Methods, apparatus, media, and electronic equipment for increasing the probability of order acceptance.

Abstract

This disclosure relates to a method, apparatus, medium, and electronic device for increasing the probability of order acceptance. The method includes: acquiring end-of-period order recall feature parameters for the current period; determining a target end-of-period order probability threshold for the current period based on the end-of-period order recall feature parameters; acquiring the end-of-period order probability of orders to be generated within the current period; identifying orders to be generated with end-of-period order probabilities greater than the target end-of-period order probability threshold as potential end-of-period orders; and adjusting the delivery parameters of the potential end-of-period orders to increase the probability of them being accepted. In this way, the target end-of-period order probability threshold for the current period can be automatically determined, avoiding the drawbacks of manually determining the target end-of-period order probability threshold in related technologies and improving the efficiency of adjusting delivery parameters. Furthermore, since the target end-of-period order probability threshold is determined for each period, the determined target end-of-period order probability threshold can better meet the needs of users within the current period, thereby achieving the goal of effectively suppressing end-of-period orders without affecting the order volume.

Description

Technical Field

[0001] This disclosure relates to the field of Internet technology, and more specifically, to a method, apparatus, medium, and electronic device for increasing the probability of an order being accepted. Background Technology

[0002] With the continuous development of the internet, users can place orders online through delivery platforms to purchase needed items or reserve vehicles. In actual business scenarios, some orders will remain unaccepted for a long time; these orders are called "last orders." For example, in ride-hailing services, orders that haven't been accepted for an extended period can be considered last orders; similarly, in delivery services, orders that haven't been accepted for an extended period can also be considered last orders. The emergence of last orders not only affects the user experience but also reduces the order acceptance rate in that business scenario.

[0003] For last-minute order scheduling, the common approach is to offer service providers (delivery personnel, drivers) extra service fees. This incentivizes them to accept orders while simultaneously discouraging user orders, thus reducing the number of last-minute orders. However, because the business platform cannot predict in advance whether an order will become a last-minute order, it cannot collect the extra service fee from users beforehand. Once an order becomes a last-minute order, since the service fee has already been charged and the order is valid, the business platform cannot charge the user again; it must provide the extra service fee. Therefore, it is necessary to assess whether an order is a last-minute order before generating it. In related technologies, the estimation of whether an order is a last-minute order is often based on a manually determined probability threshold.

[0004] However, in related technologies, a fixed probability threshold for last-minute orders is determined manually based on experience. This fixed threshold can only ensure the suppression of last-minute order volume in specific scenarios. Furthermore, the manually determined probability threshold needs to be observed during use to determine whether the volume of last-minute orders determined based on this threshold meets user needs, resulting in low efficiency. Summary of the Invention

[0005] The purpose of this disclosure is to provide a method, apparatus, medium, and electronic device for increasing the probability of order acceptance, in order to solve the problems existing in the related art.

[0006] To achieve the above objectives, a first aspect of this disclosure provides a method for increasing the probability of an order being accepted, comprising:

[0007] Obtain the tail-end order recall feature parameters for the current period, wherein the tail-end order recall feature parameters are determined based on orders that were not accepted in historical periods prior to the current period;

[0008] Based on the tail order recall feature parameters, determine the target tail order probability threshold for the current period;

[0009] Obtain the probability of remaining orders within the current period;

[0010] Orders to be generated that have a last-minute order probability greater than the target last-minute order probability threshold are identified as potential last-minute orders.

[0011] Adjust the delivery parameters of the potential last-minute orders to increase the probability that the potential last-minute orders will be accepted.

[0012] Optionally, determining the target last-order probability threshold for the current period based on the last-order recall feature parameters includes:

[0013] The tail order recall feature parameters are input into the tail order probability threshold determination model to obtain the target tail order probability threshold for the current period.

[0014] Optionally, determining the target last-order probability threshold for the current period based on the last-order recall feature parameters includes:

[0015] Based on the tail order recall feature parameters, determine the adjustment value of the tail order probability threshold for the current period;

[0016] The sum of the target tail order probability threshold of the previous period and the adjustment value is determined as the target tail order probability threshold of the current period.

[0017] Optionally, the tail order recall feature parameters include at least one of the tail order recall ratio feature parameters, tail order recall integral feature parameters, and tail order recall differential feature parameters;

[0018] Wherein, the characteristic parameter of the tail-order recall ratio term is the first cumulative tail-order recall rate error, which is the difference between the first cumulative tail-order recall rate accumulated over the first preset number of historical periods before the current period and the preset target tail-order recall rate; the characteristic parameter of the tail-order recall integral term is the sum of the first cumulative tail-order recall rate errors of the second preset number of historical periods before the current period and each period in the current period; the characteristic parameter of the tail-order recall differential term is the difference between the first cumulative tail-order recall rate error of the current period and the first cumulative tail-order recall rate error of the previous period.

[0019] Optionally, determining the target last-order probability threshold for the current period based on the last-order recall feature parameters includes:

[0020] Based on the tail order recall feature parameters, determine the tail order probability threshold for the current period;

[0021] The minimum of the tail order probability threshold and the preset probability threshold is determined as the target tail order probability threshold for the current period.

[0022] Optionally, the tail order probability threshold determination model is obtained in the following way:

[0023] A sample model is determined by obtaining a pre-set probability threshold for last-order orders, wherein the sample model for determining the probability threshold for last-order orders includes at least one preset model parameter.

[0024] Obtain the preset parameter set corresponding to each preset model parameter, wherein the preset parameter set includes multiple pre-set alternative parameter values;

[0025] Based on the candidate parameter values ​​and the tail order probability threshold, a sample model is determined, and multiple tail order probability threshold determination models to be trained are determined.

[0026] Obtain the feature parameters of the last-order recall samples from the third preset number of historical periods;

[0027] The tail order probability threshold determination model is determined based on the tail order recall sample feature parameters and multiple tail order probability threshold determination models to be trained.

[0028] Optionally, determining the tail-order probability threshold determination model based on the tail-order recall sample feature parameters and multiple tail-order probability threshold determination models to be trained includes:

[0029] For each model to be trained to determine the probability threshold of the last few orders, the following training steps are performed:

[0030] The feature parameters of the last order recall samples in each historical period are input into the tail order probability threshold determination model to be trained, so as to obtain the target tail order probability threshold of each historical period output by the tail order probability threshold determination model to be trained.

[0031] For each historical period, a second cumulative tail order recall rate is determined based on the target tail order probability threshold of the historical period, wherein the second cumulative tail order recall rate is the cumulative tail order recall rate of the fourth preset number of historical periods prior to the historical period.

[0032] Based on the second cumulative tail order recall rate of each historical period, the evaluation index parameters corresponding to the third preset number of historical periods of the tail order probability threshold determination model are determined.

[0033] Determine whether the training termination condition is met;

[0034] If the training termination condition is met, the training model with the optimal evaluation index parameters is determined as the tail order probability threshold determination model; and

[0035] If the training termination condition is not met, the training model with the optimal evaluation index parameters is determined as the optimal tail order probability threshold determination model. Based on the optimal parameter values ​​corresponding to the optimal tail order probability threshold determination model, multiple candidate tail order probability threshold determination models are determined. The candidate tail order probability threshold determination models are determined as the training model, and the above training steps are repeated until the training termination condition is met.

