不足条件；在所述目标订单满足所述运力不足条 件的情况下，将所述订单特征信息输入运力不足情形的概率预估模型和普通情形的概率预估模型；在所述目标订单不满足所述运力不足条件的情况下，将所述订单特征信息输入普通情形的概率预估模型；所述概率确定模块，用于在所述目标订单满足所述运力不足条件的情况下，获取所述运力不足情形的概率预估模型输出的第一概率和所述普通情形的概率预估模型输出的第二概率，并基于参考结合规则将所述第一概率和所述第二概率结合为所述接单概率；在所述目标订单不满足所述运力不足条件的情况下，获取所述普通情形的概率预估模型输出的所述接单概率。
可选地，所述装置还包括目标确定模块，用于确定所述目标时间段大于或等于第一参考目标时间段；所述条件确定模块，还用于在所述目标时间段大于所述第二参考目标时间段的情况下，则生成所述订单操作建议，所述第二参考目标时间段大于所述第一参考目标时间段。
可选地，所述装置还包括条件确定模块，用于确定所述目标订单满足参考提示条件；其中，所述参考提示条件包括以下至少一种：所述目标订单的预计配送时间小于参考配送时间阈值；所述目标订单的配送区域中的超时订单数目与所述配送区域中的配送员数目比值小于参考比例阈值；可配送所述目标订单的配送员数目大于参考数目阈值。
关于上述实施例中的装置，其中各个模块执行操作的具体方式已经在有关该方法的实施例中进行了详细描述，此处将不做详细阐述说明。
通过上述技术方案，至少可以达到以下技术效果：
在用户下单后，可以将订单的订单特征信息输入预训练的概率预估模型，基于概率预估模型的输出确定订单有配送员接单的接单概率，并在该接单概率小于参考概率阈值时生成订单操作建议，这样，可以通过订单的特征来判断订单无人接单的可能性是否较高，并将判断结果和与之对应的操作建议呈现给用户，使用户对订单的状态更明确，从而对订单做出合理的操作处理，这样，用户可以取消接单概率较低的订单，继续等待接单概率较高的订单，平衡外卖平台对订单的调度资源，减少商家及平台的损失，并且可以提升用户的使用体验。
图4是根据一示例性实施例示出的一种电子设备400的框图。该电子设备400可以被提供为一服务器。如图4所示，该电子设备400可以包括：处理器401，存储器402。该电子设备400还可以包括多媒体组件403，输入/输出(I/O) 接口404，以及通信组件405中的一者或多者。
其中，处理器401用于控制该电子设备400的整体操作，以完成上述的订单状态提示方法中的全部或部分步骤。存储器402用于存储各种类型的数据以支持在该电子设备400的操作，这些数据例如可以包括用于在该电子设备400上操作的任何应用程序或方法的指令，以及应用程序相关的数据。该存储器402可以由任何类型的易失性或非易失性存储设备或者它们的组合实现，例如静态随机存取存储器(Static Random Access Memory，简称SRAM)，电可擦除可编程只读存储器(Electrically Erasable Programmable Read-Only Memory，简称EEPROM)，可擦除可编程只读存储器(Erasable Programmable Read-Only Memory，简称EPROM)，可编程只读存储器(Programmable Read-Only Memory，简称PROM)，只读存储器(Read-Only Memory，简称ROM)，磁存储器，快闪存储器，磁盘或光盘。