Response duration prediction model training and prediction method, device, medium and product

Abstract

The embodiments of the present disclosure disclose a method and device for training and predicting a response duration prediction model, a medium and a product. The method comprises: obtaining a set of sample data from a sample set as a training set, wherein the sample data comprises feature information of a sample service request and a real response duration of the sample service request; training a plurality of initial multi-objective models based on the sample data in the training set to obtain a plurality of multi-objective models, wherein the learning objectives of the multi-objective models comprise a probability that a service request response duration is within a predetermined range and a predicted response duration of a service request whose response duration is within the predetermined range, and each multi-objective model corresponds to a different predetermined range. The technical solution can more accurately predict the response duration of a service request.

Description

Technical Field

[0001] This disclosure relates to the field of Internet technology, specifically to a method, device, medium, and product for training and predicting a response duration prediction model. Background Technology

[0002] For ride-hailing platforms, passenger order response time is highly valuable information, reflecting the response status of orders and the surrounding area. If the response time can be predicted when a passenger places an order, the platform can decide whether to provide subsidies. For example, subsidies can be sent to long-waiting orders that are ultimately canceled by passengers or the system to encourage drivers to accept orders quickly, thus improving the overall order response rate. Currently, there are many methods for predicting order response times, such as LightGradient Boosting Machine (GBM) regression models and Deep Neural Networks (DNNs). These methods predict order response times by directly learning the relationship between the features of all sample orders and their response times. However, order response times themselves have a long-tailed distribution with a wide range, making it difficult for these models to achieve good learning results, resulting in low accuracy in their predictions. Summary of the Invention

[0003] This disclosure provides a method, device, medium, and product for training and predicting a predictive model for response duration.

[0004] In a first aspect, this disclosure provides a method for training a prediction model for response duration.

[0005] Specifically, the method for training the prediction model for response duration includes:

[0006] A set of sample data is obtained from the sample set as a training set, and the sample data includes the feature information of the sample service request and the actual response time of the sample service request.

[0007] Based on the sample data in the training set, multiple initial multi-objective models are trained to obtain multiple multi-objective models. The learning objectives of the multi-objective models include the probability that the service request response time is within a predetermined range, and the predicted response time of service requests with response times within the predetermined range. The predetermined range is different for each multi-objective model.

[0008] In one possible implementation, the step of training multiple initial multi-objective models based on sample data in the training set to obtain multiple multi-objective models includes:

[0009] The classification branch of the initial multi-objective model is trained based on all sample data in the training set, and the regression branch of the initial multi-objective model is trained based on sample data in the training set whose actual response time is within a predetermined range, to obtain the multi-objective model.

[0010] The output of the classification branch in the multi-objective model is the probability that the service request response time is within a predetermined range, and the output of the regression branch in the multi-objective model is the predicted response time of the service request whose response time is within a predetermined range.

[0011] In one possible implementation, the feature information of the sample service request includes the feature information of the service requested by the sample service request and the feature information of the corresponding service requester.

[0012] In one possible implementation, the predetermined range corresponding to the multi-objective model includes less than or equal to K seconds, and the K value is different for each multi-objective model.

[0013] Secondly, this disclosure provides a method for predicting response duration.

[0014] Specifically, the method for predicting the response duration includes:

[0015] Obtain the characteristic information of the current service request;

[0016] The feature information of the current service request is input into multiple preset multi-objective models to obtain the output results of each multi-objective model. The output results of the multi-objective model include the probability that the response time of the current service request is within the corresponding predetermined range, and the predicted response time of the current service request when the response time is within the predetermined range. The predetermined range is different for each multi-objective model.

[0017] The predicted response time of the multi-objective model with the highest probability among the output results of the multiple multi-objective models is taken as the predicted response time of the current service request.

[0018] In one possible implementation, the method further includes:

[0019] The multiple multi-objective models are obtained based on the above-described method for training prediction models of response duration.

[0020] Thirdly, this disclosure provides a service request response processing method.

[0021] Specifically, the service request response processing method includes:

[0022] Based on the above method for predicting response duration, the predicted response duration for the current service request is obtained;

[0023] Based on the predicted response time of the current service request, a service provider is assigned to respond to the current service request at the appropriate time.

[0024] In one possible implementation, it also includes:

[0025] If the predicted response time of the current service request exceeds the preset time, then a corresponding service value-added service is configured for the current service request based on the predicted response time of the current service request. The service value corresponding to the current service request includes the original service value and the service value-added service.

[0026] Fourthly, this disclosure provides a predictive model training device for response duration.

[0027] Specifically, the predictive model training device for response duration includes:

[0028] The first acquisition module is configured to acquire a set of sample data from the sample set as a training set, wherein the sample data includes feature information of sample service requests and the actual response time of sample service requests.

