Method, device and electronic device for determining order recommendation list

Abstract

This specification provides a method, device, and electronic device for determining an order recommendation list, wherein the method includes: after receiving an order acceptance request from the driver, obtaining information on multiple passenger orders in a waiting state in real time; inputting each passenger order information and the order acceptance request information into a predetermined network model to obtain the order completion rate corresponding to each passenger order; wherein the predetermined network model includes a first subnetwork for predicting the order acceptance rate and a second subnetwork for predicting the order completion rate, and the passenger order information and the driver's order acceptance request information are used as inputs of the two subnetworks; the predetermined operation results of the output values ​​of the first subnetwork and the second subnetwork are used as the order completion rate; and according to the order completion rate corresponding to the passenger order, each passenger order is sorted in descending order to obtain an order recommendation list. This solution can take into account both the order acceptance rate and the order completion rate, and will not cause the problem of one indicator being higher than the other indicator being lower.

Description

Technical Field

[0001] The present application relates to the field of Internet technology, and in particular to a method, device, and electronic device for determining an order recommendation list. Background Art

[0002] Currently, when online ride-hailing services recommend orders to driver-side users, they usually first predict the order completion rate corresponding to each passenger-side order, sort each passenger-side order according to the order acceptance rate, or the order completion rate from high to low, and select the first few passenger orders as the order recommendation list and send it to the driver-side user.

[0003] However, if the driver accepts orders according to the recommended list and accepts the first few orders in the recommended list each time, the driver's order acceptance rate will be high but the order completion rate will be low; or the order completion rate will be high but the order acceptance rate will be low, making it difficult to balance the order acceptance rate and the order completion rate. Summary of the Invention

[0004] The purpose of this application is to provide a method, device and electronic device for determining an order recommendation list to solve the problem that existing methods are difficult to balance the order acceptance rate and the order completion rate.

[0005] In order to solve the above technical problems, the first aspect of this specification provides a method for determining an order recommendation list, including: after receiving the order acceptance request information from the driver side, obtaining information on multiple passenger orders in a waiting state in real time; inputting each passenger order information and the order acceptance request information into a predetermined network model respectively to obtain the order completion rate corresponding to each passenger order; wherein, the predetermined network model includes a first subnetwork and a second subnetwork, and the passenger order information and the driver side's order acceptance request information are both inputs of the first subnetwork and inputs of the second subnetwork; the first subnetwork is used to predict the order acceptance rate, and the second subnetwork is used to predict the order completion rate, and the predetermined operation result between the output values ​​of the first subnetwork and the second subnetwork is used as the order completion rate; according to the order completion rate corresponding to the passenger order, each passenger order is sorted in descending order to obtain an order recommendation list, and the order recommendation list is used to be sent to the driver side so that the driver side can select an order from the recommendation list to accept the order.

[0006] In some embodiments, the first sub-network and the second sub-network have the same structure, and the parameters are located at the same position in the first sub-network and the second sub-network.

[0007] In some embodiments, the output of the reservation network model also includes the order completion rate; according to the order completion rate corresponding to the passenger order, each passenger order is sorted in descending order to obtain an order recommendation list, and also includes: according to the order completion rate and order completion rate corresponding to the passenger order, each passenger order is sorted in descending order to obtain an order recommendation list.

[0008] In some embodiments, each passenger order is sorted in descending order according to the order completion rate and order completion rate corresponding to the passenger order to obtain an order recommendation list, including: taking the weighted sum of the order completion rate and order completion rate corresponding to each passenger order; and sorting each passenger order in descending order according to the sum result to obtain a recommendation list.

[0009] In some embodiments, the order completion rate corresponds to a first weight, and the order completion rate corresponds to a second weight; accordingly, before weighted summing of the order completion rate and the order completion rate corresponding to each passenger order, it also includes: receiving the values ​​of the first weight and the second weight set by the user.

[0010] In some embodiments, receiving the values ​​of the first weight and the second weight set by the user includes: receiving the order acceptance tendency selected by the driver, the order acceptance tendency including a tendency to accept more orders and a tendency to have a high order completion rate; setting the first weight and the second weight to the first preset weight and the second preset weight corresponding to the order acceptance tendency selected by the driver.

[0011] In some embodiments, receiving the values ​​of the first weight and the second weight set by the user includes: receiving the order acceptance tendency and the tendency value selected by the driver, the order acceptance tendency including the tendency to accept more orders and the tendency to have a high order completion rate, and each order acceptance tendency including multiple tendency values; setting the first weight and the second weight to the first preset weight and the second preset weight corresponding to the tendency value under the order acceptance tendency selected by the driver.

