Order processing method, device, storage medium and electronic equipment

Abstract

The present disclosure relates to an order processing method, device, storage medium and electronic device, the method comprising: collecting at least two feature information of the current order; using the feature information as the input of each prediction model in a multidimensional prediction model to obtain a multidimensional prediction result, wherein the multidimensional prediction model includes at least two prediction models, and different prediction models are used to output the prediction results of the current order under decision factors of different dimensions; determining the delivery difficulty parameter of the current order based on the multidimensional prediction result, and the delivery difficulty parameter is used to characterize the delivery difficulty of the current order; determining the processing strategy of the current order based on the delivery difficulty parameter of the current order. Since the present disclosure adopts a multidimensional prediction model when determining the delivery difficulty of the current order, the delivery difficulty of the current order is predicted from decision factors of different dimensions, and more and more comprehensive factors that characterize the order delivery difficulty are taken into consideration, and thus the order delivery difficulty can be accurately determined.

Description

Technical Field

[0001] The present disclosure relates to the field of information management, and in particular, to an order processing method, device, storage medium, and electronic device. Background Art

[0002] With the development of internet technology, online food delivery has greatly facilitated people's lives. More and more users are choosing to place their orders online, with online platforms typically dispatching delivery drivers. When delivering orders, online platforms typically need to determine the delivery difficulty of the order. However, numerous factors influence this difficulty, such as delivery time, merchant location, and food item type. Inaccurate delivery difficulty assessments can lead to irrational decisions by online platforms, such as inappropriate pricing for merchants and users, and inappropriate delivery driver scheduling. Summary of the Invention

[0003] The main purpose of the present disclosure is to provide an order processing method, device, storage medium and electronic device that can accurately determine the delivery difficulty of an order.

[0004] In order to achieve the above objectives, the present disclosure provides, in a first aspect, an order processing method, the method comprising:

[0005] Collect at least two characteristic information of the current order;

[0006] Using the feature information as input to each prediction model in a multidimensional prediction model to obtain a multidimensional prediction result, wherein the multidimensional prediction model includes at least two prediction models, and different prediction models are used to output prediction results of the current order under decision factors of different dimensions;

[0007] Determining a delivery difficulty parameter of the current order based on the multi-dimensional prediction result, where the delivery difficulty parameter is used to characterize the delivery difficulty of the current order;

[0008] A processing strategy for the current order is determined based on a delivery difficulty parameter of the current order.

[0009] Optionally, determining the delivery difficulty parameter of the current order according to the multi-dimensional prediction result includes:

[0010] Obtaining a model weight of a prediction model for each dimension in the multidimensional prediction model;

[0011] The delivery difficulty parameter of the current order is determined based on the multi-dimensional prediction results and the model weight of the prediction model of each dimension.

[0012] Optionally, obtaining the model weight of the prediction model of each dimension in the multidimensional prediction model includes:

[0013] Obtain a model weight of a prediction model for each dimension in the pre-configured multi-dimensional prediction model.

[0014] Optionally, obtaining the model weight of the prediction model of each dimension in the multidimensional prediction model includes:

[0015] Determine the current delivery scenario;

[0016] A model weight is configured for the prediction model of each dimension in the multidimensional prediction model according to the current delivery scenario.

[0017] Optionally, determining the current delivery scenario includes:

[0018] Determine the current supply and demand status of transportation capacity based on the planned delivery data and historical order data for the area where the current order is located;

[0019] The current delivery scenario is determined based on the current transportation capacity supply and demand status.

[0020] Optionally, determining the delivery difficulty parameter of the current order based on the multi-dimensional prediction results and the model weight of the prediction model of each dimension includes:

[0021] The prediction result of the current order in each dimension is multiplied by the model weight of the prediction model of the corresponding dimension, and the maximum value of the obtained results is taken as the delivery difficulty parameter of the current order.

[0022] Optionally, determining the delivery difficulty parameter of the current order based on the multi-dimensional prediction results and the model weight corresponding to the prediction model of each dimension includes:

[0023] The prediction result of the current order in each dimension is multiplied by the model weight of the prediction model of the corresponding dimension, and the sum is taken as the delivery difficulty parameter of the current order.

[0024] Optionally, the multidimensional prediction model is trained according to the following method:

[0025] Collect multiple feature information of historical orders and construct samples;

[0026] Set up multidimensional decision factors;

[0027] The sample and the actual decision results of the sample under the decision factors of each dimension in the multidimensional decision factors are used as training data to perform model training to obtain the multidimensional prediction model.

[0028] Optionally, after setting the multidimensional decision factors, the following is also included:

[0029] Binarize the actual decision results of the sample under the decision factors of each dimension in the multidimensional decision factors.

[0030] A second aspect of the present disclosure provides an order processing device, the device comprising:

[0031] A collection module is configured to collect at least two feature information of the current order;

[0032] a prediction module configured to use the feature information as input to each prediction model in a multidimensional prediction model to obtain a multidimensional prediction result, wherein the multidimensional prediction model includes at least two prediction models, and different prediction models are used to output prediction results of the current order under decision factors of different dimensions;

[0033] A first determining module is configured to determine a delivery difficulty parameter of the current order based on the multi-dimensional prediction result, wherein the delivery difficulty parameter is used to represent the delivery difficulty of the current order;

[0034] The second determination module is configured to determine a processing strategy for the current order according to a delivery difficulty parameter of the current order.

[0035] Optionally, the first determining module includes:

[0036] an acquisition submodule, configured to acquire a model weight of a prediction model of each dimension in the multidimensional prediction model;

[0037] The determination submodule is configured to determine the delivery difficulty parameter of the current order based on the multi-dimensional prediction results and the model weight of the prediction model of each dimension.