[0036] Optionally, the step of determining the optimal parameter value corresponding to the model based on the optimal tail order probability threshold, and determining multiple candidate tail order probability thresholds to determine the model, includes:

[0037] Based on the optimal tail order probability threshold, determine each optimal parameter value corresponding to the model, and determine the optimal parameter range corresponding to each optimal parameter value;

[0038] Based on each of the optimal parameter ranges, generate the optimal parameter set corresponding to each of the optimal parameter values;

[0039] Each parameter value in the target optimal parameter set is combined with each parameter value in each other optimal parameter set to obtain multiple sets of candidate parameter combinations, wherein the target optimal parameter set is any one of the optimal parameter sets;

[0040] Based on the combination of multiple candidate parameters and the optimal tail order probability threshold determination model, multiple candidate tail order probability threshold determination models are determined.

[0041] Optionally, the evaluation index parameters include at least one of the following parameters: the average value of the second cumulative tail order recall rate error over the third preset number of historical periods, the standard deviation of the second cumulative tail order recall rate error over the third preset number of historical periods, and the ratio of the number of periods in the third preset number of historical periods where the second cumulative tail order recall rate is greater than a preset tail order recall rate threshold to the third preset number, wherein the second cumulative tail order recall rate error is the difference between the second cumulative tail order recall rate and the preset target tail order recall rate.

[0042] A second aspect of this disclosure provides an apparatus for increasing the probability of an order being accepted, comprising:

[0043] The first acquisition module is configured to acquire the tail order recall feature parameters for the current period, wherein the tail order recall feature parameters are determined based on orders that have not been accepted in historical periods prior to the current period.

[0044] The first determining module is configured to determine the target tail order probability threshold for the current period based on the tail order recall feature parameters.

[0045] The second acquisition module is configured to acquire the probability of last-minute orders to be generated within the current period;

[0046] The second determining module is configured to determine the orders to be generated that have a tail-end order probability greater than the target tail-end order probability threshold as potential tail-end orders.

[0047] The adjustment module is configured to adjust the delivery parameters of the potential last-minute orders in order to increase the probability that the potential last-minute orders will be accepted.

[0048] Optionally, the first determining module is configured to: input the tail order recall feature parameters into the tail order probability threshold determining model to obtain the target tail order probability threshold for the current period.

[0049] Optionally, the first determining module is configured to: determine an adjustment value for the probability threshold of the last order in the current period based on the last order recall feature parameters; and determine the sum of the target probability threshold of the last order in the previous period and the adjustment value as the target probability threshold of the last order in the current period.

[0050] Optionally, the first determining module is configured to: determine the probability threshold of the last order in the current period based on the last order recall feature parameters; and determine the minimum of the last order probability threshold and the preset probability threshold as the target last order probability threshold for the current period.

[0051] Optionally, the device further includes:

[0052] The third acquisition module is configured to acquire a pre-set sample model for determining the probability threshold of tail orders, wherein the sample model for determining the probability threshold of tail orders includes at least one preset model parameter.

[0053] The fourth acquisition module is configured to acquire a set of preset parameters corresponding to each of the preset model parameters, wherein the set of preset parameters includes a plurality of pre-set alternative parameter values;

[0054] The third determining module is configured to determine a sample model based on the candidate parameter values ​​and the tail order probability threshold, and to determine multiple tail order probability threshold determination models to be trained.

[0055] The fifth acquisition module is configured to acquire feature parameters of the last order recall samples from a third preset number of historical periods;

[0056] The fourth determining module is configured to determine the tail order probability threshold determining model based on the tail order recall sample feature parameters and multiple tail order probability threshold determining models to be trained.

[0057] Optionally, the fourth determining module includes:

[0058] The first execution submodule is configured to perform the following training steps for each tail-order probability threshold determination model to be trained: inputting the tail-order recall sample feature parameters of each historical period into the tail-order probability threshold determination model to be trained, so as to obtain the target tail-order probability threshold of each historical period output by each tail-order probability threshold determination model to be trained; for each historical period, determining the second cumulative tail-order recall rate of the historical period based on the target tail-order probability threshold of the historical period, wherein the second cumulative tail-order recall rate is the cumulative tail-order recall rate of the fourth preset number of historical periods before the historical period; and determining the evaluation index parameters of the tail-order probability threshold determination model to be trained for the third preset number of historical periods based on the second cumulative tail-order recall rate of each historical period.

[0059] The judgment submodule is configured to determine whether the training termination condition is met.

[0060] The first determining submodule is configured to, when the training termination condition is met, determine the training tail-single probability threshold determination model with the optimal evaluation index parameters as the tail-single probability threshold determination model; and

[0061] The second determining submodule is configured to, in the absence of the training termination condition, determine the tail order probability threshold determination model with the optimal evaluation index parameters as the optimal tail order probability threshold determination model, determine multiple candidate tail order probability threshold determination models based on the optimal parameter values ​​corresponding to the optimal tail order probability threshold determination model, determine the candidate tail order probability threshold determination models as the tail order probability threshold determination model to be trained, and repeat the above training steps until the training termination condition is met.

[0062] Optionally, the second determining submodule is configured to: determine each optimal parameter value corresponding to the model based on the optimal tail order probability threshold, and determine the optimal parameter interval corresponding to each optimal parameter value; generate an optimal parameter set corresponding to each optimal parameter value based on each optimal parameter interval; arrange and combine each parameter value in the target optimal parameter set with each parameter value in each other optimal parameter set to obtain multiple sets of candidate parameter combinations, wherein the target optimal parameter set is any one of the optimal parameter sets; and determine multiple candidate tail order probability threshold determining models based on the multiple sets of candidate parameter combinations and the optimal tail order probability threshold determining model.

[0063] A third aspect of this disclosure provides a computer-readable storage medium having a computer program stored thereon that, when executed by a processor, implements the steps of the method provided in the first aspect of this disclosure.

[0064] A fourth aspect of this disclosure provides an electronic device, comprising:

[0065] A memory on which computer programs are stored;

[0066] A processor for executing the computer program in the memory to implement the steps of the method provided in the first aspect of this disclosure.

[0067] By employing the aforementioned technical solution, the target last-minute order probability threshold for the current period can be automatically determined based on the last-minute order recall characteristic parameters. This avoids the drawbacks of manually determining the target last-minute order probability threshold in related technologies, thus improving the efficiency of adjusting delivery parameters. Furthermore, since a target last-minute order probability threshold is determined for each period, the determined target last-minute order probability threshold can better meet the needs of users within the current period. In this way, the goal of effectively suppressing last-minute orders can be achieved without affecting the order volume.

[0068] Other features and advantages of this disclosure will be described in detail in the following detailed description section. Attached Figure Description

[0069] The accompanying drawings are provided to further illustrate the present disclosure and form part of the specification. They are used together with the following detailed description to explain the present disclosure, but do not constitute a limitation thereof. In the drawings:

[0070] Figure 1 This is a flowchart illustrating a method for increasing the probability of an order being accepted, according to an exemplary embodiment.