[0012] The second aspect of this specification provides an order recommendation list determination device, including: an acquisition unit, which is used to obtain information on multiple passenger orders in a waiting state in real time after receiving order acceptance request information from the driver side; a processing unit, which is used to input each passenger order information and the order acceptance request information into a predetermined network model respectively to obtain the order completion rate corresponding to each passenger order; wherein the predetermined network model includes a first subnetwork and a second subnetwork, and the passenger order information and the driver side's order acceptance request information are both inputs of the first subnetwork and the second subnetwork; the first subnetwork is used to predict the order acceptance rate, and the second subnetwork is used to predict the order completion rate, and the predetermined operation result between the output values ​​of the first subnetwork and the second subnetwork is used as the order completion rate; a sorting unit, which is used to sort each passenger order in descending order according to the order completion rate corresponding to the passenger order to obtain an order recommendation list, and the order recommendation list is used to be sent to the driver side so that the driver side can select an order from the recommendation list to accept it.

[0013] In some embodiments, the first sub-network and the second sub-network have the same structure, and the parameters are located at the same position in the first sub-network and the second sub-network.

[0014] In some embodiments, the output of the reservation network model also includes the order completion rate; the sorting unit includes: a first sorting sub-unit, which is used to sort each passenger's order in descending order according to the order completion rate and order completion rate corresponding to the passenger's order to obtain an order recommendation list.

[0015] In some embodiments, the first sorting subunit includes: a summing subunit, which is used to weightedly sum the order completion rate and order completion rate corresponding to each passenger order; and a second sorting subunit, which is used to sort each passenger order in descending order according to the summation result to obtain a recommendation list.

[0016] In some embodiments, the order completion rate corresponds to the first weight, and the order completion rate corresponds to the second weight; the first sorting subunit also includes: a first receiving subunit, which is used to receive the values ​​of the first weight and the second weight set by the user.

[0017] In some embodiments, the first receiving subunit includes: a second receiving subunit, used to receive the order acceptance tendency selected by the driver, wherein the order acceptance tendency includes a tendency to accept more orders and a tendency to have a high order completion rate; a first setting subunit, used to set the first weight and the second weight to the first preset weight and the second preset weight corresponding to the order acceptance tendency selected by the driver.

[0018] In some embodiments, the first receiving subunit includes: a third receiving subunit, used to receive the order acceptance tendency and tendency value selected by the driver, the order acceptance tendency includes a tendency to accept more orders and a tendency to have a high order completion rate, and each order acceptance tendency includes multiple tendency values; a second setting subunit, used to set the first weight and the second weight to the first preset weight and the second preset weight corresponding to the tendency value under the order acceptance tendency selected by the driver.

[0019] The third aspect of this specification provides an electronic device, comprising: a memory and a processor, wherein the processor and the memory are communicatively connected to each other, the memory stores computer instructions, and the processor implements the steps of any one of the methods described in the first aspect by executing the computer instructions.

[0020] A fourth aspect of this specification provides a computer storage medium, wherein the computer storage medium stores computer program instructions, and when the computer program instructions are executed by a processor, the steps of any one of the methods described in the first aspect are implemented.

[0021] A fifth aspect of this specification provides a computer program product, which includes a computer program. When the computer program is executed by a processor, the steps of any one of the methods described in the first aspect are implemented.

[0022] The method and device for determining a recommended order list provided in this specification inputs driver-side order acceptance request information and passenger order information into a predetermined network model comprising a first subnetwork and a second subnetwork. The first subnetwork predicts the order acceptance rate, while the second subnetwork predicts the order completion rate. The predetermined calculation result between the output values ​​of the first and second subnetworks is used as the order completion rate output by the predetermined network model. Passenger orders are then sorted according to the order completion rate to generate a recommended order list. Because the order completion rate in this scheme is obtained by multiplying the order acceptance rate and order completion rate obtained from different subnetworks, it is the result of the combined effect of the order acceptance rate and the order completion rate. Therefore, selecting passenger orders based on the recommended list sorted by order completion rate balances both the order acceptance rate and the order completion rate, avoiding the issue of one indicator being too high while the other is too low. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] In order to more clearly illustrate the implementation methods of the present application or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the implementation methods or the description of the prior art. Obviously, the drawings described below are only some implementation methods recorded in this application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative labor.

[0024] Figure 1A schematic diagram of an application scenario of an embodiment of this specification is shown;

[0025] Figure 2 A flowchart of the method for determining the order recommendation list provided in this specification is shown;

[0026] Figure 3 shows a schematic structural diagram of a predetermined network model;

[0027] Figure 4 A schematic diagram showing a control for adjusting the first weight and the second weight;

[0028] Figure 5 It shows a principle block diagram of the device for determining the order recommendation list provided in this specification;

[0029] Figure 6 A functional block diagram of an electronic device provided in this specification is shown. DETAILED DESCRIPTION

[0030] In order to enable those skilled in the art to better understand the technical solutions in this application, the following will clearly and completely describe the technical solutions in the embodiments of this application in conjunction with the drawings in the embodiments of this application. Obviously, the described embodiments are only part of the embodiments of this application, not all of the embodiments. Based on the embodiments of this application, all other embodiments obtained by ordinary technicians in this field without making creative efforts should fall within the scope of protection of this application.