[0029] The training module is configured to train multiple initial multi-objective models based on sample data in the training set to obtain multiple multi-objective models. The learning objectives of the multi-objective models include the probability that the service request response time is within a predetermined range, and the predicted response time of service requests with response times within the predetermined range. The predetermined range is different for each multi-objective model.

[0030] In one possible implementation, the training module is configured as follows:

[0031] The classification branch of the initial multi-objective model is trained based on all sample data in the training set, and the regression branch of the initial multi-objective model is trained based on sample data in the training set whose actual response time is within a predetermined range, to obtain the multi-objective model.

[0032] The output of the classification branch in the multi-objective model is the probability that the service request response time is within a predetermined range, and the output of the regression branch in the multi-objective model is the predicted response time of the service request whose response time is within a predetermined range.

[0033] In one possible implementation, the feature information of the sample service request includes the feature information of the service requested by the sample service request and the feature information of the corresponding service requester.

[0034] In one possible implementation, the predetermined range corresponding to the multi-objective model includes less than or equal to K seconds, and the K value corresponding to each multi-objective model is different, wherein K is greater than 0.

[0035] Fifthly, this disclosure provides a predictive model training device for response duration.

[0036] Specifically, the predictive model training device for response duration includes:

[0037] The second acquisition module is configured to acquire the characteristic information of the current service request;

[0038] The prediction module is configured to input the feature information of the current service request into multiple preset multi-objective models respectively, and obtain the output result of each multi-objective model. The output result of the multi-objective model includes the probability that the response time of the current service request is within a corresponding predetermined range, and the predicted response time of the current service request when the response time is within the predetermined range. The predetermined range is different for each multi-objective model.

[0039] The determination module is configured to determine the predicted response time of the multi-objective model with the highest probability among the output results of the multiple multi-objective models as the predicted response time of the current service request.

[0040] In one possible implementation, the second acquisition module is configured as follows:

[0041] The multiple multi-objective models are obtained based on the above-described method for training prediction models of response duration.

[0042] Sixthly, this disclosure provides a service request response processing apparatus.

[0043] Specifically, the service request response processing device includes:

[0044] The third acquisition module is configured to acquire the predicted response time of the current service request;

[0045] The allocation module is configured to allocate a service provider to respond to the current service request at the appropriate time based on the predicted response time of the current service request.

[0046] In one possible implementation, the device further includes:

[0047] The configuration module is configured to configure a corresponding service value-added service for the current service request based on the predicted response time of the current service request if the predicted response time of the current service request exceeds a preset time. The service value corresponding to the current service request includes the original service value and the service value-added service.

[0048] In a seventh aspect, embodiments of this disclosure provide an electronic device including a memory and a processor, the memory being used to store one or more computer instructions supporting the above-described apparatus in performing the above-described methods, and the processor being configured to execute the computer instructions stored in the memory.

[0049] Eighthly, embodiments of this disclosure provide a computer-readable storage medium having computer instructions stored thereon, which, when executed by a processor, implement the steps of the method described in any of the preceding aspects.

[0050] Ninthly, embodiments of this disclosure provide a computer program product, including a computer program / instructions, wherein the computer program / instructions, when executed by a processor, implement the steps of the method described in any of the preceding aspects.

[0051] The technical solutions provided in this disclosure may have the following beneficial effects:

[0052] The above technical solution can train multiple multi-objective models to predict response duration. The learning objectives of each multi-objective model include the probability that the service request response duration falls within a predetermined range, and the predicted response duration of service requests whose response duration falls within the predetermined range. Each multi-objective model corresponds to a different predetermined range. In this way, the large-scale distribution of response duration is transformed into duration distributions within different predetermined ranges, allowing different multi-objective models to learn the relationship between the feature information of sample service requests and response durations on different duration distributions, resulting in better learning performance for these multi-objective models. Furthermore, the probability that the service request response duration falls within the predetermined range output by the multi-objective models can be used as a basis to determine the response duration distribution of the service request. When the response duration falls within the predetermined range with the highest probability, the predicted response duration of the corresponding service request is the final prediction result. Since the response duration prediction for service requests distributed within different predetermined ranges is performed by the corresponding multi-objective models, the prediction results are more accurate.

[0053] It should be understood that the above general description and the following detailed description are exemplary and explanatory only, and are not intended to limit the embodiments of this disclosure. Attached Figure Description

[0054] Other features, objects, and advantages of embodiments of this disclosure will become more apparent from the following detailed description of non-limiting implementations, taken in conjunction with the accompanying drawings. In the drawings:

[0055] Figure 1 A flowchart is shown for a method of training a prediction model for response duration according to an embodiment of the present disclosure.

[0056] Figure 2 A schematic diagram illustrating the training and testing process of a prediction model for response time according to an embodiment of the present disclosure is shown.

[0057] Figure 3 A flowchart illustrating a method for predicting response time according to an embodiment of the present disclosure is shown.