[0071] Figure 2 This is a flowchart illustrating a method for determining a probability threshold for tail orders according to an exemplary embodiment.

[0072] Figure 3 This is a flowchart illustrating a method for determining a target tail order probability threshold according to an exemplary embodiment.

[0073] Figure 4 This is a block diagram illustrating an apparatus for increasing the probability of an order being accepted, according to an exemplary embodiment.

[0074] Figure 5 This is a block diagram illustrating an electronic device according to an exemplary embodiment.

[0075] Figure 6 This is a block diagram illustrating an electronic device according to an exemplary embodiment. Detailed Implementation

[0076] As described in the background section, related technologies often involve manually inputting a fixed end-of-line probability threshold. Then, based on the end-of-line prediction model's output end-of-line probability and the threshold, it's predicted whether the order to be generated is an end-of-line order. For example, orders with an end-of-line probability greater than the threshold are predicted as end-of-line orders. This method has the following problems:

[0077] 1. Since the probability threshold for last-minute orders is set based on experience, it needs to be verified in actual use to determine whether the recall and precision rates meet the requirements. If not, the probability threshold needs to be adjusted until it does. On the one hand, the probability threshold relies on human experience, which cannot guarantee the effectiveness of suppressing last-minute orders. Furthermore, the required recall and precision rates vary depending on the scenario (e.g., different geographical environments, different seasons). Therefore, manually determining and adjusting the probability threshold for each scenario is labor-intensive and inefficient.

[0078] 2. Typically, during the Spring Festival, the number of pending orders with a probability greater than the threshold output by the end-of-season order prediction model is relatively small, resulting in fewer end-of-season order recalls and a lower adjustment ratio for service fees. Thus, the end-of-season order suppression effect is weak, and the order size remains stable. To improve end-of-season order recall, the end-of-season order probability threshold needs to be lowered. However, in winter, the number of pending orders with a probability greater than the threshold output by the end-of-season order prediction model is large, leading to more end-of-season order recalls and a higher adjustment ratio for service fees. This inhibits user orders, resulting in a loss in order size. In this case, the end-of-season order probability threshold needs to be increased. Therefore, different end-of-season order probability thresholds are needed for different scenarios to effectively suppress end-of-season orders without affecting order size. However, in related technologies, since the end-of-season order probability threshold is fixed, it is impossible to achieve the goal of effectively suppressing end-of-season orders without affecting order size.

[0079] In view of this, this disclosure provides a method, apparatus, readable storage medium, and electronic device for increasing the probability of an order being accepted, to solve the problems existing in related technologies, to achieve the purpose of automatically determining the probability threshold of leftover orders, and to adjust the delivery parameters of the order according to the automatically determined probability threshold, thereby improving the efficiency of adjusting the delivery parameters. Furthermore, it can also ensure the effect of suppressing leftover orders and maintain the stability of order volume.

[0080] The specific embodiments of this disclosure will be described in detail below with reference to the accompanying drawings. It should be understood that the specific embodiments described herein are for illustration and explanation only and are not intended to limit this disclosure.

[0081] The application scenarios disclosed herein can be various delivery scenarios, such as food delivery, express delivery, etc.

[0082] Figure 1 This is a flowchart illustrating a method for increasing the probability of an order being accepted, according to an exemplary embodiment. This method can be applied to electronic devices with processing capabilities, such as terminals or servers. Figure 1 As shown, the method may include the following steps.

[0083] In S101, obtain the tail order recall feature parameters for the current period.

[0084] In practical applications, considering that the number of last-minute order recalls is affected not only by seasonal and other cyclical factors, but also by unforeseen events such as weather, this disclosure, to ensure the stability of last-minute order recalls over a period of time, determines the recall characteristic parameters based on orders that were not accepted in historical periods prior to the current period. It is worth noting that orders that were not accepted in historical periods prior to the current period refer to the cumulative number of orders that were not accepted by delivery capacity in historical periods prior to the current period. The period can be 1 day, 1 month, etc. Delivery capacity can be delivery personnel, unmanned delivery vehicles, drones, etc.

[0085] For example, in this disclosure, the period can be 1 day, the current period is July 6, 2020, and the last-minute order recall feature parameter can be the cumulative number of orders that were not accepted by delivery personnel from July 1, 2020 to July 5, 2020.

[0086] In S102, the target last order probability threshold for the current period is determined based on the last order recall feature parameters.

[0087] In this disclosure, for each period, a target end-of-period order probability threshold can be determined to improve the accuracy of the determined target end-of-period order probability threshold, thereby enabling subsequent orders to be generated within that period to be accurately identified as end-of-period orders based on the target end-of-period order probability threshold.

[0088] In step 103, the probability of last-minute orders to be generated in the current period is obtained.

[0089] For example, orders to be generated in the current period can be input into the tail-end order prediction model to obtain the tail-end order probability output by the tail-end order prediction model. The tail-end order probability is used to indicate whether the order to be generated is a tail-end order. The larger the value, the greater the probability that the order to be generated will be a tail-end order later.

[0090] "Orders to be generated" refers to potential orders that have not yet been formally placed. For example, on a food delivery platform, a user can add food items to their shopping cart. Even though the user hasn't paid for the items in the cart, these items can still form a potential order. Similarly, on a ride-hailing platform, a user can choose one or more ride types (such as private car, taxi, express car, carpooling, etc.). Before the user completes payment, the currently selected ride type can form a potential order. These potential orders are what we refer to as "orders to be generated" in this disclosure.

[0091] It should be noted that in this disclosure, the execution order of S101, S102, and S103 is not specifically limited. For example, steps S101 and S102 can be executed first, followed by S103 (e.g., ...). Figure 1 As shown), S103 can be executed first, followed by S101 and S102, or S103 can be executed during the execution of S101 and S102.

[0092] In step 104, orders to be generated with a probability greater than the target probability threshold for last-minute orders are identified as potential last-minute orders.

[0093] It is worth noting that "end-of-line orders" refer to orders that have not been accepted after a merchant has posted them. The method provided in this disclosure for increasing the probability of orders being accepted is applied at the stage when the merchant generates an order. Therefore, based on the target end-of-line order probability threshold, it is not possible to determine whether the order to be generated is a real end-of-line order, but only whether it is a potential end-of-line order.

[0094] Furthermore, since the purpose of identifying last-minute orders is to determine the delivery parameters of the order, and orders are usually generated based on delivery parameters, in this embodiment, it is determined whether the order to be generated is a potential last-minute order.

[0095] In step 105, the delivery parameters of potential last-minute orders are adjusted to increase the probability that the potential last-minute orders will be accepted.

[0096] Potential last-minute orders are used to indicate orders that are unlikely to be accepted by delivery capacity after being placed. To prevent these potential last-minute orders from becoming actual last-minute orders, which would lead to a poor user experience, this disclosure allows for adjustments to the delivery parameters of potential last-minute orders, such as increasing delivery service fees. This can both increase the incentive for delivery capacity to accept orders and inhibit users from placing orders, thereby achieving a balance between supply and demand.