[0031] This specification provides a method for determining an order recommendation list, which can be used to determine an order recommendation list provided to a driver-side user.

[0032] Figure 1 A schematic diagram of the application scenario of an embodiment of this specification is shown, wherein A is an online car-hailing platform, B is a driver-side user, and C is a passenger-side user. When a passenger needs to take a taxi to his destination, he can place an order through a passenger terminal interface such as a mobile phone APP or a web page, or he can place an order by calling the customer service hotline. When placing an order, he provides the travel time, departure place, and destination, and waits for the driver-side user to accept the order after placing the order. The driver-side user can receive an order recommendation list sent by the online car-hailing platform through a mobile terminal such as a mobile phone APP, and select an order from the order recommendation list to accept the order. After accepting the order, the driver-side user should drive the vehicle to the departure place specified in the target passenger order, pick up the passenger, and deliver the passenger to the destination. After delivering the passenger to the destination, the target passenger order is considered to be completed. The order recommendation list determination method provided in this specification can be used for the above-mentioned online car-hailing platform to determine an order recommendation list for the driver-side user so that the driver-side user can receive more matching passenger orders.

[0033] It should be noted that the term "online ride-hailing platform" in this specification refers to a service platform built on internet technology that connects qualified vehicles and drivers to provide passengers with taxi reservation services. Passengers can hail a taxi individually, which is referred to as an "express ride," or they can share a ride with others, which is also referred to as a "ride-sharing ride." The method for determining a recommended order list provided in this specification can be used for both the express ride and ride-sharing services.

[0034] The term "ride-sharing" in this specification is also called carpooling or hitchhiking, which means that the driver wants to go from X to Y. If the vehicle is not overloaded, he can pick up and drop off passengers in the direction from X to Y to complete the order. For passengers, the price of a ride-sharing is usually cheaper than taking a taxi alone, but you may need to share the car with strangers. For online car-hailing platforms, the difference between a ride-sharing and a passenger taking a taxi alone is that, in addition to obtaining the time, departure place, and destination of the passenger's travel, it is also necessary to obtain the time, departure place, and destination of the driver's travel, and it is necessary to match the driver's and the passenger's travel routes (i.e., the driving route from the departure place to the destination) and travel time.

[0035] The "order acceptance rate" mentioned in this manual refers to: after a driver's order is matched with a passenger's order, whether the passenger's order is accepted by the driver (i.e., the order is accepted); the "order completion rate" refers to: after the driver accepts the passenger's order, whether the driver will eventually take the passenger to the destination and complete the order; the "order completion rate" refers to: after a driver's order is matched with a passenger's order, whether the driver will eventually take the passenger to the destination and complete the order.

[0036] The method for determining the order recommendation list provided in this manual is as follows: Figure 2 As shown, the following steps are included:

[0037] S110: After receiving the order request information from the driver side, obtain the information of multiple passenger orders in the waiting order status in real time.

[0038] Generally, after receiving the order request from the driver, the online car-hailing platform will promptly determine the order recommendation list for the driver. Figure 2 The method shown.

[0039] In some embodiments, such as in the express service, the order request information sent by the driver may carry the driver's current location, so that the online car-hailing platform can filter out passenger orders that can be assigned to the driver based on the driver's current location.

[0040] In some embodiments, such as in the express service, the order acceptance request information sent by the driver may also carry the driver's current order acceptance status, that is, whether it is "order accepted and being completed" or "pending order", so that the online car-hailing platform can give priority to assigning passenger orders to drivers in the pending order status based on the actual order acceptance status.

[0041] In some embodiments, such as in express and ride-sharing services, the order request information sent by the driver may also carry a travel time, which may be a specific time point or a time period (for example, tomorrow afternoon from 2 to 3).

[0042] In some embodiments, such as in a ride-sharing service, the order request information sent by the driver may also carry the driver's departure and destination, so that the online ride-hailing platform can match the driver's travel route with the passenger's travel route.

[0043] Passenger order information is sent by users through the passenger terminal, including the passenger's travel time, departure place, and destination.

[0044] Passenger order information can be classified into two statuses: Accepted and Pending. For passenger orders in the Accepted status, the driver who accepted the order will complete the order. For passenger orders in the Pending status, the driver must first accept the order before the order can be completed. In some embodiments, it is common for a passenger order to be accepted by a driver and then canceled due to driver-related reasons. In this case, the cancelled passenger order can also be considered to be in the Pending status.

[0045] In step S110, "obtaining information on multiple passenger orders in a pending order state" can be performed by filtering out passenger orders whose departure locations are in the same area as the driver's current location from the multiple passenger orders in a pending order state, or filtering out passenger orders whose departure locations are near the driver's travel route and whose travel times are within the same time period as the driver's travel time from the multiple passenger orders in a pending order state. In other words, "obtaining information on multiple passenger orders in a pending order state" in step S110 can be specifically performed by "obtaining information on multiple passenger orders in a pending order state that match the order acceptance request information."