[0038] Optionally, the acquisition submodule includes:

[0039] The weight acquisition submodule is configured to obtain the model weight of the prediction model of each dimension in the pre-configured multi-dimensional prediction model.

[0040] Optionally, the acquisition submodule includes:

[0041] A scenario determination submodule, configured to determine a current delivery scenario;

[0042] The weight configuration submodule is configured to configure the model weight for the prediction model of each dimension in the multidimensional prediction model according to the current delivery scenario.

[0043] Optionally, the scenario determination submodule is configured to determine the current transportation capacity supply and demand status based on the planned delivery data and historical order data of the area where the current order is located; and determine the current delivery scenario based on the current transportation capacity supply and demand status.

[0044] Optionally, the determination submodule is configured to specifically multiply the prediction result of the current order in each dimension by the model weight of the prediction model of the corresponding dimension, and take the maximum value of the obtained results as the delivery difficulty parameter of the current order.

[0045] Optionally, the determination submodule is configured to specifically multiply the prediction result of the current order in each dimension by the model weight of the prediction model of the corresponding dimension, and then sum them up, and use the resulting sum as the delivery difficulty parameter of the current order.

[0046] Optionally, the device further comprises:

[0047] A construction module is configured to collect at least two feature information of historical orders and construct a sample;

[0048] a setting module configured to set multidimensional decision factors;

[0049] The training module is configured to use the sample and the actual decision results of the sample under the decision factors of each dimension in the multidimensional decision factors as training data to perform model training and obtain the multidimensional prediction model.

[0050] Optionally, the device further comprises:

[0051] The binarization module is configured to binarize the actual decision result of the sample under the decision factor of each dimension in the multi-dimensional decision factor.

[0052] A third aspect of the present disclosure provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the order processing method described in the first aspect.

[0053] A fourth aspect of the present disclosure provides an electronic device, including:

[0054] a memory having a computer program stored thereon;

[0055] A processor is used to execute the computer program in the memory to implement the steps of the order processing method described in the first aspect.

[0056] The technical solution provided by this disclosure can achieve at least the following technical effects:

[0057] The technical solution provided by the embodiments of the present disclosure collects at least two feature information of the current order, uses the collected feature information as the input of each prediction model in the multidimensional prediction model, obtains a multidimensional prediction result, and determines the delivery difficulty parameter of the current order based on the multidimensional prediction result. The delivery difficulty parameter is used to characterize the delivery difficulty of the current order, and then determines the processing strategy of the current order based on the delivery difficulty parameter of the current order. Since a multidimensional prediction model is used when determining the delivery difficulty of the current order, prediction models of different dimensions are used to output prediction results of the current order under decision factors of different dimensions, thereby predicting the delivery difficulty of the current order under decision factors of different dimensions, taking into account more and more comprehensive factors that characterize the difficulty of order delivery, thereby accurately determining the difficulty of order delivery, and providing a reliable basis for the decision-making of the network platform.

[0058] Other features and advantages of the present disclosure will be described in detail in the following detailed description. BRIEF DESCRIPTION OF THE DRAWINGS

[0059] The accompanying drawings are used to provide a further understanding of the present disclosure and constitute a part of the specification. Together with the following detailed description, they are used to explain the present disclosure but do not constitute a limitation of the present disclosure. In the accompanying drawings:

[0060] Figure 1 is a flow chart of an order processing method shown in an embodiment of the present disclosure;

[0061] Figure 2 is a flowchart of a training method for a multidimensional prediction model shown in an embodiment of the present disclosure;

[0062] Figure 3 is a flow chart of another order processing method shown in an embodiment of the present disclosure;

[0063] Figure 4 is a flow chart of another order processing method shown in an embodiment of the present disclosure;

[0064] Figure 5 is a flowchart of another order processing method shown in an embodiment of the present disclosure;

[0065] Figure 6 is a structural block diagram of an order processing device shown in an embodiment of the present disclosure;

[0066] Figure 7 is a structural block diagram of another order processing device shown in an embodiment of the present disclosure;

[0067] Figure 8 It is a structural diagram of an electronic device shown in an embodiment of the present disclosure. DETAILED DESCRIPTION

[0068] The following describes the specific embodiments of the present disclosure in detail with reference to the accompanying drawings. It should be understood that the specific embodiments described herein are only used to illustrate and explain the present disclosure and are not intended to limit the present disclosure.

[0069] When delivering orders, online platforms usually need to determine the difficulty of delivering the orders. The results of the order delivery difficulty determination can provide a basis for online platforms to make decisions on user-side pricing, merchant-side pricing, rider scheduling, merchant exposure ranking, etc. Generally speaking, orders with long delivery times, no riders to take orders, delivery timeouts, slow food delivery by merchants, remote merchant locations, inconvenient access to the user's community, and difficulty in combining packages are all orders with high delivery difficulty. For orders with high delivery difficulty, for example, higher delivery fees can be charged to the merchants or users involved in the order, or rider dispatch can be delayed, or the merchants involved in the order can be ranked relatively low when the merchant list is displayed to the user. Specific decisions can be made based on actual needs.

[0070] In the relevant technologies for determining the difficulty of order delivery, a single decision factor is usually used to build a single prediction model to determine the delivery difficulty of an order. For example, the long-term order prediction model simply predicts the delivery difficulty of an order by predicting the delivery time of the order. That is, the longer the order delivery time, the higher the order delivery difficulty. Obviously, this is unreasonable. For example, the rider's willingness to accept orders, the length of time the rider waits for meals, etc., are also important factors that significantly affect the difficulty of order delivery. However, the existing single-model prediction scheme cannot take these factors into account, resulting in inaccurate results in determining the difficulty of order delivery.