多媒体组件403可以包括屏幕和音频组件。其中屏幕例如可以是触摸屏，音频组件用于输出和/或输入音频信号。例如，音频组件可以包括一个麦克风，麦克风用于接收外部音频信号。所接收的音频信号可以被进一步存储在存储器402或通过通信组件405发送。音频组件还包括至少一个扬声器，用于输出音频信号。I/O接口404为处理器401和其他接口模块之间提供接口，上述其他接口模块可以是键盘，鼠标，按钮等。这些按钮可以是虚拟按钮或者实体按钮。通信组件405用于该电子设备400与其他设备之间进行有线或无线通信。无线通信，例如Wi-Fi，蓝牙，近场通信(Near Field Communication，简称NFC)，2G、3G、4G、NB-IOT、eMTC、或其他5G等等，或它们中的一种或几种的组合，在此不做限定。因此相应的该通信组件405可以包括：Wi-Fi模块，蓝牙模块，NFC模块等等。
在一示例性实施例中，电子设备400可以被一个或多个应用专用集成电路(Application Specific Integrated Circuit，简称ASIC)、数字信号处理器(Digital Signal Processor，简称DSP)、数字信号处理设备(Digital Signal Processing Device，简称DSPD)、可编程逻辑器件(Programmable Logic Device，简称PLD)、现场可编程门阵列(Field Programmable Gate Array，简称FPGA)、控制器、微控制器、微处理器或其他电子元件实现，用于执行上述的订单状态提示方法。
在另一示例性实施例中，还提供了一种包括程序指令的计算机可读存储介质，该程序指令被处理器执行时实现上述的订单状态提示方法的步骤。例如，该计算机可读存储介质可以为上述包括程序指令的存储器402，上述程序指令可 由电子设备400的处理器401执行以完成上述的订单状态提示方法。
在另一示例性实施例中，还提供一种计算机程序产品，该计算机程序产品包含能够由可编程的装置执行的计算机程序，该计算机程序具有当由该可编程的装置执行时用于执行上述的订单状态提示方法的代码部分。
以上结合附图详细描述了本公开的优选实施方式，但是，本公开并不限于上述实施方式中的具体细节，在本公开的技术构思范围内，可以对本公开的技术方案进行多种简单变型，这些简单变型均属于本公开的保护范围。
另外需要说明的是，在上述具体实施方式中所描述的各个具体技术特征，在不矛盾的情况下，可以通过任何合适的方式进行组合，为了避免不必要的重复，本公开对各种可能的组合方式不再另行说明。
此外，本公开的各种不同的实施方式之间也可以进行任意组合，只要其不违背本公开的思想，其同样应当视为本公开所公开的内容。