[0046] In some embodiments, after receiving a driver-side order acceptance request and before the driver-side confirms the order, information about passenger orders in the pending order status can be obtained in real time. For example, if a driver-side order acceptance request is received at time t and the driver-side confirms the order acceptance at time t+100, then passenger orders in the pending order status can be continuously obtained between time t and time t+100. During this period, passengers may submit new orders as pending order requests, and some passenger orders in the pending order status may be accepted by other drivers, changing their status to accepted order status.

[0047] In order to prevent the driver from finding that the order he accepted has been accepted by other drivers after accepting the order, and to enable the driver to accept more orders, step S110 performs the following operations in real time after receiving the order acceptance request from the driver and before the driver confirms to accept the order: obtain information on passenger orders in the waiting state, and generate an order recommendation list in real time based on the obtained content, and send the order recommendation list to the driver.

[0048] In some embodiments, the order acceptance request on the driver's side may also carry the driver's portrait information, such as the driver's gender, age, and the number of orders received and completed within a predetermined time period; the passenger order information may carry the passenger's portrait information, such as the passenger's gender, age, and the number of orders issued within a predetermined time period.

[0049] S120: Each passenger's order information and order acceptance request information are input into the reservation network model to obtain the order completion rate corresponding to each passenger's order. The reservation network model includes a first subnetwork and a second subnetwork. The passenger's order information and the driver's order acceptance request information serve as both inputs to the first and second subnetworks. The first subnetwork is used to predict the order acceptance rate, while the second subnetwork is used to predict the order completion rate. The result of the predetermined operation between the output values ​​of the first and second subnetworks is used as the order completion rate.

[0050] In some embodiments, the predetermined operation result between the output values ​​of the first subnetwork and the second subnetwork can be the product of the output values ​​of the first subnetwork and the second subnetwork, that is, the product of the order acceptance rate and the order completion rate, that is, taking order acceptance and order completion as the driver's action sequence, and calculating the probability of the sequence occurring.

[0051] In some embodiments, the predetermined operation result between the output values ​​of the first sub-network and the second sub-network may also be the result of performing other operations, such as a ratio operation, a square difference operation, etc.

[0052] In some embodiments, passenger order information and order request information can be organized in the form of a matrix, and continuous features can be organized into one matrix, while categorical features can be organized into another matrix. Continuous features refer to features whose values ​​are continuous data, for example, the distance between the departure point of a passenger order and the driver. The number of possible values ​​for categorical features is usually limited, and can also be called sparse features. For example, the value of gender is 0 (indicating male) and 1 (indicating female), and the value of time in passenger information or order request information is any integer from 0 to 23, for a total of 24 values.

[0053] After obtaining sparse features, we can construct cross-features by pairwise combining them. Cross-features are essentially sparse features. For example, the cross-feature of driver gender and driver dispatch time can be constructed by determining whether male and female drivers dispatch between 0:00 and 23:00. The value can be represented by a number between 0 and 47, for a total of 2 × 24 = 48 values.

[0054] Sparse features and cross-features are processed one-hot and then mapped into embedding vectors of a certain dimension. Because it is complex for deep models to directly learn sparse features, sparse features are mapped into continuous vectors to represent them. For example, if the driver's gender is 0 for male and 1 for female, the one-hot processing is [1, 0] for male and [0, 1] for female. Then, a matrix is ​​used to represent the one-hot vectors as [0.831456, 0.654385] for male and [0.467321, 0.548974] for female, similar to how continuous features are used for deep model learning. The vector values ​​here are initialized using a normal distribution with a mean of 0 and a variance of 1 and are automatically learned in the model. The dimension of the embedding vector can be adjusted, for example, it can be set to 4 dimensions.

[0055] Embedding features and continuous features are horizontally concatenated. For example, if the driver's gender is a 4-dimensional vector, the departure time is a 4-dimensional vector, and the continuous feature is 100-dimensional, after horizontal concatenation, the total feature dimension is 108.

[0056] The passenger order information and order request information are input into the reservation network model in the form of the above-mentioned horizontal splicing result. Figure 3 As shown in the figure, the feature layer refers to the input layer of the predetermined network model, that is, the input is: cross features, embedding vectors converted after splicing sparse features, and continuous features.

[0057] The predefined network model includes a first subnetwork N1 and a second subnetwork N2. The first subnetwork N1 is used to predict the order placement and acceptance rate, while the second subnetwork N2 is used to predict the order completion rate. The output of the predefined network model is the product of the outputs of the first and second subnetworks N1 and N2, which is the order placement and completion rate.

[0058] Each sub-network in the first sub-network N1 and the second sub-network N2 can be a combination of a linear model and a multi-layer perceptron, and the final prediction result is obtained through a fully connected layer.

[0059] In some embodiments, the first sub-network and the second sub-network have the same structure, and the parameters are located at the same position in the first sub-network and the second sub-network, the difference being that the values ​​of the parameters at the same position are different (i.e., some parameter values ​​are different, or all parameter values ​​are different).