[0058] Figure 4 A flowchart illustrating a service request response processing method according to an embodiment of the present disclosure is shown.

[0059] Figure 5 A structural block diagram of a predictive model training apparatus for response duration according to an embodiment of the present disclosure is shown.

[0060] Figure 6 A structural block diagram of a response duration prediction device according to an embodiment of the present disclosure is shown.

[0061] Figure 7 A structural block diagram of a service request response processing apparatus according to an embodiment of the present disclosure is shown.

[0062] Figure 8 A structural block diagram of an electronic device according to an embodiment of the present disclosure is shown.

[0063] Figure 9 This is a schematic diagram of the structure of a computer system suitable for implementing the method according to embodiments of the present disclosure. Detailed Implementation

[0064] Hereinafter, exemplary embodiments of the present disclosure will be described in detail with reference to the accompanying drawings to enable those skilled in the art to readily implement them. Furthermore, for clarity, portions unrelated to the description of the exemplary embodiments have been omitted from the drawings.

[0065] In embodiments disclosed herein, it should be understood that terms such as “comprising” or “having” are intended to indicate the presence of features, numbers, steps, behaviors, components, portions or combinations thereof disclosed herein, and are not intended to exclude the possibility that one or more other features, numbers, steps, behaviors, components, portions or combinations thereof are present or added.

[0066] It should also be noted that, unless otherwise specified, the embodiments and features described in this disclosure can be combined with each other. The embodiments of this disclosure will now be described in detail with reference to the accompanying drawings and examples.

[0067] As mentioned above, passenger order response time is highly valuable information for ride-hailing platforms, reflecting the response status of orders and the surrounding area. If the response time can be predicted when a passenger places an order, the platform can decide whether to provide subsidies. For example, subsidies can be sent to long-waiting orders that are ultimately canceled by passengers or the system to encourage drivers to accept orders quickly, thus improving the overall order response rate. Currently, there are many methods for predicting order response times, such as the Light Gradient Boosting Machine (GBM) regression model and Deep Neural Networks (DNNs) in machine learning. These methods predict order response times by directly learning the relationship between the features of all sample orders and their response times. However, order response times themselves have a long-tailed distribution with a wide range, making it difficult for these models to achieve good learning results, resulting in low accuracy in the predictions.

[0068] To address the aforementioned issues, this disclosure provides a training scheme for a response duration prediction model. This scheme trains multiple multi-objective models to predict response durations. Each multi-objective model's learning objective includes the probability that the service request response duration falls within a predetermined range, and the predicted response duration for service requests with response durations within the predetermined range. Each multi-objective model corresponds to a different predetermined range. This transforms the broad distribution of response durations into duration distributions within different predetermined ranges, allowing different multi-objective models to learn the relationship between the feature information of sample service requests and their response durations on different duration distributions. This results in better learning performance for these multi-objective models. Furthermore, the probability of a service request response duration falling within a predetermined range, output by the multi-objective models, can be used as the basis to determine the response duration distribution of the service request. The predicted response duration of the service request corresponding to the distribution with the highest probability within the predetermined range is the final prediction result. Since service requests distributed within different predetermined ranges are predicted by corresponding multi-objective models, the prediction results are more accurate.

[0069] Figure 1 A flowchart illustrating a method for training a prediction model for response duration according to an embodiment of the present disclosure is shown, as follows: Figure 1 As shown, the method for training the prediction model for response duration includes the following steps S101-S102:

[0070] In step S101, a set of sample data is obtained from the sample set as a training set. The sample data includes the feature information of the sample service request and the actual response time of the sample service request.

[0071] In step S102, based on the sample data in the training set, multiple initial multi-objective models are trained to obtain multiple multi-objective models. The output of each multi-objective model includes the probability that the service request response time is within a predetermined range, and the predicted response time of the service request with a response time within the predetermined range. The predetermined range is different for each multi-objective model.

[0072] In one possible implementation, the subject executing the prediction model training method for response duration can be a computer, a server, or a server cluster, etc.

[0073] In one possible implementation, the service refers to a service provided by a service provider to a service requester. When a service requester wants a certain service, they can initiate a service request on the service platform. The timer starts when the service requester initiates the service request and stops when a service provider agrees to provide the service, the service requester cancels the service request, or the service platform cancels the service request. This timeout period is the response time of the service. For example, the service could be a ride-hailing service, where the service requester is the passenger and the service provider is the driver. If a driver accepts the order after the passenger places it, the response time is the time between the passenger placing the order and the driver accepting it. If the passenger cancels the order after waiting for a period of time, the response time is the time between the passenger placing the order and the cancellation. If the passenger waits for a long time and no one accepts the order, the ride-hailing platform cancels the order, the response time is the time between the passenger placing the order and the platform canceling the order. Of course, the service could also be a consultation service on a human Q&A platform, a delivery service on a delivery platform, etc.