[0097] In one embodiment, the delivery parameters to be adjusted can be determined based on how much the last-minute order probability exceeds the target last-minute order probability threshold. For example, if the difference between the last-minute order probability and the target last-minute order probability threshold is 10%, the adjusted delivery parameter is one yuan; if the difference is 20%, the adjusted delivery parameter is two yuan, and so on.

[0098] By adopting the above technical solution, the target last-minute order probability threshold for the current period can be automatically determined based on the last-minute order recall characteristic parameters. This avoids the drawbacks of manually determining the target last-minute order probability threshold in related technologies and improves the efficiency of adjusting delivery parameters. Furthermore, since a target last-minute order probability threshold is determined for each period, the determined target last-minute order probability threshold can better meet the needs of users within the current period. Thus, it is possible to effectively suppress last-minute orders without affecting the order volume.

[0099] In one embodiment, the target tail-order probability threshold for the current period can be determined using machine learning. For example, Figure 1 S102 in the process may further include: inputting the tail order recall feature parameters into the tail order probability threshold determination model to obtain the target tail order probability threshold for the current period.

[0100] First, the tail-order recall feature parameters in this embodiment will be described. The tail-order recall feature parameters may include at least one of the following: tail-order recall proportional feature parameters, tail-order recall integral feature parameters, and tail-order recall differential feature parameters. The following explanation will use a tail-order probability threshold determination model as a PID (Proportion-Integral-Derivative) model, where the tail-order recall feature parameters include tail-order recall proportional feature parameters, tail-order recall integral feature parameters, and tail-order recall differential feature parameters.

[0101] The characteristic parameter of the last-minute order recall ratio is the first cumulative last-minute order recall rate error. This first cumulative last-minute order recall rate error is the difference between the first cumulative last-minute order recall rate accumulated over the first preset number of historical periods before the current period and the preset target last-minute order recall rate. For example, assuming the period is 1 day, the current period i is July 6, 2020, and the first preset number is 5, then the last-minute orders accumulated from July 1, 2020 to July 5, 2020 (i.e., orders not accepted by delivery personnel during this period) are counted. Assuming the accumulated number of last-minute orders is M1, and the number of orders predicted as potential last-minute orders within M1 is M2, where M2 is less than M1, then the first cumulative last-minute order recall rate y for the current period i is calculated. i =M2 / M1, the preset target recall rate of last-minute orders is y*, then the error of the first cumulative recall rate of last-minute orders in the current period i is e. i =|y i -y * |, the characteristic parameter b of the tail order recall ratio in the current period i i =e i .

[0102] The feature parameter of the tail-order recall integral term is the sum of the errors of the first cumulative tail-order recall rate for the second preset number of historical periods before the current period and for each period in the current period. For example, y1 is the first cumulative tail-order recall rate on July 1, 2020, y2 is the first cumulative tail-order recall rate on July 2, 2020, ..., y6 is the first cumulative tail-order recall rate on July 6, 2020. The errors between the first cumulative tail-order recall rate from July 1, 2020 to July 6, 2020 and the preset target tail-order recall rate y* are calculated as e1, e2, ..., e6. Then, the feature parameter of the tail-order recall integral term for the current period i is... It is worth noting that in this embodiment, the second preset quantity is the same as the first preset quantity. In practical applications, the second preset quantity and the first preset quantity may also be different.

[0103] The differential feature parameter of tail-order recall is the difference between the first cumulative tail-order recall error of the current period and the first cumulative tail-order recall error of the previous period. For example, if the current period i is July 6, 2020, the first cumulative tail-order recall error of the current period is e6, and the first cumulative tail-order recall error of the previous period is e5, then the differential feature parameter w of tail-order recall for the current period i is... i =e6-e5.

[0104] Next, the model for determining the tail order probability threshold used in this embodiment will be explained. Specifically, as follows... Figure 2 As shown, the model for determining the probability threshold of tail orders can be pre-trained through the following steps:

[0105] In S201: Obtain a pre-set probability threshold for determining the tail order to determine the sample model. This sample model may include at least one preset model parameter (i.e., a pre-set model parameter). For example, if the probability threshold determination model for the tail order is a PID model, the preset model parameter may be the proportional coefficient Kb, the integral coefficient Kj, and the derivative coefficient Kw.

[0106] In S202: Obtain the preset parameter set corresponding to each preset model parameter. The preset parameter set includes multiple pre-set alternative parameter values. For example, m proportional coefficients Kb, m integral coefficients Kj, and m differential coefficients Kw can be pre-generated randomly within the range of 0 to 1. The m proportional coefficients Kb form the preset parameter set corresponding to the proportional coefficients, the m integral coefficients Kj form the preset parameter set corresponding to the integral coefficients, and the m differential coefficients Kw form the preset parameter set corresponding to the differential coefficients, where m is an integer greater than 1.

[0107] In S203: A sample model is determined based on the candidate parameter values ​​and the tail-order probability threshold, thus determining multiple tail-order probability threshold determination models to be trained. For example, firstly, each candidate parameter value in the target preset parameter set is combined with each candidate parameter value in each other preset parameter set to obtain multiple parameter combinations. Each parameter combination includes a proportional term coefficient, an integral term coefficient, and a differential term coefficient. The target preset parameter set can be any preset parameter set, and the other preset parameter sets can include preset parameter sets other than the target preset parameter set. Then, based on these multiple parameter combinations and the tail-order probability threshold, a sample model is determined, thus determining multiple tail-order probability threshold determination models to be trained.

[0108] In S204: Obtain the feature parameters of the last-order recall samples from the third preset number of historical periods. (Refer to...) Figure 1 The method for obtaining the tail-end order recall feature parameters in the current period is to obtain the tail-end order recall sample feature parameters. The third preset quantity can be 10, 20, etc.

[0109] In S205: The model is determined based on the feature parameters of the last-order recall samples and multiple last-order probability thresholds to be trained. The model is determined by the last-order probability thresholds.

[0110] The specific implementation method of step 205 may include:

[0111] For each model to be trained to determine the probability threshold of the last few orders, the following training steps are performed:

[0112] First, the feature parameters of the last order recall samples in each historical period are input into the model to be trained to determine the probability threshold of the last order, so as to obtain the target probability threshold of the last order for each historical period output by the model to be trained.

[0113] Next, for each historical period, a second cumulative tail-order recall rate is determined based on the target tail-order probability threshold for that historical period. This second cumulative tail-order recall rate is the cumulative tail-order recall rate of the fourth preset number of historical periods preceding the historical period. In this embodiment, the fourth preset number can be the same as or different from the first preset number. Furthermore, after determining the target tail-order probability threshold for the historical period, the number of potential tail-orders can be determined from the cumulative tail-orders of the fourth preset number of historical periods preceding the historical period, based on the target tail-order probability threshold, thereby determining the second cumulative tail-order recall rate. For example, if the fourth preset number is 4, and the historical period is July 5, 2020, assuming that the unaccepted tail-orders identified between July 1, 2020 and July 4, 2020 are Order 1, Order 2, Order 3, Order 4... Order 10, then the cumulative number of tail-orders is 10. Of these 10 last-minute orders, orders 2, 4, 5, and 8 were identified as potential last-minute orders during the order generation period. Therefore, the second cumulative last-minute order recall rate is 40% (4 / 10 = 40%).