[0071] The inventor has noticed this problem and proposed an order processing method, which is as follows:

[0072] See also Figure 1 , Figure 1 This is a flow chart of an order processing method shown in an embodiment of the present disclosure. The execution subject of the order processing method shown in the embodiment of the present disclosure may be the order processing device provided in the embodiment of the present disclosure, or an electronic device integrated with the order processing device, wherein the electronic device may be a backend server of a network platform, such as Figure 1 As shown, the method includes the following steps:

[0073] Step 101: Collect at least two feature information of the current order.

[0074] A current order can be an order that is currently in process of being transacted. For example, if a user searches for "milk tea" on an online platform, and the online platform displays a list of merchants offering milk tea based on the user's search, the user may then transact with any of the merchants in the list, resulting in an order for milk tea. This order is considered a currently in process of being transacted. Alternatively, a current order can be an order that has already been transacted, such as a user having already placed an order for a particular dish at a particular merchant.

[0075] For example, the characteristic information of an order can be divided into different dimensions, including but not limited to: merchant dimension, user dimension, rider dimension, order dimension, and merchant and user intersection dimension. Each dimension may include one or more characteristic information, and the number of characteristic information contained in each dimension may be the same or different.

[0076] Among them, the characteristic information of the merchant dimension includes: the merchant's address, the area where the merchant is located, the business district where the merchant is located, the longitude and latitude of the merchant's location, the merchant's food delivery speed, etc.; the characteristic information of the user dimension includes: the user's address, the area where the user is located, the longitude and latitude of the user's location, whether the user's community is convenient to enter and exit, the degree of traffic congestion in the user's area, etc.; the characteristic information of the rider dimension includes: the number of riders, the number of orders received by the rider, etc.; the characteristic information of the order dimension includes: the types of food involved in the order, etc.; the characteristic information of the cross-dimensionality of merchants and users includes: the distance between the merchant and the user, the number of historical transaction orders between the merchant and the user, etc.

[0077] In a specific implementation, at least two feature information of the current order can be collected from a real-time feature platform or from an offline feature platform, that is, the at least two feature information collected can be real-time feature information related to the current order or historical feature information related to the current order. The at least two feature information of the current order collected can belong to the same dimension or different dimensions, and no specific limitation is made here.

[0078] Step 102: Use the feature information as input to each prediction model in the multidimensional prediction model to obtain a multidimensional prediction result.

[0079] The multidimensional prediction model includes at least two prediction models, each of which outputs a prediction result. Different prediction models are used to output the prediction results of the current order under decision factors of different dimensions.

[0080] For example, decision factors of different dimensions may involve factors that affect the difficulty of delivery in multiple dimensions, such as the willingness of the rider to accept the order, the time the rider waits for the meal, the delivery time of the rider, the loss of the delivery meal, and the delivery time of the order. Specifically, decision factors of different dimensions may be, for example: whether there is no rider willing to accept the order within 15 minutes, whether the time the rider waits for the meal exceeds 10 minutes, whether the delivery time of the rider exceeds 5 minutes, whether the loss of the delivery meal exceeds 10 yuan, whether the delivery time of the order exceeds 20 minutes, etc. In a specific embodiment, the multidimensional prediction model may be shown in Table 1 below.

[0081] Dimensions Prediction Model Decision Factors 1 Rider order acceptance willingness prediction model Is there no rider willing to accept the order within 15 minutes? 2 Rider waiting time prediction model Whether the rider waits for the meal for more than 10 minutes 3 Rider delivery time prediction model Whether the delivery time of the rider exceeds 5 minutes 4 Order delivery time prediction model Is the order delivery time longer than 20 minutes? 5 Delivery timeout prediction model Will the order delivery be delayed? 6 Delivery meal loss prediction model Whether the meal loss compensation for delivery exceeds 10 yuan 7 Package probability prediction model Is the probability of combining packages greater than 60%? 8 Capacity occupancy time prediction model Whether the capacity occupancy time exceeds the preset value 9 Delivery tail order prediction model Is it the last order? …… …… ……

[0082] Table 1

[0083] It should be noted that the multidimensional prediction models shown in Table 1 are for illustration only and are not intended to limit the types and number of models.

[0084] Among them, when at least two feature information of the current order are used as the input of each prediction model in the multidimensional prediction model, the feature information of the current order input to each prediction model can be the same or different, depending on the features taken during the training of the multidimensional prediction model. For example, if the at least two feature information of the current order include five feature information, A, B, C, D, and E, when making predictions, the feature information input to the rider's willingness to accept the order prediction model can be A, B, and C, the feature information input to the rider's waiting time prediction model can be B, C, D, and E, and the feature information input to the order delivery time prediction model can be A, B, C, D, and E.

[0085] Taking the multidimensional prediction model shown in Table 1 as an example, when at least two feature information of the current order are used as input to the rider's willingness to accept the order prediction model, the output of the prediction model is the prediction result of whether no rider is willing to accept the current order within 15 minutes; when at least two feature information of the current order are used as input to the rider's waiting time prediction model, the output of the prediction model is the prediction result of whether the rider's waiting time for the meal for the current order exceeds 10 minutes; when at least two feature information of the current order are used as input to the rider's delivery time prediction model, the output of the prediction model is the prediction result of whether the rider's delivery time for the current order exceeds 5 minutes.

[0086] The prediction results output by each prediction model can be represented by specific numerical values 0 or 1. For example, a positive prediction result can be represented by 1, and a negative prediction result can be represented by 0.

[0087] Step 103: Determine the delivery difficulty parameter of the current order based on the multi-dimensional prediction result.