[0074] In one possible implementation, the sample data refers to data on services whose service requests have been answered or canceled. This sample set contains a massive amount of sample data, and a portion of the sample data can be selected as a training set to train multiple multi-objective models. Preferably, the actual response times of the sample data in the training set can be uniformly distributed.

[0075] In one possible implementation, the feature information of the sample service request may include the feature information of the service requested by the sample service request and the feature information of the corresponding service requester. For example, taking the service requested by the sample service request as a ride-hailing service as an example, the feature information of the requested service includes at least one of the following: feature information of the pick-up and drop-off points, such as whether the pick-up and drop-off points are located in popular ride-hailing areas, the historical ride-hailing response rate of the pick-up and drop-off points, the historical ride-hailing cancellation rate, etc.; ride-hailing price, distance, duration, service brand of the requested service, etc.; and feature information of the service requester, including passenger identity information, passenger's historical ride-hailing response rate, cancellation rate, etc.

[0076] In one possible implementation, the multi-gate Mixture-of-Experts (MMoE) model has two learning objectives: one is classification learning, which aims to predict the probability that the response time is within a predetermined range; the other is regression learning, which aims to predict the response time of a service request when the response time is within the predetermined range.

[0077] In one possible implementation, different multi-objective models correspond to different predetermined ranges, and different predetermined ranges can be configured for multiple multi-objective models according to the actual response time of sample service requests in the sample set.

[0078] In this embodiment, the response duration of the long-tailed distribution can be divided into multiple different predetermined ranges. Then, each multi-objective model learns the correspondence between feature information and response duration in different predetermined ranges. In this way, the prediction effect of different multi-objective models on response durations distributed in different predetermined ranges will also be different. The probability of the service request response duration output by the multi-objective model being in the predetermined range can be used as the basis to determine the response duration distribution of the service request. When the response duration is in the predetermined range with the highest probability, the predicted response duration of the corresponding service request is the final prediction result, and the prediction result is more accurate.

[0079] Step S102 in the above-mentioned method for training the prediction model of response duration, namely, training multiple initial multi-objective models based on the sample data in the training set to obtain multiple multi-objective models, may include the following steps:

[0080] The classification branch of the initial multi-objective model is trained based on all sample data in the training set, and the regression branch of the initial multi-objective model is trained based on sample data in the training set whose actual response time is within a predetermined range, to obtain the multi-objective model.

[0081] The output of the classification branch in the multi-objective model is the probability that the service request response time is within a predetermined range, and the output of the regression branch in the multi-objective model is the predicted response time of the service request whose response time is within a predetermined range.

[0082] In this implementation, the initial multi-objective model includes a main branch, a classification branch, and a regression branch. The main branch is used to perform feature extraction and other processing on the feature information of the sample service request to obtain the feature vector of the sample service request. The input of the classification branch is the feature vector of the sample service request, and the output is the probability that the service request response time is within a predetermined range. The input of the regression branch is the feature vector of the sample service request, and the output is the predicted response time of the service request with a response time within the predetermined range. Since the learning objective of the regression branch is the predicted response time of the service request with a response time within the predetermined range, in order to learn a better model, when training the regression branch, only the feature vectors corresponding to the sample data with actual response times within the predetermined range are used for training. The learning objective of the classification branch is the probability that the service request response time is within the predetermined range, and it needs to use the feature vectors corresponding to the sample data with actual response times outside the predetermined range as negative samples for training.

[0083] In this embodiment, when training the multi-objective model, the feature information of the sample service request in the sample data of the training set can be input into the multi-objective model to obtain two output results of the multi-objective model. The accuracy of the two output results is determined respectively. If neither accuracy reaches the preset threshold, the parameters in the multi-objective model are continuously adjusted until the two output results of the multi-objective model reach the preset threshold respectively. In this way, the multi-objective model is trained.

[0084] In one possible implementation, the predetermined range corresponding to the multi-objective model includes less than or equal to K seconds, and the K value corresponding to each multi-objective model is different, wherein K is greater than 0.

[0085] For example, assume that the actual response time of sample service requests in the sample set is greater than 0 and less than T. N There are N multi-objective models M1, M2...M N ,like Figure 2 As shown, the predetermined range configured for the multi-objective model M1 can be less than or equal to 30 seconds. The learning objective of the multi-objective model M1 is to predict the probability that the response time is less than or equal to 30 seconds and to predict the response time of the service request when the response time falls within the distribution range of less than or equal to 30 seconds. The predetermined range configured for the multi-objective model M2 can be less than or equal to 50 seconds. The learning objective of the multi-objective model M2 is to predict the probability that the response time is less than or equal to 50 seconds and to predict the response time of the service request when the response time falls within the distribution range of less than or equal to 50 seconds. ... The predetermined range configured for the multi-objective model M2 can be less than or equal to 50 seconds. N The predefined range of the configuration can be less than or equal to T. N Seconds, the multi-objective model M N The learning objective is to predict response times less than or equal to T. NThe probability of a second and the predicted response time being less than or equal to T N The response time for a service request is within the range of seconds.