[0114] Next, based on the second cumulative tail-order recall rate for each historical period, the evaluation metric parameters corresponding to the third preset number of historical periods for determining the tail-order probability threshold of the model to be trained are determined. These evaluation metric parameters are used to evaluate whether the tail-order probability threshold output by the model can meet the user's requirements for tail-order recall rate.

[0115] For example, the third preset quantity can be 10, and the third preset quantity of historical periods can be from June 20, 2020 to June 29, 2020. For each training model to determine the probability threshold of tail orders, the target probability threshold of tail orders for each day from June 20, 2020 to June 29, 2020 can be determined, and then the second cumulative tail order recall rate for each day can be determined.

[0116] In this disclosure, the evaluation index parameters may include at least one of the following parameters: the average value of the second cumulative tail order recall rate error over a third preset number of historical periods, the standard deviation of the second cumulative tail order recall rate error over a third preset number of historical periods, and the ratio of the number of periods in which the second cumulative tail order recall rate is greater than a preset tail order recall rate threshold to the third preset number, wherein the second cumulative tail order recall rate error is the difference between the second cumulative tail order recall rate and the target tail order recall rate.

[0117] For example, following the example above, the average value of the second cumulative tail order recall rate error for the third preset number of historical periods can be determined using formula (1). Among them, y i The second cumulative last-order recall rate represents the total recall rate for each historical period, y* represents the preset target last-order recall rate, and N is the third preset quantity, where y i The method for determining the first cumulative last order recall rate can be referred to the above description.

[0118]

[0119] The second cumulative tail-order recall error for each historical period was calculated separately. Specifically, the second cumulative tail-order recall error for each day from June 20, 2020 to June 29, 2020 was calculated. Then, the standard deviation σ(e) of these 10 second cumulative tail-order recall errors was calculated. i ), where the value of i ranges from [1, N], and N = 10.

[0120] In addition, the ratio A of the number of periods in which the second cumulative last-order recall rate is greater than the preset last-order recall rate threshold in the third preset number of historical periods can be calculated using formula (2) to the third preset number. α For control coefficients, αy * Characterizes the preset last-order recall rate threshold, and α For numbers greater than 1, The number of periods in which the second cumulative last-order recall rate is greater than the preset last-order recall rate threshold:

[0121]

[0122] In one embodiment, the evaluation index parameter includes any of the following parameters. In this embodiment, after the evaluation index parameter is determined in the manner described above, the tail order probability threshold determination model can be determined from multiple tail order probability threshold determination models to be trained based on the evaluation index parameter.

[0123] In another embodiment, the evaluation index parameters include two or three of the above parameters. In this embodiment, the two or three determined above can be summed to obtain the evaluation index parameters corresponding to the third preset number of historical periods of the tail order probability threshold determination model to be trained.

[0124] For example, let's take the evaluation index parameter Z, which includes three of the above parameters, as an example. The various evaluation index parameters determined above can be summed, as shown in formula (3):

[0125]

[0126] Thus, the evaluation index parameters corresponding to the third preset number of historical periods for each tail-order probability threshold determination model to be trained can be determined. In this embodiment, the smaller the Z value of the evaluation index parameter, the better the target tail-order probability threshold determined by the model can meet the user's needs for tail-order recall rate. Therefore, in this disclosure, the tail-order probability threshold determination model can be determined based on the tail-order probability threshold determination model to be trained with the optimal evaluation index parameters.

[0127] In one embodiment, the tail order probability threshold determination model with the optimal evaluation index parameters (i.e., the smallest evaluation index parameters) can be directly determined as the final tail order probability threshold determination model to be used.

[0128] In actual training, the more training iterations, the more accurate the resulting model. Therefore, multiple training iterations are typically performed to obtain a more accurate model. For example, in another embodiment, after obtaining the evaluation index parameters corresponding to the third preset number of historical periods for each tail-order probability threshold-determined model during each training round, it is necessary to determine whether the training termination condition is met. This termination condition can be that the number of training iterations reaches a preset number, or that the evaluation index parameters corresponding to the third preset number of historical periods for the tail-order probability threshold-determined model reach a preset value, etc.

[0129] In this embodiment, if the training termination condition is met, the training model with the optimal evaluation index parameters is directly selected as the final training model for determining the tail order probability threshold, and the training ends. If the training termination condition is not met, training can be repeated until the training termination condition is met.

[0130] Considering that the candidate parameter values ​​included in the above-mentioned preset parameter set are human-selected and somewhat one-sided values, in one embodiment, if the training termination condition is not met, the training model with the smallest evaluation index parameter can be determined as the optimal tail order probability threshold determination model. Based on the optimal parameter value corresponding to the optimal tail order probability threshold determination model, multiple candidate tail order probability threshold determination models are determined, and the candidate tail order probability threshold determination models are determined as the tail order probability threshold determination models to be trained. The above training steps are repeated until the training termination condition is met, and the training model with the optimal evaluation index parameter is determined as the tail order probability threshold determination model.

[0131] For example, determining the optimal parameter values ​​for the model based on the optimal tail-order probability threshold, and determining the model by identifying multiple candidate tail-order probability thresholds, may include:

[0132] First, determine the optimal parameter value for each model based on the optimal tail order probability threshold, and then determine the optimal parameter range corresponding to each optimal parameter value.

[0133] For example, suppose the optimal parameter combination corresponding to the optimal tail-end probability threshold for determining the model is [Kb1, Kj1, Kw1], where Kb1 is the coefficient of the first proportional term pre-generated randomly within the range of 0 to 1, Kj1 is the coefficient of the first integral term pre-generated randomly within the range of 0 to 1, and Kw1 is the coefficient of the first differential term pre-generated randomly within the range of 0 to 1. Then the optimal parameter values ​​are Kb1, Kj1, and Kw1. For each optimal parameter value, the optimal parameter interval corresponding to that optimal parameter value is determined. For example, based on the optimal parameter values ​​and the spacing parameter k in the current training process... now The optimal parameter range corresponding to Kb1 can be determined as (Kb1 / k now ,Kb1*k now The optimal parameter range corresponding to Kj1 can be (Kj1 / k). now ,Kj1*k now The optimal parameter range corresponding to Kw1 can be in the range (Kw1 / k). now ,Kw1*k now ), where the spacing parameter k is used in each training process. now They can be the same or different.

[0134] Next, based on each optimal parameter range, a set of optimal parameters corresponding to each optimal parameter value is generated.

[0135] For example, m values ​​can be randomly selected from each optimal parameter range to generate the optimal parameter set corresponding to each optimal parameter value.

[0136] Next, each parameter value in the target optimal parameter set is combined with each parameter value in each other optimal parameter set to obtain multiple sets of candidate parameter combinations, where the target optimal parameter set is any one of the optimal parameter sets.