Claims (22)

  1. 一种订单状态提示方法，其中，所述方法包括：
    将目标订单的订单特征信息输入预训练的目标概率预估模型，所述目标概率预估模型用于输出订单在从当前时刻起的第一参考时间段内有配送员接单的概率；
    基于所述目标概率预估模型的输出确定所述目标订单的接单概率；
    在所述接单概率小于参考概率阈值的情况下，生成用于呈现给用户的订单操作建议。
  2. 根据权利要求1所述的方法，其中，所述方法还包括：
    确定所述目标订单从下单时刻起至当前时刻为止所经过的目标时间段；
    所述将目标订单的订单特征信息输入预训练的目标概率预估模型，包括：
    从预训练的多个概率预估模型中，确定所述目标概率预估模型，所述目标概率模型的样本时间段与所述目标时间段相同，其中，每一所述概率预估模型用于输出订单在下单时刻起的第二参考时间段内有配送员接单的概率，所述第二参考时间段为所述目标时间段与所述第一参考时间段之和；
    将所述目标订单的订单特征信息输入所述目标概率预估模型。
  3. 根据权利要求2所述的方法，其中，每一所述概率预估模型是由样本订单集合训练得到，所述样本订单集合包括在所述概率预估模型对应的所述样本时间段内有配送员接单的正样本订单，及在所述样本时间段内无配送员接单的负样本订单，其中，所述负样本订单的训练权重高于所述正样本订单。
  4. 根据权利要求2所述的方法，其中，每一所述概率预估模型由样本训练集训练得到，所述样本训练集包括从所述当前时刻起的所述第一参考时间段内有配送员接单的正样本订单，及从所述当前时刻起的所述第一参考时间段内无配送员接单的负样本订单，其中，所述负样本订单的训练权重高于所述正样本订单。
  5. 根据权利要求1或2任一项所述的方法，其中，所述在所述接单概率小于参考概率阈值的情况下，生成用于呈现给用户的订单操作建议包括：
    在所述接单概率小于参考概率阈值的情况下，生成用于取消订单的服务通道以及用于呈现给用户的订单取消操作建议。
  6. 根据权利要求1所述的方法，其中，在所述生成用于呈现给用户的订单操作建议之后，所述方法还包括：
    根据目标时间段确定所述订单操作建议的提示方式，所述提示方式包括提示条展示、订单取消入口展示、弹窗提示、短信提示中的至少一种，所述目标时间段包括所述目标订单从下单时刻起至当前时刻为止所经过的时间段。
  7. 根据权利要求5所述的方法，其中，所述方法还包括：
    根据损失预算值确定所述参考概率阈值，其中，被取消的订单造成的损失值之和不超过所述损失预算值。
  8. 根据权利要求1所述的方法，其中，所述将目标订单的订单特征信息输入预训练的目标概率预估模型，包括：
    根据所述目标订单的订单特征信息，确定所述目标订单是否满足运力不足条件；
    在所述目标订单满足所述运力不足条件的情况下，将所述订单特征信息输入运力不足情形的概率预估模型和普通情形的概率预估模型；
    在所述目标订单不满足所述运力不足条件的情况下，将所述订单特征信息输入普通情形的概率预估模型；
    所述基于所述目标概率预估模型的输出确定所述目标订单的接单概率，包括：
    在所述目标订单满足所述运力不足条件的情况下，获取所述运力不足情形的概率预估模型输出的第一概率和所述普通情形的概率预估模型输出的第二概率，并基于参考结合规则将所述第一概率和所述第二概率结合为所述接单概率；
    在所述目标订单不满足所述运力不足条件的情况下，获取所述普通情形的概率预估模型输出的所述接单概率。
  9. 根据权利要求2所述的方法，其中，所述从预训练的多个概率预估模型中，确定所述目标概率预估模型之前，所述方法还包括：
    确定所述目标时间段大于或等于第一参考目标时间段；
    所述方法还包括：
    在所述目标时间段大于所述第二参考目标时间段的情况下，则生成所述订单操作建议，所述第二参考目标时间段大于所述第一参考目标时间段。
  10. 根据权利要求1所述的方法，其中，在所述生成用于呈现给用户的订单操作建议之前，所述方法还包括：
    确定所述目标订单满足参考提示条件；
    其中，所述参考提示条件包括以下至少一种：
    所述目标订单的预计配送时间小于参考配送时间阈值；
    所述目标订单的配送区域中的超时订单数目与所述配送区域中的配送员数目比值小于参考比例阈值；
    可配送所述目标订单的配送员数目大于参考数目阈值。
  11. 一种订单状态提示装置，其中，所述装置包括：
    特征输入模块，用于将目标订单的订单特征信息输入预训练的目标概率预估模型，所述目标概率预估模型用于输出订单在从当前时刻起的第一参考时间段内有配送员接单的概率；