[0086] In one possible implementation, after training the multiple multi-objective models, the method further includes the following steps:

[0087] Obtain another set of sample data from the sample set as the test set;

[0088] The multiple multi-objective models are tested using sample data from the test set to determine their prediction errors.

[0089] In this implementation, another set of sample data can be selected from the sample set as a test set. The sample data in the test set does not overlap with the sample data in the training set, and the number of sample data in the training set is much larger than the number of sample data in the test set. For example, the ratio of the number of sample data in the training set to the number of sample data in the test set can be 7:3 or 6:4.

[0090] In one possible implementation, the step of testing the plurality of multi-objective models using sample data from the test set and determining the prediction error of the plurality of multi-objective models may include the following steps:

[0091] The feature information of the sample service requests in the test set is input into the multiple multi-objective models respectively to obtain the output results of each multi-objective model;

[0092] The prediction response time corresponding to the highest probability value among the output results of the multiple multi-objective models is taken as the final prediction result;

[0093] Based on the final prediction result of each sample service request in the test set and the actual response time of the sample service request, the prediction error of the multiple multi-objective models is determined.

[0094] In this embodiment, such as Figure 2 As shown, for each sample service request in this test set, the feature information of the sample service request can be input into the multiple multi-objective models M1, M2...M... N The multi-objective model M1, M2...M is obtained. N The two outputs are the probability and the predicted response time, derived from M1, M2...M N Find the multi-objective model with the highest probability value in the output results, and determine the predicted response time output by the multi-objective model with the highest probability value as the final prediction result of the service request for this sample.

[0095] In this implementation, the final prediction result can be compared with the actual response time of the sample service request, and the difference between the two is the prediction error. In this way, the prediction error corresponding to each sample service request in the test set can be calculated. Based on the prediction error corresponding to each sample service request in the test set, the prediction error of the multi-objective model can be calculated. For example, the prediction error corresponding to each sample service request in the test set can be weighted and averaged to obtain the prediction error of the multiple multi-objective models; or the prediction error corresponding to each sample service request in the test set can be calculated by mean square error to obtain the prediction error of the multi-objective model; and so on.

[0096] In this embodiment, the calculated prediction error of the multiple multi-objective models can indicate the prediction accuracy of the response time prediction model. Tests on the test set show that, compared with existing models, the prediction model composed of multiple multi-objective models improves the R-squared index by 10% and has higher accuracy.

[0097] This disclosure also provides a method for predicting response time. Figure 3 A flowchart illustrating a method for predicting response time according to an embodiment of the present disclosure is shown, as follows: Figure 3 As shown, the method for predicting response duration includes the following steps S301-S303:

[0098] In step S301, the characteristic information of the current service request is obtained;

[0099] In step S302, the feature information of the current service request is input into a plurality of preset multi-objective models respectively to obtain the output results of each multi-objective model. The output results of the multi-objective model include the probability that the response time of the current service request is within the corresponding predetermined range, and the predicted response time of the current service request when the response time is within the predetermined range. The predetermined range is different for each multi-objective model.

[0100] In step S303, the predicted response time of the multi-objective model with the highest probability among the output results of the multiple multi-objective models is taken as the predicted response time of the current service request.

[0101] In one possible implementation, the entity executing the response time prediction method can be a computer, server, or server cluster, etc. For example, it could be the backend server of a service platform.

[0102] In one possible implementation, the current service request refers to a service request initiated by the current service requester at the current moment. The characteristic information of the current service request may include the characteristic information of the service requested by the current service request and the characteristic information of the corresponding service requester. For example, taking the service requested by the current service request as a ride-hailing service as an example, the characteristic information of the requested service includes at least one of the following: the characteristic information of the pick-up and drop-off points, such as whether the pick-up and drop-off points are located in popular ride-hailing areas, the historical ride-hailing response rate of the pick-up and drop-off points, the historical ride-hailing cancellation rate, etc.; the ride-hailing price, distance, duration, the service brand of the requested service, etc.; the characteristic information of the service requester includes the passenger's identity information, the passenger's historical ride-hailing response rate, cancellation rate, etc.