[0137] Finally, based on the combination of multiple candidate parameters and the optimal tail order probability threshold, multiple candidate tail order probability determination models are determined.

[0138] After identifying multiple candidate tail-single probability determination models as described above, each candidate tail-single probability threshold determination model is selected as the tail-single probability threshold determination model to be trained, and the above training steps are repeated until the training termination condition is met. Upon meeting the training termination condition, the tail-single probability threshold determination model with the optimal evaluation index parameters is selected as the final tail-single probability threshold determination model. Thus, the tail-single probability threshold determination model can be pre-trained.

[0139] By using the above training method, the most suitable parameter value can be selected from a relatively large number of parameter values, thereby obtaining a more accurate model for determining the probability threshold of the last order, thus improving the accuracy of determining the probability threshold of the target last order.

[0140] In another embodiment, the target tail-end probability threshold for the current period can also be determined by other means. For example, such as Figure 3 As shown, Figure 1 S102 may further include S1021 and step S1022.

[0141] In step 1021, the adjustment value of the last order probability threshold for the current period is determined based on the last order recall feature parameters.

[0142] In step 1022, the sum of the target tail order probability threshold and the adjustment value of the previous period is determined as the target tail order probability threshold of the current period.

[0143] In this embodiment, the target tail order probability threshold for the current period is determined based on the target tail order probability threshold for the previous period.

[0144] For example, suppose the tail-order recall feature parameters include the tail-order recall ratio feature parameter b. i The feature parameter j of the end-order recall integral term i The characteristic parameter w of the differential term of the tail single recall i , can be used for b i j i and w i Weighted summation to obtain the adjustment value u i For example, the target tail order probability threshold for the current period can be obtained through formula (4):

[0145]

[0146] Among them, threshold i-1 Threshold is the target tail order probability threshold of the previous period i-1 in the current period i. i Let Kb, Kj, and Kw be the target tail order probability thresholds for the current period i, and let b, Kj, and Kw be the values ​​of b, Kj, and Kw, respectively. i j i and w i The corresponding coefficients, where Kb, Kj, and Kw can be empirical values ​​or optimal parameter values ​​determined using the training method described above. The determination method for Kb, Kj, and Kw will not be elaborated here.

[0147] It is worth noting that the initial value of the target last-minute order probability threshold can be determined by the user based on the probability of last-minute orders in each pending order location in the historical period, the number of last-minute orders that were not accepted after the order was generated, and the expected last-minute order recall rate.

[0148] In another embodiment, to avoid low recall accuracy for last-minute orders, before determining the target last-minute order probability threshold, it is also necessary to determine whether the determined probability threshold for the current period is too large. For example, Figure 1 The specific implementation method of S102 can be as follows: determine the probability threshold of the last order in the current period based on the last order recall feature parameters; determine the minimum of the last order probability threshold and the preset probability threshold as the target last order probability threshold for the current period.

[0149] It is worth noting that the tail order probability threshold for the current period can be determined by using the tail order probability threshold determination model in the above machine learning method, or it can be determined by the above formula (4). Then, the determined tail order probability threshold for the current period is compared with the preset probability threshold. If the tail order probability threshold is greater than the preset probability threshold, the preset probability threshold is determined as the target tail order probability threshold for the current period; if the tail order probability threshold is not greater than the preset probability threshold, the tail order probability threshold is determined as the target tail order probability threshold for the current period. Among them, the preset probability threshold can be the maximum tail order probability threshold determined by the user based on the tail order data in the historical period, or it can be the tail order probability threshold determined by the user based on experience.

[0150] By adopting the above technical solution, after determining the probability threshold of the last order in the current period, the probability threshold of the last order is compared with the preset probability threshold, and the smaller of the two is determined as the target probability threshold of the last order. In this way, the target probability threshold of the last order can be avoided to be too large, and the accuracy of the last order recall can be ensured.

[0151] Based on the same inventive concept, this disclosure provides a device for increasing the probability of an order being accepted. Figure 4 This is a block diagram illustrating an apparatus for increasing the probability of an order being accepted, according to an exemplary embodiment. Figure 4 As shown, the device 400 for increasing the probability of an order being accepted may include:

[0152] The first acquisition module 401 is configured to acquire the tail order recall feature parameters of the current period, wherein the tail order recall feature parameters are determined based on orders that have not been accepted in historical periods before the current period.

[0153] The first determining module 402 is configured to determine the target tail order probability threshold for the current period based on the tail order recall feature parameters.

[0154] The second acquisition module 403 is configured to acquire the probability of last-minute orders to be generated in the current period;

[0155] The second determining module 404 is configured to determine the orders to be generated that have a tail-end order probability greater than the target tail-end order probability threshold as potential tail-end orders.

[0156] The adjustment module 405 is configured to adjust the delivery parameters of the potential last-minute orders in order to increase the probability that the potential last-minute orders will be accepted.

[0157] Optionally, the first determining module 402 is configured to: input the tail order recall feature parameters into the tail order probability threshold determining model to obtain the target tail order probability threshold for the current period.

[0158] Optionally, the first determining module 402 is configured to: determine an adjustment value for the probability threshold of the last order in the current period based on the last order recall feature parameters; and determine the sum of the target probability threshold of the last order in the previous period and the adjustment value as the target probability threshold of the last order in the current period.

[0159] Optionally, the first determining module 402 is configured to: determine the last-order probability threshold of the current period based on the last-order recall feature parameters; and determine the minimum of the last-order probability threshold and the preset probability threshold as the target last-order probability threshold of the current period.

[0160] Optionally, the device further includes:

[0161] The third acquisition module is configured to acquire a pre-set sample model for determining the probability threshold of tail orders, wherein the sample model for determining the probability threshold of tail orders includes at least one preset model parameter.

[0162] The fourth acquisition module is configured to acquire a set of preset parameters corresponding to each of the preset model parameters, wherein the set of preset parameters includes a plurality of pre-set alternative parameter values;

[0163] The third determining module is configured to determine a sample model based on the candidate parameter values ​​and the tail order probability threshold, and to determine multiple tail order probability threshold determination models to be trained.

[0164] The fifth acquisition module is configured to acquire feature parameters of the last order recall samples from a third preset number of historical periods;

[0165] The fourth determining module is configured to determine the tail order probability threshold determining model based on the tail order recall sample feature parameters and multiple tail order probability threshold determining models to be trained.

[0166] Optionally, the fourth determining module includes:

[0167] The first execution submodule is configured to perform the following training steps for each tail-order probability threshold determination model to be trained: inputting the tail-order recall sample feature parameters of each historical period into the tail-order probability threshold determination model to be trained, so as to obtain the target tail-order probability threshold of each historical period output by each tail-order probability threshold determination model to be trained; for each historical period, determining the second cumulative tail-order recall rate of the historical period based on the target tail-order probability threshold of the historical period, wherein the second cumulative tail-order recall rate is the cumulative tail-order recall rate of the fourth preset number of historical periods before the historical period; and determining the evaluation index parameters of the tail-order probability threshold determination model to be trained for the third preset number of historical periods based on the second cumulative tail-order recall rate of each historical period.