[0060] On the one hand, the subnetwork structures for predicting the order acceptance rate and the order completion rate are the same, and the parameters are located at the same position in the first subnetwork and the second subnetwork, which will not lead to the "seesaw" phenomenon in which the output increase of one subnetwork will inevitably lead to the output decrease of the other subnetwork during model training or updating.

[0061] On the other hand, the subnetwork structures for predicting the order placement acceptance rate and the order completion rate are identical, and the parameters are located in the same positions in the first and second subnetworks. This prevents differences caused by different models from being amplified, ensuring that the predicted values ​​of the order placement acceptance rate and order completion rate are of the same magnitude, with minimal difference. This prevents the situation where, when combining the order placement acceptance rate and order completion rate to determine the order recommendation list, one metric has little or no influence on the result due to the significant difference in magnitude between the two metrics. For example, if the order placement acceptance rate is between 0.8-0.9 and the order completion rate is between 0.05-0.1, the order recommendation list determined based on the sum of the order placement acceptance rate and the order completion rate will be almost identical to the order recommendation list determined based solely on the order placement acceptance rate. In other words, the order completion rate has little influence on the result.

[0062] Furthermore, since the first and second sub-networks have identical inputs and structures, different parameter values ​​at the same position in the sub-networks can lead to differences in the predicted results, allowing the predicted results to represent two different metrics: order acceptance rate and order completion rate. In some embodiments, the first and second sub-networks can be trained separately or jointly.

[0063] S130: Sort the passenger orders in descending order according to the order completion rate corresponding to the passenger orders to obtain an order recommendation list, which is sent to the driver side so that the driver side can select an order from the recommendation list to accept the order.

[0064] In some embodiments, a recommended order list may be generated based solely on the order completion rate corresponding to a passenger's order, or based on both the order completion rate and the order completion rate. For example, in some embodiments, passenger orders are sorted in descending order based on their order completion rates and order completion rates to generate a recommended order list.

[0065] In some embodiments, for each passenger order, the sum of the corresponding order completion rate and the order completion rate can be calculated, and then the sum results corresponding to multiple passenger orders can be sorted in descending order to obtain an order recommendation list.

[0066] Furthermore, in some embodiments, the order completion rate and order completion rate corresponding to each passenger order can be weighted and summed, and the passenger orders can be sorted in descending order according to the sum result to obtain a recommendation list.

[0067] The order completion rate corresponds to the first weight, and the order completion rate corresponds to the second weight. The first and second weights can be pre-set in the online car-hailing platform or set by the driver. For example, the driver's APP can set the first and second weights to default values, or it can display a control for adjusting the weights, and the driver can adjust the weights himself. For example, the control for adjusting the weights can be as follows Figure 4 As shown, the left side of the slider D represents the first weight, and the right side of the slider represents the second weight. The driver user can adjust the position of the slider on the slide bar E to adjust the first weight and the second weight.

[0068] In some embodiments, the driver user is unclear about the meaning of the first weight and the second weight, so it is usually difficult to adjust the weight to the required position according to their own needs. In this regard, the order acceptance tendency can be set on the driver side for the driver user to choose, wherein the order acceptance tendency includes a tendency to accept more orders and a tendency to have a high completion rate. The driver user can choose one of the order acceptance tendencies according to their needs. On the online car-hailing platform, the first preset weight and the second preset weight corresponding to each order acceptance tendency can be pre-stored. After receiving the order acceptance tendency selected by the driver, the first weight and the second weight can be set to the first preset weight and the second preset weight corresponding to the order acceptance tendency selected by the driver.

[0069] Furthermore, in some embodiments, each order acceptance tendency may also include a tendency value. For example, each order acceptance tendency may include five tendency values: 1-2-3-4-5. Accordingly, after receiving the order acceptance tendency and tendency value selected by the driver, the first weight and the second weight may be set to the first preset weight and the second preset weight corresponding to the order acceptance tendency and tendency value selected by the driver.

[0070] In some embodiments, for each passenger order, in addition to summing the order completion rate and the order completion rate as described above, other calculation methods may be used, such as multiplication, inverse trigonometric function, etc.

[0071] After obtaining the order recommendation list, when sending the order recommendation list to the driver side, you can only send a predetermined number of passenger orders that are ranked at the top, or you can send this order recommendation list to the driver side, and the driver side will determine which passenger orders to present to the driver.

[0072] Gradient descent can be used to sort passenger orders in descending order. The following describes the training process of the reservation network model, using the task of predicting the order acceptance rate or order completion rate for a passenger order as an example.