    概率确定模块，用于基于所述目标概率预估模型的输出确定所述目标订单的接单概率；
    建议生成模块，用于在所述接单概率小于参考概率阈值的情况下，生成用于呈现给用户的订单操作建议。
  12. 根据权利要求11所述的装置，其中，所述装置还包括：时间确定模块，用于确定所述目标订单从下单时刻起至当前时刻为止所经过的目标时间段；所述特征输入模块，用于从预训练的多个概率预估模型中，确定所述目标概率预估模型，所述目标概率模型的样本时间段与所述目标时间段相同，其中，每一所述概率预估模型用于输出订单在下单时刻起的第二参考时间段内有配送员接单的概率，所述第二参考时间段为所述目标时间段与所述第一参考时间段之和，并将所述目标订单的订单特征信息输入所述目标概率预估模型。
  13. 根据权利要求12所述的装置，其中，每一所述概率预估模型是由样本订单集合训练得到，所述样本订单集合包括在所述概率预估模型对应的所述样本时间段内有配送员接单的正样本订单，及在所述样本时间段内无配送员接单的负样本订单，其中，所述负样本订单的训练权重高于所述正样本订单。
  14. 根据权利要求12所述的装置，其中，每一所述概率预估模型由样本训练集训练得到，所述样本训练集包括从所述当前时刻起的所述第一参考时间段内有配送员接单的正样本订单，及从所述当前时刻起的所述第一参考时间段内无配送员接单的负样本订单，其中，所述负样本订单的训练权重高于所述正样本订单。
  15. 根据权利要求11或12所述的装置，其中，所述建议生成模块，用于在所述接单概率小于参考概率阈值的情况下，生成用于取消订单的服务通道以及用于呈现给用户的订单取消操作建议。
  16. 根据权利要求11所述的装置，其中，所述装置还包括提示模块，用于根据目标时间段确定所述订单操作建议的提示方式，所述提示方式包括提示条展示、订单取消入口展示、弹窗提示、短信提示中的至少一种，所述目标时间段包括所述目标订单从下单时刻起至当前时刻为止所经过的时间段。
  17. 根据权利要求15所述的装置，其中，所述装置还包括，阈值确定模块，用于根据损失预算值确定所述参考概率阈值，其中，被取消的订单造成的损失值之和不超过所述损失预算值。
  18. 根据权利要求11所述的装置，其中，所述特征输入模块，用于根据所述目标订单的订单特征信息，确定所述目标订单是否满足运力不足条件；在所述目标订单满足所述运力不足条件的情况下，将所述订单特征信息输入运力不足情形的概率预估模型和普通情形的概率预估模型；在所述目标订单不满足所述运力不足条件的情况下，将所述订单特征信息输入普通情形的概率预估模型；所述概率确定模块，用于在所述目标订单满足所述运力不足条件的情况下，获取所述运力不足情形的概率预估模型输出的第一概率和所述普通情形的概率预 估模型输出的第二概率，并基于参考结合规则将所述第一概率和所述第二概率结合为所述接单概率；在所述目标订单不满足所述运力不足条件的情况下，获取所述普通情形的概率预估模型输出的所述接单概率。
  19. 根据权利要求12所述的装置，其中，所述装置还包括目标确定模块，用于确定所述目标时间段大于或等于第一参考目标时间段；所述条件确定模块，还用于在所述目标时间段大于所述第二参考目标时间段的情况下，则生成所述订单操作建议，所述第二参考目标时间段大于所述第一参考目标时间段。
  20. 根据权利要求11所述的装置，其中，所述装置还包括条件确定模块，用于确定所述目标订单满足参考提示条件；其中，所述参考提示条件包括以下至少一种：所述目标订单的预计配送时间小于参考配送时间阈值；所述目标订单的配送区域中的超时订单数目与所述配送区域中的配送员数目比值小于参考比例阈值；可配送所述目标订单的配送员数目大于参考数目阈值。
  21. 一种计算机可读存储介质，其上存储有计算机程序，其特征在于，该程序被处理器执行时实现权利要求1-10中任一项所述方法的步骤。
  22. 一种电子设备，其中，包括：
    存储器，其上存储有计算机程序；
    处理器，用于执行所述存储器中的所述计算机程序，以实现权利要求1-10中任一项所述方法的步骤。
PCT/CN2021/077486 2020-03-13 2021-02-23 订单状态提示 WO2021179906A1 (zh)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN202010176669.1 2020-03-13
CN202010176669.1A CN113393284A (zh) 2020-03-13 2020-03-13 订单状态提示方法和装置、存储介质和电子设备