[0168] The judgment submodule is configured to determine whether the training termination condition is met.

[0169] The first determining submodule is configured to, when the training termination condition is met, determine the training tail-single probability threshold determination model with the optimal evaluation index parameters as the tail-single probability threshold determination model; and

[0170] The second determining submodule is configured to, in the absence of the training termination condition, determine the tail order probability threshold determination model with the optimal evaluation index parameters as the optimal tail order probability threshold determination model, determine multiple candidate tail order probability threshold determination models based on the optimal parameter values ​​corresponding to the optimal tail order probability threshold determination model, determine the candidate tail order probability threshold determination models as the tail order probability threshold determination model to be trained, and repeat the above training steps until the training termination condition is met.

[0171] Optionally, the second determining submodule is configured to: determine each optimal parameter value corresponding to the model based on the optimal tail order probability threshold, and determine the optimal parameter interval corresponding to each optimal parameter value; generate an optimal parameter set corresponding to each optimal parameter value based on each optimal parameter interval; arrange and combine each parameter value in the target optimal parameter set with each parameter value in each other optimal parameter set to obtain multiple sets of candidate parameter combinations, wherein the target optimal parameter set is any one of the optimal parameter sets; and determine multiple candidate tail order probability threshold determining models based on the multiple sets of candidate parameter combinations and the optimal tail order probability threshold determining model.

[0172] Regarding the apparatus in the above embodiments, the specific manner in which each module performs its operation has been described in detail in the embodiments related to the method, and will not be elaborated upon here.

[0173] Figure 5 This is a block diagram illustrating an electronic device 500 according to an exemplary embodiment. For example... Figure 5 As shown, the electronic device 500 may include a processor 501 and a memory 502. The electronic device 500 may also include one or more of a multimedia component 503, an input / output (I / O) interface 504, and a communication component 505.

[0174] The processor 501 controls the overall operation of the electronic device 500 to complete all or part of the steps in the method described above for increasing the probability of order acceptance. The memory 502 stores various types of data to support the operation of the electronic device 500. This data may include, for example, instructions for any application or method operating on the electronic device 500, and application-related data such as contact data, sent and received messages, images, audio, video, etc. The memory 502 can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as Static Random Access Memory (SRAM), Electrically Erasable Programmable Read-Only Memory (EEPROM), Erasable Programmable Read-Only Memory (EPROM), Programmable Read-Only Memory (PROM), Read-Only Memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk. Multimedia component 503 may include a screen and an audio component. The screen may be, for example, a touchscreen, and the audio component is used to output and / or input audio signals. For example, the audio component may include a microphone for receiving external audio signals. The received audio signals may be further stored in memory 502 or transmitted via communication component 505. The audio component also includes at least one speaker for outputting audio signals. I / O interface 504 provides an interface between processor 501 and other interface modules, such as a keyboard, mouse, buttons, etc. These buttons may be virtual or physical buttons. Communication component 505 is used for wired or wireless communication between the electronic device 500 and other devices. Wireless communication, such as Wi-Fi, Bluetooth, Near Field Communication (NFC), 2G, 3G, 4G, NB-IoT, eMTC, or other 5G technologies, or combinations thereof, is not limited here. Therefore, the corresponding communication component 505 may include: a Wi-Fi module, a Bluetooth module, an NFC module, etc.

[0175] In an exemplary embodiment, the electronic device 500 may be implemented by one or more application-specific integrated circuits (ASICs), digital signal processors (DSPs), digital signal processing devices (DSPDs), programmable logic devices (PLDs), field-programmable gate arrays (FPGAs), controllers, microcontrollers, microprocessors, or other electronic components to perform the above-described method for increasing the probability of order acceptance.

[0176] In another exemplary embodiment, a computer-readable storage medium including program instructions is also provided. When executed by a processor, these program instructions implement the steps of the method for increasing the probability of order acceptance described above. For example, the computer-readable storage medium may be the memory 502 including the program instructions described above, which may be executed by the processor 501 of the electronic device 500 to complete the method for increasing the probability of order acceptance described above.

[0177] Figure 6 This is a block diagram illustrating an electronic device 600 according to an exemplary embodiment. For example, the electronic device 600 may be provided as a server. (Refer to...) Figure 6 The electronic device 600 includes a processor 622, which may be one or more, and a memory 632 for storing computer programs executable by the processor 622. The computer programs stored in the memory 632 may include one or more modules, each corresponding to a set of instructions. Furthermore, the processor 622 may be configured to execute the computer program to perform the aforementioned methods for increasing the probability of order acceptance.

[0178] Additionally, the electronic device 600 may also include a power supply component 626 and a communication component 650. The power supply component 626 can be configured to perform power management of the electronic device 600, and the communication component 650 can be configured to enable communication of the electronic device 600, such as wired or wireless communication. Furthermore, the electronic device 600 may also include an input / output (I / O) interface 658. The electronic device 600 can operate on an operating system, such as Windows Server, stored in memory 632. TM Mac OSX TM Unix TM Linux TM etc.

[0179] In another exemplary embodiment, a computer-readable storage medium including program instructions is also provided, which, when executed by a processor, implement the steps of the method for increasing the probability of order acceptance described above. For example, the computer-readable storage medium may be the memory 632 including the program instructions described above, which may be executed by the processor 622 of the electronic device 600 to complete the method for increasing the probability of order acceptance described above.

[0180] In another exemplary embodiment, a computer program product is also provided, the computer program product comprising a computer program executable by a programmable device, the computer program having a code portion for performing the above-described method for increasing the probability of order acceptance when executed by the programmable device.

[0181] The preferred embodiments of this disclosure have been described in detail above with reference to the accompanying drawings. However, this disclosure is not limited to the specific details of the above embodiments. Within the scope of the technical concept of this disclosure, various simple modifications can be made to the technical solutions of this disclosure, and these simple modifications all fall within the protection scope of this disclosure.

[0182] It should also be noted that the various specific technical features described in the above embodiments can be combined in any suitable manner without contradiction. To avoid unnecessary repetition, this disclosure will not describe the various possible combinations separately.

[0183] Furthermore, various different embodiments of this disclosure can be combined in any way, as long as they do not violate the spirit of this disclosure, they should also be regarded as the content disclosed in this disclosure.

Claims

1. A method for increasing the probability of an order being accepted, characterized in that, include: Obtain the tail-end order recall feature parameters for the current period, wherein the tail-end order recall feature parameters are determined based on orders that were not accepted in historical periods prior to the current period; Based on the tail-order recall feature parameters, the target tail-order probability threshold for the current period is determined. To ensure stable tail-order recall over a period, the target tail-order probability threshold for the current period is the sum of the target tail-order probability threshold for the previous period and an adjustment value. The tail-order probability threshold is determined using a proportional-integral-differential (PI-DI) model. The tail-order recall feature parameters include a proportional term feature parameter, an integral term feature parameter, and a differential term feature parameter. The adjustment value is obtained by weighted summation of these three parameters. The characteristic parameter of the tail-order recall ratio is the first cumulative tail-order recall rate error, which is the difference between the first cumulative tail-order recall rate accumulated over the first preset number of historical periods before the current period and the preset target tail-order recall rate. The feature parameter of the tail-order recall integral term is the sum of the first cumulative tail-order recall rate error of the second preset number of historical periods before the current period and the first cumulative tail-order recall rate error of each period in the current period. The differential feature parameter of the tail order recall is the difference between the first cumulative tail order recall rate error of the current period and the first cumulative tail order recall rate error of the previous period. Obtain the probability of remaining orders within the current period; Orders to be generated that have a last-minute order probability greater than the target last-minute order probability threshold are identified as potential last-minute orders. Adjust the delivery parameters of the potential last-minute orders to increase the probability that the potential last-minute orders will be accepted.