[0073] During offline training, a probability prediction value between 0 and 1 is output for each task, and the corresponding true label of the task is 0 or 1, indicating whether the task is completed. The task loss value can be obtained using the probability prediction value and the true label of the task. Multiple tasks will obtain multiple loss values, and multiple tasks are trained in parallel. The loss of the model in this solution is composed of the weighted order placement and order acceptance loss and order placement and completion loss. Because it is estimated from the entire space starting from the order placement, and because the order placement and completion estimate includes the probability of order acceptance and completion, the loss of order acceptance and completion is not considered in the overall loss. The final loss function formula is w 发单接单 *L 发单接单 +w 发单完单 *L 发单完单 +w 2 发单接单 +w 2 发单完单 , where * represents a multiplication operation, weights w are automatically learned by the model, and the initial weights are randomly initialized using a normal distribution. L is the loss for each task. The weights w here are used during model training and are different from the first and second weights mentioned above. After obtaining the overall model loss function, the network parameters for each task are adjusted through reverse gradient optimization to minimize the overall loss function, thereby achieving the goal of the model's predicted values ​​approaching the true labels of each task.

[0074] During offline training of the reservation network model, order matching logs are obtained and parsed to derive model input data. These include driver and passenger order matching features, such as departure time difference, starting and ending distances, and order placement time. Passenger and driver profile features are also obtained, including age, gender, and the number of orders placed, accepted, and completed in the past month. These features are used as model inputs for offline training. Once the optimal model is obtained, it is uploaded to the ride-hailing platform's online order ranking system for use. When the reservation network model is used online, real-time driver requests are received. The aforementioned features are used to predict the order acceptance rate and order completion rate. The scores are then summed using specific weights to generate a final score. Passenger orders are then ranked based on this score to create a recommended order list, which is then returned to the driver. The recommended order list and the passenger orders that the driver accepts from the recommended order list are logged for modeling.

[0075] The method for determining a recommended order list provided in this specification inputs driver-side order acceptance request information and passenger order information into a predetermined network model comprising a first subnetwork and a second subnetwork. The first subnetwork predicts the order acceptance rate, while the second subnetwork predicts the order completion rate. The product of the outputs of the first and second subnetworks is used as the order completion rate output by the predetermined network model. Passenger orders are then sorted according to the order completion rate to generate a recommended order list. Because the order completion rate in this solution is derived by multiplying the order acceptance rate and order completion rate obtained from different subnetworks, it is the result of the combined effect of these two rates. Therefore, selecting passenger orders based on the recommended list sorted by order completion rate balances both the order acceptance rate and the order completion rate, avoiding the issue of one indicator being too high while the other is too low.

[0076] This specification provides a device for determining an order recommendation list, which can be used to implement Figure 2 As shown in the method. Figure 5 As shown, the device includes an acquisition unit 10, a processing unit 20 and a sorting unit 30.

[0077] The acquisition unit 10 is used to obtain information on multiple passenger orders in a waiting state in real time after receiving the order acceptance request information from the driver side.

[0078] The processing unit 20 is used to input each passenger's order information and the order acceptance request information into the reservation network model respectively to obtain the order completion rate corresponding to each passenger order; wherein, the reservation network model includes a first subnetwork and a second subnetwork, and the passenger order information and the driver's order acceptance request information are both inputs of the first subnetwork and inputs of the second subnetwork; the first subnetwork is used to predict the order acceptance rate, and the second subnetwork is used to predict the order completion rate, and the product of the output values ​​of the first subnetwork and the second subnetwork is used as the order completion rate.

[0079] The sorting unit 30 is used to sort each passenger order in descending order according to the order completion rate corresponding to the passenger order, and obtain an order recommendation list. The order recommendation list is used to send it to the driver side so that the driver side can select an order from the recommendation list to accept the order.

[0080] In some embodiments, the first sub-network and the second sub-network have the same structure, and the values ​​of parameters at the same position are different.

[0081] In some embodiments, the output of the reservation network model also includes the order completion rate; the sorting unit includes: a first sorting sub-unit, which is used to sort each passenger's order in descending order according to the order completion rate and order completion rate corresponding to the passenger's order to obtain an order recommendation list.

[0082] In some embodiments, the first sorting subunit includes a summing subunit and a second sorting subunit.

[0083] The summing subunit is used to perform a weighted sum of the order completion rate and the order completion rate corresponding to each passenger order. The second sorting subunit is used to sort the passenger orders in descending order according to the summing result to obtain a recommendation list.

[0084] In some embodiments, the order completion rate corresponds to the first weight, and the order completion rate corresponds to the second weight; the first sorting subunit also includes: a first receiving subunit, which is used to receive the values ​​of the first weight and the second weight set by the user.

[0085] In some embodiments, the first receiving subunit includes a second receiving subunit and a first setting subunit.

[0086] The second receiving subunit is configured to receive the order acceptance tendency selected by the driver, wherein the order acceptance tendency includes a preference for accepting more orders and a preference for a high order completion rate. The first setting subunit is configured to set the first weight and the second weight to the first preset weight and the second preset weight corresponding to the order acceptance tendency selected by the driver.