Publications (1)

Publication Number Publication Date
WO2021179906A1 true WO2021179906A1 (zh) 2021-09-16

Family

ID=77616125

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2021/077486 WO2021179906A1 (zh) 2020-03-13 2021-02-23 订单状态提示

Country Status (2)

Country Link
CN (1) CN113393284A (zh)
WO (1) WO2021179906A1 (zh)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN114066105A (zh) * 2022-01-11 2022-02-18 浙江口碑网络技术有限公司 运单配送超时预估模型的训练方法，存储介质和电子设备

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2010037394A1 (en) * 2008-10-01 2010-04-08 Just-Eat Group Limited Ordering system with terminal unit having barcode scanner
CN106971326A (zh) * 2016-09-07 2017-07-21 蔚来汽车有限公司 订单的用户预期管理系统和方法
CN109376942A (zh) * 2018-11-12 2019-02-22 达疆网络科技(上海)有限公司 订单处理方法、存储介质和装置
CN110119847A (zh) * 2019-05-14 2019-08-13 拉扎斯网络科技（上海）有限公司 一种配送时长的预测方法、装置、存储介质和电子设备

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2010037394A1 (en) * 2008-10-01 2010-04-08 Just-Eat Group Limited Ordering system with terminal unit having barcode scanner
CN106971326A (zh) * 2016-09-07 2017-07-21 蔚来汽车有限公司 订单的用户预期管理系统和方法
CN109376942A (zh) * 2018-11-12 2019-02-22 达疆网络科技(上海)有限公司 订单处理方法、存储介质和装置
CN110119847A (zh) * 2019-05-14 2019-08-13 拉扎斯网络科技（上海）有限公司 一种配送时长的预测方法、装置、存储介质和电子设备

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN114066105A (zh) * 2022-01-11 2022-02-18 浙江口碑网络技术有限公司 运单配送超时预估模型的训练方法，存储介质和电子设备

Also Published As

Publication number Publication date
CN113393284A (zh) 2021-09-14

Similar Documents

Publication Publication Date Title
US10346889B1 (en) Determining courier effort for deliveries
US10467602B2 (en) Facilitating sending, receiving, and updating of payments using message and payment queues
US20160117651A1 (en) Facilitating sending and receiving of payments between users in a group
US20210090017A1 (en) Feedback-based management of delivery orders
AU2014410285A1 (en) Facilitating sending and receiving of payments using message-based contextual prompts
CN104103113A (zh) 排队服务器以及排队方法
CN108550090A (zh) 一种确定房源价格信息的处理方法及系统
WO2020034851A1 (zh) 推送信息的方法、装置及设备
JP5746554B2 (ja) 情報提供システム、情報提供方法、プログラム及び情報記録媒体
WO2018157808A1 (zh) 互联网征信评估方法和系统
CN111247782B (zh) 用于自动创建即时ad-hoc日历事件的方法和系统
WO2020199961A1 (zh) 候餐时长实时预估方法和系统
US20180374029A1 (en) Selection of customer service requests
WO2021179906A1 (zh) 订单状态提示
US11651320B2 (en) Presentation device and presentation method
CN103348375B (zh) 终端控制装置
CN109961328A (zh) 确定订单冷静期的方法和装置
CN111382359A (zh) 基于强化学习的服务策略推荐方法、装置、电子设备
CN116821513B (zh) 一种推荐场景下的参数搜索方法、装置、设备和介质
WO2018046001A1 (zh) 信息过滤方法、交易控制方法及服务器
CN116307552A (zh) 基于工作量数据信息的工作内容匹配方法
CN109784659A (zh) 服务请求的处理方法、装置、存储介质及电子设备
CN110826823A (zh) 定价策略的评价方法和系统
CN114897584A (zh) 网约车订单处理方法、装置、服务器及存储介质
KR102229727B1 (ko) 스마트 주문처리 장치

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 21767428

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

122 Ep: pct application non-entry in european phase

Ref document number: 21767428

Country of ref document: EP

Kind code of ref document: A1