[0088] Specifically, the prediction results output by multiple prediction models can be combined to determine the delivery difficulty parameter of the current order. The delivery difficulty parameter of the current order is used to characterize the delivery difficulty of the current order. If the positive prediction result is represented by 1 and the negative prediction result is represented by 0, the larger the delivery difficulty parameter of the current order, the higher the delivery difficulty of the current order.

[0089] Step 104: Determine a processing strategy for the current order based on the delivery difficulty parameter of the current order.

[0090] For example, the delivery difficulty parameter determined for the current order can provide a basis for online platforms to make decisions on pricing, rider scheduling, and merchant exposure ranking. For orders with high delivery difficulty, for example, higher delivery fees can be charged to the merchants or users involved in the order, rider scheduling can be delayed, or the merchants involved in the order can be ranked lower when displaying the merchant list to the user. Specific decisions can be made based on actual needs.

[0091] The above method collects at least two feature information items of the current order and uses the collected feature information as input to each prediction model in the multidimensional prediction model to obtain a multidimensional prediction result. Based on the multidimensional prediction result, a delivery difficulty parameter for the current order is determined. This delivery difficulty parameter is used to characterize the delivery difficulty of the current order. Because a multidimensional prediction model is used to determine the delivery difficulty of the current order, prediction models of different dimensions are used to output prediction results for the current order under different decision factors. This allows prediction of the delivery difficulty of the current order based on different decision factors, incorporating a wider and more comprehensive range of factors that characterize order delivery difficulty. This allows accurate determination of order delivery difficulty, providing a reliable basis for decision-making on the online platform.

[0092] The following describes the training method of the multidimensional prediction model shown in the embodiment of the present disclosure. Figure 2 As shown, the training method includes:

[0093] Step 201: Collect at least two feature information of historical orders and construct a sample.

[0094] Historical orders can be orders that have already been delivered. For example, at least two feature information items from a preset number of historical orders within a preset time period can be collected to construct a sample. The feature information of one historical order can be used as a sample. When training a model, the more samples taken, the more accurate the model prediction results, and correspondingly, the higher the training cost. Therefore, the number of samples used can be reasonably selected based on actual conditions. The at least two feature information items of the historical orders can be the same as the at least two feature information items of the current order in terms of both dimension and quantity. For details, please refer to the description of the at least two feature information items of the current order and will not be repeated here.

[0095] Step 202: Setting multi-dimensional decision factors.

[0096] Specifically, multidimensional decision factors can be set based on experience and actual needs. The multidimensional decision factors include at least two decision factors. The multidimensional decision factors may involve factors affecting the delivery difficulty in multiple dimensions, such as the rider's willingness to accept orders, the length of time the rider waits for meals, the length of time the rider delivers the meals, meal damage during delivery, and the length of time it takes to deliver the order. For example, the multidimensional decision factors set may include: whether there are no riders willing to accept the order within 15 minutes, whether the rider waits for meals for more than 10 minutes, whether the rider delivery time exceeds 5 minutes, whether the meal damage compensation for delivery exceeds 10 yuan, and whether the order delivery time exceeds 20 minutes. In this way, the order delivery difficulty prediction problem is transformed into a binary classification problem.

[0097] Step 203: binarize the actual decision results of the sample under the decision factors of each dimension in the multidimensional decision factors.

[0098] Since the sample is composed of feature information from historical orders, and historical orders are orders that have already been delivered, the actual decision results of the sample under the decision factors of each dimension are known. For example, the historical order involved in the sample is a milk tea transaction order. After the user placed the order, a rider accepted the order within 3 minutes. After the rider arrived at the merchant's store, he waited for 2 minutes before picking up the milk tea. The rider took 25 minutes to reach the user and delivered the milk tea to the user in 1 minute. Therefore, the actual decision result of this sample under the decision factor "whether there is no rider willing to accept the order within 15 minutes" is no, the actual decision result under the decision factor "whether the rider waits for the food for more than 10 minutes" is no, the actual decision result under the decision factor "whether the order delivery time exceeds 20 minutes" is yes, and the actual decision result under the decision factor "whether the rider delivery time exceeds 5 minutes" is no.

[0099] Since this solution involves multiple prediction models, in order to facilitate the subsequent integration of the prediction results of multiple models, the actual decision results of the samples under the decision factors of each dimension in the multidimensional decision factors can be binarized. For example, the positive actual decision results are represented by 1, and the negative actual decision results are represented by 0.

[0100] Step 204 : Using the sample and the actual decision result of the sample under the decision factors of each dimension in the multidimensional decision factors as training data, model training is performed to obtain the multidimensional prediction model.

[0101] For example, the multidimensional prediction model can be a model implemented using algorithms such as Gradient Boosting Decision Tree (GBDT), deep learning, and random forest. The various prediction models included in the multidimensional prediction model can be implemented using the same algorithm or different algorithms. The sample features used in the training of the various prediction models included in the multidimensional prediction model can be the same or different, depending on specific needs.

[0102] The model training process is to find the optimal model parameters of the preset classification model under the decision factors of each dimension, so that the output results of the sample in the preset classification model are as close as possible to the actual decision results of the sample under the decision factors of each dimension. When the output results of the sample in the preset classification model are as close as possible to the actual decision results of the sample under the decision factors of one dimension, a prediction model is obtained. By traversing the decision factors of all dimensions, the multidimensional prediction model is obtained.

[0103] Figure 3 is a flow chart of another order processing method shown in an embodiment of the present disclosure, such as Figure 3 As shown, the method includes:

[0104] Step 301: Collect at least two feature information of the current order.

[0105] Step 302: Use the feature information as input to each prediction model in the multidimensional prediction model to obtain a multidimensional prediction result.

[0106] Steps 301 and 302 are the same as the aforementioned steps 101 and 102 and will not be repeated here.