[0087] In some embodiments, the first receiving subunit includes: a third receiving subunit, used to receive the order acceptance tendency and tendency value selected by the driver, the order acceptance tendency includes a tendency to accept more orders and a tendency to have a high order completion rate, and each order acceptance tendency includes multiple tendency values; a second setting subunit, used to set the first weight and the second weight to the first preset weight and the second preset weight corresponding to the tendency value under the order acceptance tendency selected by the driver.

[0088] The description and beneficial effects of the above-mentioned order recommendation list determination device can be found in the description and beneficial effects of the method part, which will not be repeated here.

[0089] The embodiment of the present invention further provides an electronic device, such as Figure 6 As shown, the electronic device may include a processor 601 and a memory 602, wherein the processor 601 and the memory 602 may be connected via a bus or other means. Figure 6 The bus connection is taken as an example.

[0090] The processor 601 may be a central processing unit (CPU). The processor 601 may also be other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, or a combination of the above chips.

[0091] The memory 602 is a non-transitory computer-readable storage medium that can be used to store non-transitory software programs, non-transitory computer executable programs, and modules, such as the program instructions / modules corresponding to the method for determining the order recommendation list in the embodiment of the present invention (for example, Figure 5 The processor 601 executes various functional applications and data classification of the processor by running the non-transitory software programs, instructions, and modules stored in the memory 602, thereby implementing the order recommendation list determination method in the above method embodiment.

[0092] The memory 602 may include a program storage area and a data storage area, wherein the program storage area may store an operating system and application programs required for at least one function; the data storage area may store data created by the processor 601, etc. In addition, the memory 602 may include a high-speed random access memory, and may also include a non-volatile memory, such as at least one disk storage device, a flash memory device, or other non-volatile solid-state storage device. In some embodiments, the memory 602 may optionally include a memory remotely located relative to the processor 601, and these remote memories may be connected to the processor 601 via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

[0093] The one or more modules are stored in the memory 602 and when executed by the processor 601, perform the following steps: Figure 2 The method for determining the order recommendation list in the illustrated embodiment.

[0094] For details of the above electronic equipment, please refer to Figure 2 The relevant descriptions and effects in the corresponding embodiments have been understood and will not be repeated here.

[0095] This specification provides a computer storage medium that stores computer program instructions. When the computer program instructions are executed by a processor, Figure 2Steps of the method shown.

[0096] This specification provides a computer program product, which includes a computer program that, when executed by a processor, implements Figure 2 Steps of the method shown.

[0097] Those skilled in the art will appreciate that all or part of the processes in the above-mentioned embodiments can be implemented by instructing the relevant hardware through a computer program. The program can be stored in a computer-readable storage medium. When the program is executed, it can include the processes of the embodiments of the above-mentioned methods. The storage medium can be a magnetic disk, an optical disk, a read-only memory (ROM), a random access memory (RAM), a flash memory, a hard disk drive (HDD), or a solid-state drive (SSD). The storage medium can also include a combination of the above-mentioned types of memory.

[0098] The various embodiments in this specification are described in a progressive manner, and the same or similar parts between the various embodiments can be referred to each other. Each embodiment focuses on the differences from other embodiments.

[0099] The systems, devices, modules or units described in the above embodiments may be implemented by computer chips or entities, or by products with certain functions.

[0100] For the convenience of description, the above devices are described as being divided into various units according to their functions. Of course, when implementing this application, the functions of each unit can be implemented in the same or multiple software and / or hardware.

[0101] Through the description of the above embodiments, it can be seen that those skilled in the art can clearly understand that the present application can be implemented by means of software plus the necessary general hardware platform. Based on this understanding, the technical solution of the present application, or the part that contributes to the prior art, can be embodied in the form of a software product. The computer software product can be stored in a storage medium, such as ROM / RAM, a magnetic disk, an optical disk, etc., and includes a number of instructions for enabling a computer device (which can be a personal computer, a server, or a network device, etc.) to execute certain parts of the methods of each embodiment of the present application.

[0102] The present application can be used in a wide variety of general-purpose or specialized computer system environments or configurations, such as personal computers, server computers, handheld or portable devices, tablet devices, multiprocessor systems, microprocessor-based systems, set-top boxes, programmable consumer electronics, network PCs, minicomputers, mainframe computers, and distributed computing environments that include any of the above.

[0103] The present application may be described in the general context of computer-executable instructions executed by a computer, such as program modules. Generally, program modules include routines, programs, objects, components, data structures, etc. that perform specific tasks or implement specific abstract data types. The present application may also be practiced in distributed computing environments where tasks are performed by remote processing devices connected through a communications network. In a distributed computing environment, program modules may be located in local and remote computer storage media, including storage devices.

[0104] Although the present application has been described through embodiments, those skilled in the art will appreciate that there are many modifications and variations to the present application without departing from the spirit of the present application. It is intended that the appended claims include these modifications and variations without departing from the spirit of the present application.