[0103] In one possible implementation, the input of the multi-objective model is the feature information of the current service request, and the output of the multi-objective model has two parts: one is the probability that the service request response time is within the corresponding predetermined range, and the other is the predicted response time of the service request when the response time is within the predetermined range. Assume there are multiple multi-objective models M1, M2, and M3. M1 corresponds to a predetermined range of less than or equal to 30 seconds, M2 corresponds to a predetermined range of less than or equal to 1 minute, and M3 corresponds to a predetermined range of less than or equal to 3 minutes. The output of M1 shows a 90% probability that the service request response time is less than or equal to 30 seconds, and a predicted response time of 15 seconds when the response time is less than or equal to 30 seconds. The output of M2 shows a 40% probability that the service request response time is less than or equal to 1 minute, and a predicted response time of 40 seconds when the response time is less than or equal to 1 second. The output of M3 shows a 38% probability that the service request response time is less than or equal to 3 minutes, and a predicted response time of 1 minute when the response time is less than or equal to 3 minutes. Therefore, the predicted response time of 15 seconds for the current service request can be taken from the multi-objective model M1, which has the highest probability (90%).

[0104] In one possible implementation, the multiple multi-objective models described herein can be trained using the aforementioned response duration prediction model training method, the specific training process of which will not be elaborated here. These multiple multi-objective models can be trained by the execution entity of this prediction method using the aforementioned response duration prediction model training method, or they can be trained by other devices using the aforementioned response duration prediction model training method and then sent to the execution entity of this prediction method for use; no limitation is imposed here.

[0105] This disclosure also provides a service request response processing method. Figure 4 A flowchart illustrating a service request response processing method according to an embodiment of the present disclosure is shown, such as... Figure 4As shown, the service request response processing method includes the following steps S401-S402:

[0106] In step S401, the predicted response time for the current service request is obtained;

[0107] In step S402, based on the predicted response time of the current service request, a service provider is assigned to respond to the current service request at the corresponding time.

[0108] In one possible implementation, the above-described method for predicting response duration can be used to obtain the predicted response duration for the current service request.

[0109] In one possible implementation, after obtaining the predicted response time for the current service request, a service provider can be assigned to respond to the current service request at a corresponding time based on the predicted response time. For example, if the predicted response time is long, to prevent the service requester from waiting too long, a service provider can be assigned to respond to the current service request at a time when the interval between the initiation time of the current service request and the predicted response time is shorter than the predicted response time. Alternatively, to prevent the service requester from canceling the service request, a service provider can be assigned to respond to the current service request at a time when the interval between the initiation time of the current service request and the predicted response time exceeds a predetermined duration of the predicted response time. In this way, service providers can be precisely scheduled to provide services to the service requester based on the predicted response time of the current service request.

[0110] For example, assuming the service is a ride-hailing service, after a passenger places an order, the predicted response time for their ride request can be estimated. If the predicted response time is as long as 5 minutes, to prevent the service requester from waiting too long, a driver can be directly assigned to the passenger at the corresponding time when the interval between the passenger's order placement and the actual time is 3 minutes, and no driver has accepted the order yet. Alternatively, if the predicted response time for the passenger's ride request is 2 minutes, to prevent the passenger from canceling the ride order, a driver can be directly assigned to the passenger when the interval between the passenger's order placement and the actual time is more than 2 minutes, reaching 2 minutes and 30 seconds, and no driver has accepted the order yet.

[0111] In one possible implementation, the above service request response processing method may further include the following steps:

[0112] If the predicted response time of the current service request exceeds the preset time, then a corresponding service value-added service is configured for the current service request based on the predicted response time of the current service request. The service value corresponding to the current service request includes the original service value and the service value-added service.

[0113] In this implementation, if the predicted response time for a current service request exceeds a preset time, it indicates that the requested service is difficult to provide, and service providers are unwilling to respond. In this case, to increase the service providers' willingness to respond and improve the response rate, a corresponding service value-added service can be configured for the current service request based on the original service value. This can be done by following the rule that the longer the predicted response time, the higher the corresponding service value-added service. It is understood that the service value-added service can be a fee or other form of service, such as the quality or quantity of tasks the service provider can obtain. The specifics can be determined based on the actual situation and are not limited here.

[0114] The following are embodiments of the apparatus disclosed herein, which can be used to execute embodiments of the method disclosed herein.

[0115] Figure 5 This diagram illustrates a structural block diagram of a predictive model training apparatus for response time according to an embodiment of the present disclosure. This apparatus can be implemented as part or all of an electronic device through software, hardware, or a combination of both. Figure 5 As shown, the predictive model training device for response duration includes:

[0116] The first acquisition module 501 is configured to acquire a set of sample data from the sample set as a training set, wherein the sample data includes feature information of sample service requests and the actual response time of sample service requests.

[0117] Training module 502 is configured to train multiple initial multi-objective models based on sample data in the training set to obtain multiple multi-objective models. The learning objectives of the multi-objective models include the probability that the service request response time is within a predetermined range, and the predicted response time of service requests with response times within the predetermined range. The predetermined range is different for each multi-objective model.

[0118] In one possible implementation, the training module 502 is configured as follows:

[0119] The classification branch of the initial multi-objective model is trained based on all sample data in the training set, and the regression branch of the initial multi-objective model is trained based on sample data in the training set whose actual response time is within a predetermined range, to obtain the multi-objective model.