[0107] Step 303: Obtain the model weight of the prediction model of each dimension in the pre-configured multi-dimensional prediction model.

[0108] In a specific implementation, the backend management personnel of the network platform can pre-configure the model weights for the prediction models of each dimension in the multi-dimensional prediction model according to business needs. For example, if the user experience is to be improved, the prediction models that can reflect the user experience (such as the order delivery time prediction model, the delivery timeout prediction model, the delivery meal loss prediction model, etc.) can be configured with higher model weights, and other prediction models can be configured with lower model weights. For another example, if the delivery efficiency is to be improved, the prediction models that can reflect the delivery efficiency (such as the capacity occupancy time prediction model, the package combination probability prediction model, etc.) can be configured with higher model weights, and other prediction models can be configured with lower model weights. The backend server can obtain the model weights pre-configured for the prediction models of each dimension in the multi-dimensional prediction model. In particular, if the model weight of a prediction model is configured to 0, it can be said that the prediction model is not enabled.

[0109] Step 304 : Determine the delivery difficulty parameter of the current order based on the multi-dimensional prediction results and the model weight of the prediction model of each dimension.

[0110] Specifically, the fusion algorithm F can be set to process the multidimensional prediction result and the model weight of the prediction model of each dimension, then the delivery difficulty parameter O of the current order = F(R, W), where O represents the delivery difficulty parameter, R represents the multidimensional prediction result, and W represents the model weight.

[0111] In a specific embodiment, the fusion algorithm F can be a maximum fusion algorithm, that is, the prediction result of the current order in each dimension can be multiplied by the model weight of the prediction model of the corresponding dimension, and the maximum value of the obtained results can be taken as the delivery difficulty parameter of the current order.

[0112] In a specific embodiment, the fusion algorithm F can also be a linear weighted fusion algorithm, that is, the prediction result of the current order in each dimension can be multiplied by the model weight of the prediction model of the corresponding dimension and then summed up, and the resulting sum value is used as the delivery difficulty parameter of the current order.

[0113] The delivery difficulty parameter is used to characterize the delivery difficulty of the current order. If a positive prediction result is represented by 1 and a negative prediction result is represented by 0, the larger the delivery difficulty parameter of the current order, the higher the delivery difficulty of the current order.

[0114] Step 305: Determine a processing strategy for the current order based on the delivery difficulty parameter of the current order.

[0115] For example, the delivery difficulty parameter determined for the current order can provide a basis for online platforms to make decisions on pricing, rider scheduling, and merchant exposure ranking. For orders with high delivery difficulty, for example, higher delivery fees can be charged to the merchants or users involved in the order, rider scheduling can be delayed, or the merchants involved in the order can be ranked lower when displaying the merchant list to the user. Specific decisions can be made based on actual needs.

[0116] Using the above method, a multidimensional prediction model is adopted when determining the delivery difficulty of the current order. The prediction models of different dimensions are used to output the prediction results of the current order under the decision factors of different dimensions, so as to predict the delivery difficulty of the current order from the decision factors of different dimensions, taking into account more and more comprehensive factors that characterize the difficulty of order delivery, thereby achieving accurate determination of the order delivery difficulty and providing a reliable basis for the decision-making of the network platform.

[0117] Furthermore, by configuring model weights for each prediction model, it is possible to differentiate the difficulty of order delivery according to different business needs and scenarios, making it more versatile.

[0118] Figure 4 This is a flow chart of another order processing method shown in an embodiment of the present disclosure. Figure 4 As shown, the method includes:

[0119] Step 401: Collect at least two feature information of the current order.

[0120] Step 402: Use the feature information as input to each prediction model in the multi-dimensional prediction model to obtain a multi-dimensional prediction result.

[0121] Steps 401 and 402 are the same as the aforementioned steps 101 and 102 and will not be repeated here.

[0122] Step 403: Determine the current transportation capacity supply and demand status based on the planned delivery data and historical order data for the area where the current order is located.

[0123] Planned delivery data includes: the planned delivery time for orders in the current region, the planned delivery completion rate for orders in the current region; historical order data includes: the rider acceptance rate for orders in the current region, and the number of undelivered orders in the current region. Generally speaking, the shorter the planned delivery time for orders in the current region, the higher the planned delivery completion rate for orders in the current region, the higher the rider acceptance rate for orders in the current region, or the greater the number of undelivered orders in the current region, the greater the demand for transportation capacity.

[0124] Specifically, the current transportation capacity supply and demand status determined based on the planned delivery data and historical order data of the area where the current order is located may include: insufficient transportation capacity supply, balanced transportation capacity supply and demand, and sufficient transportation capacity supply.

[0125] Step 404: Determine the current delivery scenario based on the current transportation capacity supply and demand status.

[0126] When the supply of transportation capacity is insufficient, the platform will pay more attention to delivery efficiency, and the current delivery scenario can be determined as a delivery efficiency priority scenario; when the supply of transportation capacity is sufficient, the platform will pay more attention to user experience, and the current delivery scenario can be determined as a user experience priority scenario; when the supply and demand of transportation capacity is balanced, the platform may pay attention to the rider experience, and the current delivery scenario can be determined as a rider experience priority scenario.

[0127] Step 405 : configuring a model weight for the prediction model of each dimension in the multi-dimensional prediction model according to the current delivery scenario.