Claims

1. A method for determining an order recommendation list, characterized in that: include: After receiving the order request information from the driver, obtain the information of multiple passenger orders in the waiting state in real time; Each passenger order information and the order acceptance request information are input into the reservation network model respectively to obtain the order completion rate corresponding to each passenger order; wherein, the reservation network model includes a first subnetwork and a second subnetwork, and the passenger order information and the order acceptance request information of the driver side are both the input of the first subnetwork and the input of the second subnetwork; the first subnetwork and the second subnetwork have the same structure, and the parameters are located at the same position of the first subnetwork and the second subnetwork, the difference is that the values ​​of the parameters at the same position are different; the first subnetwork and the second subnetwork are trained simultaneously, and the loss of the reservation network model during training includes order issuance and acceptance loss and order completion loss; the first subnetwork is used to predict the order issuance and acceptance rate, and the second subnetwork is used to predict the order completion rate, and the predetermined operation result between the output values ​​of the first subnetwork and the second subnetwork is used as the order completion rate; the order issuance and acceptance rate refers to the probability that a passenger order is accepted by the driver after a driver order is matched with a passenger order; the order completion rate refers to the probability that the driver will eventually take the passenger to the destination and complete the order after receiving the passenger order; According to the order completion rate corresponding to the passenger orders, each passenger order is sorted in descending order to obtain an order recommendation list, which is sent to the driver side so that the driver side can select an order from the recommendation list to accept the order.

2. The method according to claim 1, characterized in that The output of the reservation network model also includes an order completion rate; and according to the order completion rate corresponding to the passenger order, each passenger order is sorted in descending order to obtain an order recommendation list, which also includes: According to the order completion rate and order completion rate corresponding to the passenger orders, each passenger order is sorted in descending order to obtain an order recommendation list.

3. The method according to claim 2, characterized in that According to the order completion rate and order completion rate corresponding to the passenger orders, each passenger order is sorted in descending order to obtain an order recommendation list, including: Take the weighted sum of the order completion rate and order acceptance rate for each passenger order; According to the summation result, sort the orders of each passenger in descending order to obtain a recommendation list.

4. The method according to claim 3, characterized in that The order completion rate corresponds to the first weight, and the order completion rate corresponds to the second weight. Accordingly, before the weighted sum of the order completion rate and the order completion rate corresponding to each passenger order, the following is also included: Receive the values ​​of the first weight and the second weight set by the user.

5. The method according to claim 4, characterized in that Receiving the values ​​of the first weight and the second weight set by the user, including: receiving a driver's selected order acceptance tendency, wherein the order acceptance tendency includes a preference for accepting more orders and a preference for a high order completion rate; The first weight and the second weight are set to the first preset weight and the second preset weight corresponding to the order acceptance tendency selected by the driver.

6. The method according to claim 5, characterized in that Receiving the values ​​of the first weight and the second weight set by the user, including: Receive the order acceptance tendency and tendency value selected by the driver, wherein the order acceptance tendency includes a tendency to accept more orders and a tendency to have a high order completion rate, and each order acceptance tendency includes multiple tendency values; The first weight and the second weight are set to the first preset weight and the second preset weight corresponding to the tendency value under the order acceptance tendency selected by the driver.

7. A device for determining an order recommendation list, characterized in that: include: An acquisition unit, configured to acquire information of multiple passenger orders in a waiting order state in real time after receiving an order acceptance request from the driver; The processing unit is used to input each passenger order information and the order acceptance request information into the reservation network model respectively to obtain the order completion rate corresponding to each passenger order; wherein, the reservation network model includes a first subnetwork and a second subnetwork, and the passenger order information and the order acceptance request information of the driver side are both inputs of the first subnetwork and the second subnetwork; the first subnetwork and the second subnetwork have the same structure, and the parameters are located at the same position of the first subnetwork and the second subnetwork, except that the values ​​of the parameters at the same position are different; the first subnetwork and the second subnetwork are trained simultaneously, and the loss of the reservation network model during training includes order issuance and acceptance loss and order completion loss; the first subnetwork is used to predict the order issuance and acceptance rate, and the second subnetwork is used to predict the order completion rate, and the predetermined operation result between the output values ​​of the first subnetwork and the second subnetwork is used as the order completion rate; the order issuance and acceptance rate refers to the probability that a passenger order is accepted by the driver after a driver order is matched with a passenger order; the order completion rate refers to the probability that the driver will eventually take the passenger to the destination and complete the order after receiving the passenger order; The sorting unit is used to sort each passenger order in descending order according to the order completion rate corresponding to the passenger order to obtain an order recommendation list, which is sent to the driver side so that the driver side can select an order from the recommendation list to accept the order.

8. An electronic device, characterized in that: include: A memory and a processor, wherein the processor and the memory are communicatively connected to each other, the memory stores computer instructions, and the processor implements the steps of the method according to any one of claims 1 to 6 by executing the computer instructions.

9. A computer storage medium, characterized in that The computer storage medium stores computer program instructions, and when the computer program instructions are executed by a processor, the steps of the method according to any one of claims 1 to 6 are implemented.

Images