[0120] The output of the classification branch in the multi-objective model is the probability that the service request response time is within a predetermined range, and the output of the regression branch in the multi-objective model is the predicted response time of the service request whose response time is within a predetermined range.

[0121] In one possible implementation, the feature information of the sample service request includes the feature information of the service requested by the sample service request and the feature information of the corresponding service requester.

[0122] In one possible implementation, the predetermined range corresponding to the multi-objective model includes less than or equal to K seconds, and the K value corresponding to each multi-objective model is different, wherein K is greater than 0.

[0123] Figure 5 The technical terms and technical features involved in the illustrated and related embodiments are consistent with Figures 1 to 4 The technical terms and technical features mentioned in the illustrated and related embodiments are the same or similar, for Figure 5 The explanations and descriptions of the technical terms and features involved in the illustrated and related embodiments can be found in the above-mentioned... Figures 1 to 4 The explanations of the illustrated and related embodiments are not repeated here.

[0124] Figure 6 This diagram illustrates a structural block diagram of a response time prediction device according to an embodiment of the present disclosure. This device can be implemented as part or all of an electronic device through software, hardware, or a combination of both. Figure 6 As shown, the response duration prediction device includes:

[0125] The second acquisition module 601 is configured to acquire the characteristic information of the current service request;

[0126] The prediction module 602 is configured to input the feature information of the current service request into a plurality of preset multi-objective models respectively, and obtain the output result of each multi-objective model. The output result of the multi-objective model includes the probability that the response time of the current service request is within a corresponding predetermined range, and the predicted response time of the current service request when the response time is within the predetermined range. The predetermined range is different for each multi-objective model.

[0127] The determination module 603 is configured to determine the predicted response time of the multi-objective model with the highest probability among the output results of the multiple multi-objective models as the predicted response time of the current service request.

[0128] In one possible implementation, the second acquisition module 601 is configured to:

[0129] The multiple multi-objective models are obtained based on the above-described method for training prediction models of response duration.

[0130] Figure 6 The technical terms and technical features involved in the illustrated and related embodiments are consistent with Figures 1 to 4The technical terms and technical features mentioned in the illustrated and related embodiments are the same or similar, for Figure 6 The explanations and descriptions of the technical terms and features involved in the illustrated and related embodiments can be found in the above-mentioned... Figures 1 to 4 The explanations of the illustrated and related embodiments are not repeated here.

[0131] Figure 7 This diagram illustrates a structural block diagram of a service request response processing apparatus according to an embodiment of the present disclosure. This apparatus can be implemented as part or all of an electronic device through software, hardware, or a combination of both. Figure 7 As shown, the service request response processing device includes:

[0132] The third acquisition module 701 is configured to acquire the predicted response time of the current service request;

[0133] The allocation module 702 is configured to allocate a service provider to respond to the current service request at a corresponding time based on the predicted response time of the current service request.

[0134] In one possible implementation, the device further includes:

[0135] The configuration module is configured to configure a corresponding service value-added service for the current service request based on the predicted response time of the current service request if the predicted response time of the current service request exceeds a preset time. The service value corresponding to the current service request includes the original service value and the service value-added service.

[0136] Figure 7 The technical terms and technical features involved in the illustrated and related embodiments are consistent with Figures 1 to 4 The technical terms and technical features mentioned in the illustrated and related embodiments are the same or similar, for Figure 7 The explanations and descriptions of the technical terms and features involved in the illustrated and related embodiments can be found in the above-mentioned... Figures 1 to 4 The explanations of the illustrated and related embodiments are not repeated here.

[0137] This disclosure also discloses an electronic device. Figure 8 A structural block diagram of an electronic device according to an embodiment of the present disclosure is shown, such as... Figure 8 As shown, the electronic device 800 includes a memory 801 and a processor 802; wherein,

[0138] The memory 801 is used to store one or more computer instructions, wherein the one or more computer instructions are executed by the processor 802 to implement the above method steps.

[0139] Figure 9This is a schematic diagram of the structure of a computer system suitable for implementing the method according to embodiments of the present disclosure.

[0140] like Figure 9 As shown, the computer system 900 includes a processing unit 901, which can execute various processes described above based on a program stored in a read-only memory (ROM) 902 or a program loaded from a storage section 908 into a random access memory (RAM) 903. The RAM 903 also stores various programs and data required for the operation of the system 900. The processing unit 901, ROM 902, and RAM 903 are interconnected via a bus 904. An input / output (I / O) interface 905 is also connected to the bus 904.