[0128] In a specific embodiment, the trained prediction models may be pre-grouped according to the delivery scenario. The specific groupings are shown in Table 2 below:

[0129]

[0130] Table 2

[0131] After determining the current delivery scenario, the model weights can be automatically configured for the prediction models of the corresponding groups according to the current delivery scenario. For example, if the current delivery scenario is a delivery efficiency priority scenario, the first model weight can be configured for the prediction model of the group corresponding to the delivery efficiency priority scenario, and the second model weight can be configured for the prediction models of other groups. If the current delivery scenario is a user experience priority scenario, the first model weight can be configured for the prediction model of the group corresponding to the user experience priority scenario, and the second model weight can be configured for other prediction models. If the current delivery scenario is a rider experience priority scenario, the first model weight can be configured for the prediction model of the group corresponding to the rider experience priority scenario, and the second model weight can be configured for other prediction models, wherein the first model weight is greater than the second model weight. In particular, if the model weights of certain prediction models are configured to 0, it can be indicated that the prediction model of this group is not enabled.

[0132] Step 406 : Determine the delivery difficulty parameter of the current order based on the multi-dimensional prediction results and the model weight of the prediction model of each dimension.

[0133] Specifically, the fusion algorithm F can be set to process the multidimensional prediction result and the model weight of the prediction model of each dimension, then the delivery difficulty parameter O of the current order = F(R, W), where O represents the delivery difficulty parameter, R represents the multidimensional prediction result, and W represents the model weight.

[0134] In a specific embodiment, the fusion algorithm F can be a maximum fusion algorithm, that is, the prediction result of the current order in each dimension can be multiplied by the model weight of the prediction model of the corresponding dimension, and the maximum value of the obtained results can be taken as the delivery difficulty parameter of the current order.

[0135] In a specific embodiment, the fusion algorithm F can also be a linear weighted fusion algorithm, that is, the prediction result of the current order in each dimension can be multiplied by the model weight of the prediction model of the corresponding dimension and then summed up, and the resulting sum value is used as the delivery difficulty parameter of the current order.

[0136] The delivery difficulty parameter is used to characterize the delivery difficulty of the current order. If a positive prediction result is represented by 1 and a negative prediction result is represented by 0, the larger the delivery difficulty parameter of the current order, the higher the delivery difficulty of the current order.

[0137] Step 407: Determine a processing strategy for the current order based on the delivery difficulty parameter of the current order.

[0138] For example, the delivery difficulty parameter determined for the current order can provide a basis for online platforms to make decisions on pricing, rider scheduling, and merchant exposure ranking. For orders with high delivery difficulty, for example, higher delivery fees can be charged to the merchants or users involved in the order, rider scheduling can be delayed, or the merchants involved in the order can be ranked lower when displaying the merchant list to the user. Specific decisions can be made based on actual needs.

[0139] Using the above method, a multidimensional prediction model is adopted when determining the delivery difficulty of the current order. The prediction models of different dimensions are used to output the prediction results of the current order under the decision factors of different dimensions, so as to predict the delivery difficulty of the current order from the decision factors of different dimensions, taking into account more and more comprehensive factors that characterize the difficulty of order delivery, thereby achieving accurate determination of the order delivery difficulty and providing a reliable basis for the decision-making of the network platform.

[0140] Furthermore, by configuring model weights for each prediction model, it is possible to differentiate the difficulty of order delivery according to different business needs and scenarios, making it more versatile.

[0141] The following is a specific example to illustrate the order processing method shown in the embodiment of the present disclosure. Figure 5 For example, the collected at least two characteristic information of the current order include M features, namely feature 1, feature 2...feature M, and the trained multidimensional prediction models include N models, namely model 1, model 2...model N, where M and N are integers greater than 2. Then, these M features can be used as the input of each of the N models, and N prediction results, namely prediction result R1, prediction result R2...prediction result RN, can be obtained. If the model weights of the N models are W1, W2,...WN respectively, then the delivery difficulty parameter of the current order determined according to these N prediction results and N model weights can be expressed as: O=F(Ri,Wi), where O represents the delivery difficulty parameter, F represents the fusion algorithm, R represents the prediction result, W represents the model weight, and i∈[1,N].

[0142] If the fusion algorithm selects the maximum fusion algorithm, that is, multiplying the prediction result of the current order in each dimension with the model weight of the prediction model of the corresponding dimension, and taking the maximum value of the obtained results as the delivery difficulty parameter of the current order, then O=max(R1*W1,R2*W2…RN*WN). If the fusion algorithm selects the linear weighted fusion algorithm, that is, multiplying the prediction result of the current order in each dimension with the model weight of the prediction model of the corresponding dimension, and taking the sum as the delivery difficulty parameter of the current order, then

[0143] Finally, the processing strategy for the current order can be determined based on the delivery difficulty parameters of the current order. For example, the delivery difficulty parameters of the current order can provide a decision-making basis for the network platform to determine pricing, rider scheduling, merchant exposure ranking, etc. For orders with high delivery difficulty, for example, higher delivery fees can be charged to the merchants or users involved in the order, or rider dispatch can be delayed, or the merchants involved in the order can be ranked relatively low when displaying the merchant list to the user. Specific decisions can be made based on actual needs.

[0144] Figure 6 This is a structural block diagram of an order processing device shown in an embodiment of the present disclosure. Figure 6 As shown, the apparatus 500 may include:

[0145] The collection module 501 is configured to collect at least two feature information of the current order;

[0146] Prediction module 502 is configured to use the feature information as input to each prediction model in a multi-dimensional prediction model to obtain a multi-dimensional prediction result, wherein the multi-dimensional prediction model includes at least two prediction models, and different prediction models are used to output prediction results for the current order under decision factors of different dimensions;

[0147] A first determining module 503 is configured to determine a delivery difficulty parameter of the current order based on the multi-dimensional prediction result, where the delivery difficulty parameter is used to represent the delivery difficulty of the current order;

[0148] The second determining module 504 is configured to determine a processing strategy for the current order according to the delivery difficulty parameter of the current order.