2. The method according to claim 1, characterized in that, The step of determining the target last-order probability threshold for the current period based on the last-order recall feature parameters includes: The tail order recall feature parameters are input into the tail order probability threshold determination model to obtain the target tail order probability threshold for the current period.

3. The method according to claim 1, characterized in that, The step of determining the target last-order probability threshold for the current period based on the last-order recall feature parameters includes: Based on the tail order recall feature parameters, determine the adjustment value of the tail order probability threshold for the current period; The sum of the target tail order probability threshold of the previous period and the adjustment value is determined as the target tail order probability threshold of the current period.

4. The method according to claim 1, characterized in that, The step of determining the target last-order probability threshold for the current period based on the last-order recall feature parameters includes: Based on the tail order recall feature parameters, determine the tail order probability threshold for the current period; The minimum of the tail order probability threshold and the preset probability threshold is determined as the target tail order probability threshold for the current period.

5. The method according to claim 2, characterized in that, The tail order probability threshold determination model is obtained through the following method: A sample model is determined by obtaining a pre-set probability threshold for last-order orders, wherein the sample model for determining the probability threshold for last-order orders includes at least one preset model parameter. Obtain the preset parameter set corresponding to each preset model parameter, wherein the preset parameter set includes multiple pre-set alternative parameter values; Based on the candidate parameter values ​​and the tail order probability threshold, a sample model is determined, and multiple tail order probability threshold determination models to be trained are determined. Obtain the feature parameters of the last-order recall samples from the third preset number of historical periods; The tail order probability threshold determination model is determined based on the tail order recall sample feature parameters and multiple tail order probability threshold determination models to be trained.

6. The method according to claim 5, characterized in that, The step of determining the tail-order probability threshold determination model based on the tail-order recall sample feature parameters and multiple tail-order probability threshold determination models to be trained includes: For each model to be trained to determine the probability threshold of the last few orders, the following training steps are performed: The feature parameters of the last order recall samples in each historical period are input into the tail order probability threshold determination model to be trained, so as to obtain the target tail order probability threshold of each historical period output by the tail order probability threshold determination model to be trained. For each historical period, a second cumulative tail order recall rate is determined based on the target tail order probability threshold of the historical period, wherein the second cumulative tail order recall rate is the cumulative tail order recall rate of the fourth preset number of historical periods prior to the historical period. Based on the second cumulative tail order recall rate of each historical period, the evaluation index parameters corresponding to the third preset number of historical periods of the tail order probability threshold determination model are determined. Determine whether the training termination condition is met; If the training termination condition is met, the training model with the optimal evaluation index parameters is determined as the tail order probability threshold determination model; and If the training termination condition is not met, the training model with the optimal evaluation index parameters is determined as the optimal tail order probability threshold determination model. Based on the optimal parameter values ​​corresponding to the optimal tail order probability threshold determination model, multiple candidate tail order probability threshold determination models are determined. The candidate tail order probability threshold determination models are determined as the training model, and the above training steps are repeated until the training termination condition is met.

7. The method according to claim 6, characterized in that, The step of determining the optimal parameter value corresponding to the model based on the optimal tail order probability threshold, and determining multiple candidate tail order probability thresholds to determine the model, includes: Based on the optimal tail order probability threshold, determine each optimal parameter value corresponding to the model, and determine the optimal parameter range corresponding to each optimal parameter value; Based on each of the optimal parameter ranges, generate the optimal parameter set corresponding to each of the optimal parameter values; Each parameter value in the target optimal parameter set is combined with each parameter value in each other optimal parameter set to obtain multiple sets of candidate parameter combinations, wherein the target optimal parameter set is any one of the optimal parameter sets; Based on the combination of multiple candidate parameters and the optimal tail order probability threshold determination model, multiple candidate tail order probability threshold determination models are determined.

8. The method according to claim 6, characterized in that, The evaluation index parameters include at least one of the following parameters: the average value of the second cumulative tail order recall rate error of the third preset number of historical periods, the standard deviation of the second cumulative tail order recall rate error of the third preset number of historical periods, and the ratio of the number of periods in the third preset number of historical periods where the second cumulative tail order recall rate is greater than the preset tail order recall rate threshold to the third preset number, wherein the second cumulative tail order recall rate error is the difference between the second cumulative tail order recall rate and the preset target tail order recall rate.

9. An apparatus for increasing the probability of an order being accepted, characterized in that, include: The first acquisition module is configured to acquire the tail order recall feature parameters for the current period, wherein the tail order recall feature parameters are determined based on orders that have not been accepted in historical periods prior to the current period. The first determining module is configured to determine the target tail-order probability threshold for the current period based on the tail-order recall feature parameters. To ensure the stability of tail-order recall over a period of time, the target tail-order probability threshold for the current period is the sum of the target tail-order probability threshold for the previous period and an adjustment value. The tail-order probability threshold is determined using a proportional-integral-differential (PI-DI) model. The tail-order recall feature parameters include a tail-order recall proportional term feature parameter, a tail-order recall integral term feature parameter, and a tail-order recall differential term feature parameter. The adjustment value is obtained by weighted summation of the tail-order recall proportional term feature parameter, tail-order recall integral term feature parameter, and tail-order recall differential term feature parameter. The characteristic parameter of the tail-order recall ratio is the first cumulative tail-order recall rate error, which is the difference between the first cumulative tail-order recall rate accumulated over the first preset number of historical periods before the current period and the preset target tail-order recall rate. The feature parameter of the tail-order recall integral term is the sum of the first cumulative tail-order recall rate error of the second preset number of historical periods before the current period and the first cumulative tail-order recall rate error of each period in the current period. The differential feature parameter of the tail order recall is the difference between the first cumulative tail order recall rate error of the current period and the first cumulative tail order recall rate error of the previous period. The second acquisition module is configured to acquire the probability of last-minute orders to be generated within the current period; The second determining module is configured to determine the orders to be generated that have a tail-end order probability greater than the target tail-end order probability threshold as potential tail-end orders. The adjustment module is configured to adjust the delivery parameters of the potential last-minute orders in order to increase the probability that the potential last-minute orders will be accepted.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When executed by a processor, the program implements the steps of the method described in any one of claims 1-8.

11. An electronic device, characterized in that, include: A memory on which computer programs are stored; A processor for executing the computer program in the memory to implement the steps of the method according to any one of claims 1-8.

Images