[0141] The following components are connected to I / O interface 905: an input section 906 including a keyboard, mouse, etc.; an output section 907 including a cathode ray tube (CRT), liquid crystal display (LCD), etc., and a speaker, etc.; a storage section 908 including a hard disk, etc.; and a communication section 909 including a network interface card such as a LAN card, modem, etc. The communication section 909 performs communication processing via a network such as the Internet. A drive 910 is also connected to I / O interface 905 as needed. A removable medium 911, such as a disk, optical disk, magneto-optical disk, semiconductor memory, etc., is installed on drive 910 as needed so that computer programs read from it can be installed into storage section 908 as needed. The processing unit 901 can be implemented as a CPU, GPU, TPU, FPGA, NPU, etc.

[0142] In particular, according to embodiments of this disclosure, the methods described above can be implemented as computer software programs. For example, embodiments of this disclosure include a computer program product comprising a computer program tangibly embodied on a readable medium thereof, the computer program containing program code for performing the methods described above. In such embodiments, the computer program can be downloaded and installed from a network via communication section 909, and / or installed from removable medium 911.

[0143] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of this disclosure. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code containing one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions indicated in the blocks may occur in a different order than those indicated in the drawings. For example, two consecutively indicated blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in the block diagrams and / or flowcharts, and combinations of blocks in the block diagrams and / or flowcharts, can be implemented using a dedicated hardware-based system that performs the specified function or operation, or using a combination of dedicated hardware and computer instructions.

[0144] The units or modules described in the embodiments of this disclosure can be implemented in software or hardware. The described units or modules can also be located in a processor, and the names of these units or modules do not necessarily constitute a limitation on the unit or module itself.

[0145] In another aspect, embodiments of this disclosure also provide a computer-readable storage medium, which may be a computer-readable storage medium included in the apparatus described in the above embodiments; or it may be a standalone computer-readable storage medium not assembled into a device. The computer-readable storage medium stores one or more programs, which are used by one or more processors to perform the methods described in embodiments of this disclosure.

[0146] The above description is merely a preferred embodiment of this disclosure and an explanation of the technical principles employed. Those skilled in the art should understand that the scope of the invention involved in the embodiments of this disclosure is not limited to technical solutions formed by specific combinations of the above-described technical features, but should also cover other technical solutions formed by arbitrary combinations of the above-described technical features or their equivalents without departing from the inventive concept. For example, technical solutions formed by substituting the above features with (but not limited to) technical features with similar functions disclosed in the embodiments of this disclosure.

Claims

1. A method for training a prediction model for response duration, wherein, include: A set of sample data is obtained from the sample set as a training set, and the sample data includes the feature information of the sample service request and the actual response time of the sample service request. Multiple initial multi-objective models are trained based on all sample data in the training set, including classification branches. Multiple initial multi-objective models are trained based on sample data in the training set whose actual response times are within a predetermined range, resulting in multiple multi-objective models. The output of the classification branch in each multi-objective model is the probability that the service request response time is within the predetermined range. The output of the regression branch in each multi-objective model is the predicted response time of the service request whose response time is within the predetermined range. The predetermined range is different for each multi-objective model.

2. The method according to claim 1, wherein, The feature information of the sample service request includes the feature information of the service requested by the sample service request and the feature information of the corresponding service requester.

3. The method according to claim 1, wherein, The predetermined range corresponding to the multi-objective model includes less than or equal to K seconds, and the K value is different for each multi-objective model, wherein K is greater than 0.

4. A method for predicting response duration, wherein, include: Obtain the characteristic information of the current service request; The feature information of the current service request is input into multiple preset multi-objective models to obtain the output results of each multi-objective model. The output results of the multi-objective model include the probability that the response time of the current service request is within the corresponding predetermined range, and the predicted response time of the current service request when the response time is within the predetermined range. The predetermined range is different for each multi-objective model. The predicted response time of the multi-objective model with the highest probability among the output results of the multiple multi-objective models is determined as the predicted response time of the current service request; The plurality of multi-objective models are obtained based on the prediction model training method for response duration according to any one of claims 1 to 3.

5. A service request and response processing method, wherein, include: Based on the method described in claim 4, the predicted response time for the current service request is obtained; Based on the predicted response time of the current service request, a service provider is assigned to respond to the current service request at the appropriate time.

6. The method according to claim 5, wherein, Also includes: If the predicted response time of the current service request exceeds the preset time, then a corresponding service value-added service is configured for the current service request based on the predicted response time of the current service request. The service value corresponding to the current service request includes the original service value and the service value-added service.

7. An electronic device comprising a memory and a processor; wherein, The memory is used to store one or more computer instructions, wherein the one or more computer instructions are executed by the processor to implement the steps of the method according to any one of claims 1 to 6.

8. A computer-readable storage medium having stored thereon computer instructions that, when executed by a processor, implement the steps of the method according to any one of claims 1-6.

9. A computer program product comprising a computer program / instructions, wherein, When the computer program / instructions are executed by the processor, they implement the steps of the method described in any one of claims 1-6.

Images