[0149] In one embodiment, if Figure 7 As shown, the first determining module 503 includes:

[0150] An acquisition submodule 5031 is configured to acquire a model weight of a prediction model of each dimension in the multidimensional prediction model;

[0151] The determination submodule 5032 is configured to determine the delivery difficulty parameter of the current order based on the multi-dimensional prediction results and the model weight of the prediction model of each dimension.

[0152] In one embodiment, the acquisition submodule 5031 includes:

[0153] The weight acquisition submodule 50311 is configured to obtain the model weight of the prediction model of each dimension in the pre-configured multi-dimensional prediction model.

[0154] In one embodiment, the acquisition submodule 5031 includes:

[0155] The scene determination submodule 50312 is configured to determine the current delivery scene;

[0156] The weight configuration submodule 50313 is configured to configure the model weight for the prediction model of each dimension in the multi-dimensional prediction model according to the current delivery scenario.

[0157] In one embodiment, the scenario determination submodule 50312 is configured to determine the current transportation supply and demand status based on the planned delivery data and historical order data of the area where the current order is located; and determine the current delivery scenario based on the current transportation supply and demand status.

[0158] In one embodiment, the determination submodule 5032 is configured to specifically multiply the prediction result of the current order in each dimension by the model weight of the prediction model of the corresponding dimension, and take the maximum value of the obtained results as the delivery difficulty parameter of the current order.

[0159] In one embodiment, the determination submodule 5032 is configured to specifically multiply the prediction result of the current order in each dimension by the model weight of the prediction model of the corresponding dimension and then sum them up, and use the resulting sum as the delivery difficulty parameter of the current order.

[0160] In one embodiment, the device further comprises:

[0161] A construction module 505 is configured to collect at least two feature information of historical orders and construct a sample;

[0162] A setting module 506 configured to set multi-dimensional decision factors;

[0163] The training module 507 is configured to use the sample and the actual decision result of the sample under the decision factor of each dimension in the multidimensional decision factor as training data to perform model training to obtain the multidimensional prediction model.

[0164] In one embodiment, the device further comprises:

[0165] The binarization module 508 is configured to binarize the actual decision result of the sample under the decision factor of each dimension in the multi-dimensional decision factor.

[0166] Regarding the apparatus in the above embodiment, the specific manner in which each module performs operations has been described in detail in the embodiment of the method, and will not be elaborated here.

[0167] In summary, by collecting at least two feature information of the current order and using the collected feature information as the input of each prediction model in the multidimensional prediction model, a multidimensional prediction result is obtained. The delivery difficulty parameter of the current order is determined based on the multidimensional prediction result. The delivery difficulty parameter is used to characterize the delivery difficulty of the current order, and then the processing strategy of the current order is determined based on the delivery difficulty parameter of the current order. Because a multidimensional prediction model is used to determine the delivery difficulty of the current order, prediction models of different dimensions are used to output prediction results of the current order under decision factors of different dimensions, thereby predicting the delivery difficulty of the current order from the decision factors of different dimensions, taking into account more and more comprehensive factors that characterize the difficulty of order delivery, thereby accurately determining the difficulty of order delivery, and providing a reliable basis for the decision-making of the network platform.

[0168] Furthermore, by configuring model weights for each prediction model, it is possible to differentiate the difficulty of order delivery according to different business needs and scenarios, making it more versatile.

[0169] For example, Figure 8 FIG. 8 is a structural diagram of an electronic device 800 according to an embodiment of the present disclosure. Figure 8 The electronic device 800 includes a processor 801, which may be one or more, and a memory 802 for storing a computer program executable by the processor 801. The computer program stored in the memory 802 may include one or more modules, each corresponding to a set of instructions. Furthermore, the processor 801 may be configured to execute the computer program to perform the aforementioned method for determining the difficulty of order delivery.

[0170] In addition, the electronic device 800 may further include a power supply component 803 and a communication component 804. The power supply component 803 may be configured to perform power management of the electronic device 800, and the communication component 804 may be configured to implement communication, such as wired or wireless communication, of the electronic device 800. In addition, the electronic device 800 may further include an input / output (I / O) interface 905. The electronic device 800 may operate based on an operating system stored in the memory 802, such as Windows Server™, Mac OS X™, Unix™, Linux™, etc.

[0171] In another exemplary embodiment, a computer-readable storage medium including program instructions is also provided. When executed by a processor, the program instructions implement the steps of the above-described order processing method. For example, the computer-readable storage medium may be the aforementioned memory 802 including the program instructions. The program instructions may be executed by the processor 801 of the electronic device 800 to perform the above-described order processing method.

[0172] The preferred embodiments of the present disclosure are described in detail above in conjunction with the accompanying drawings. However, the present disclosure is not limited to the specific details of the above embodiments. Within the technical concept of the present disclosure, various simple modifications can be made to the technical solutions of the present disclosure, and these simple modifications all fall within the scope of protection of the present disclosure.

[0173] It should also be noted that the various specific technical features described in the above specific embodiments can be combined in any appropriate manner without contradiction. In order to avoid unnecessary repetition, the present disclosure will not further describe various possible combinations.

Claims

1. An order processing method, characterized in that: The method comprises: Collect at least two characteristic information of the current order; Using the feature information as input to each prediction model in a multidimensional prediction model to obtain a multidimensional prediction result, wherein the multidimensional prediction model includes at least two prediction models, and different prediction models are used to output prediction results of the current order under decision factors of different dimensions; Determine the current delivery scenario based on the current supply and demand status of transportation capacity, and configure the model weight for the prediction model of each dimension in the multidimensional prediction model based on the current delivery scenario. In the state of insufficient capacity supply, the current delivery scenario is determined to be a delivery efficiency priority scenario, and a first model weight is configured for the prediction model of the corresponding group of the delivery efficiency priority scenario, and the prediction model of the corresponding group of the delivery efficiency priority scenario includes one of a capacity occupancy time prediction model and a package combination probability prediction model; or in the state of sufficient capacity supply, the current delivery scenario is determined to be a user experience priority scenario, and a first model weight is configured for the prediction model of the corresponding group of the user experience priority scenario, and the prediction model of the corresponding group of the user experience priority scenario includes one of an order delivery time prediction model, a delivery timeout prediction model, a delivery last-order prediction model, and a delivery meal loss prediction model; or in the state of balanced capacity supply and demand, the current delivery scenario is determined to be a rider experience priority scenario, and a first model weight is configured for the prediction model of the corresponding group of the rider experience priority scenario, and the prediction model of the corresponding group of the rider experience priority scenario includes one of a rider order acceptance willingness prediction model, a rider waiting time prediction model, and a rider delivery time prediction model; Determining a delivery difficulty parameter of the current order based on the multi-dimensional prediction result, where the delivery difficulty parameter is used to characterize the delivery difficulty of the current order; A processing strategy for the current order is determined based on a delivery difficulty parameter of the current order.

2. The order processing method according to claim 1, characterized in that: Determining the delivery difficulty parameter of the current order according to the multi-dimensional prediction result includes: Obtaining a model weight of a prediction model for each dimension in the multidimensional prediction model; The delivery difficulty parameter of the current order is determined based on the multi-dimensional prediction results and the model weight of the prediction model of each dimension.

3. The order processing method according to claim 2, characterized in that: The obtaining of the model weight of the prediction model of each dimension in the multidimensional prediction model includes: Obtain a model weight of a prediction model for each dimension in the pre-configured multi-dimensional prediction model.

4. The order processing method according to claim 2, characterized in that: The obtaining of the model weight of the prediction model of each dimension in the multidimensional prediction model includes: Determine the current delivery scenario; A model weight is configured for the prediction model of each dimension in the multidimensional prediction model according to the current delivery scenario.

5. The order processing method according to claim 4, characterized in that: Determining the current delivery scenario includes: Determine the current supply and demand status of transportation capacity based on the planned delivery data and historical order data for the area where the current order is located; The current delivery scenario is determined based on the current transportation capacity supply and demand status.

6. The order processing method according to any one of claims 2 to 5, characterized in that: The determining of the delivery difficulty parameter of the current order based on the multi-dimensional prediction results and the model weight of the prediction model of each dimension includes: The prediction result of the current order in each dimension is multiplied by the model weight of the prediction model of the corresponding dimension, and the maximum value of the obtained results is taken as the delivery difficulty parameter of the current order.

7. The order processing method according to any one of claims 2 to 5, characterized in that: The determining of the delivery difficulty parameter of the current order based on the multi-dimensional prediction results and the model weight of the prediction model of each dimension includes: The prediction result of the current order in each dimension is multiplied by the model weight of the prediction model of the corresponding dimension, and the sum is taken as the delivery difficulty parameter of the current order.

8. The order processing method according to any one of claims 1 to 5, characterized in that: The multidimensional prediction model is trained according to the following method: Collect at least two feature information of historical orders and construct a sample; Set up multidimensional decision factors; The sample and the actual decision results of the sample under the decision factors of each dimension in the multidimensional decision factors are used as training data to perform model training to obtain the multidimensional prediction model.

9. The order processing method according to claim 8, characterized in that: After setting up the multidimensional decision factors, it also includes: Binarize the actual decision results of the sample under the decision factors of each dimension in the multidimensional decision factors.

10. An order processing device, characterized in that: The device comprises: A collection module is configured to collect at least two feature information of the current order; The prediction module is configured to use the feature information as input to each prediction model in a multidimensional prediction model to obtain a multidimensional prediction result, wherein the multidimensional prediction model includes at least two prediction models, and different prediction models are used to output prediction results of the current order under decision factors of different dimensions, determine the current delivery scenario based on the current supply and demand status of transportation capacity, and configure model weights for the prediction models of each dimension in the multidimensional prediction model based on the current delivery scenario. In the state of insufficient capacity supply, the current delivery scenario is determined to be a delivery efficiency priority scenario, and a first model weight is configured for the prediction model of the corresponding group of the delivery efficiency priority scenario, and the prediction model of the corresponding group of the delivery efficiency priority scenario includes one of a capacity occupancy time prediction model and a package combination probability prediction model; or in the state of sufficient capacity supply, the current delivery scenario is determined to be a user experience priority scenario, and a first model weight is configured for the prediction model of the corresponding group of the user experience priority scenario, and the prediction model of the corresponding group of the user experience priority scenario includes one of an order delivery time prediction model, a delivery timeout prediction model, a delivery last-order prediction model, and a delivery meal loss prediction model; or in the state of balanced capacity supply and demand, the current delivery scenario is determined to be a rider experience priority scenario, and a first model weight is configured for the prediction model of the corresponding group of the rider experience priority scenario, and the prediction model of the corresponding group of the rider experience priority scenario includes one of a rider order acceptance willingness prediction model, a rider waiting time prediction model, and a rider delivery time prediction model; A first determining module is configured to determine a delivery difficulty parameter of the current order based on the multi-dimensional prediction result, wherein the delivery difficulty parameter is used to represent the delivery difficulty of the current order; The second determination module is configured to determine a processing strategy for the current order according to a delivery difficulty parameter of the current order.

11. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the steps of the order processing method according to any one of claims 1 to 9 are implemented.

12. An electronic device, characterized in that: include: a memory having a computer program stored thereon; A processor, configured to execute the computer program in the memory to implement the steps of the order processing method according to any one of claims 1 to 9.

Images