Delivery Method Discrimination Method, Device, Storage Medium and Electronic Device

Abstract

The present disclosure relates to a delivery method discrimination method, apparatus, storage medium, and electronic device. The delivery method discrimination method includes: obtaining order information of completed deliveries within a target AOI, where the order information includes Wi-Fi feature data, AOI feature data, and pedometer feature data of each order within the target AOI; inputting the Wi-Fi feature data, the AOI feature data, and the pedometer feature data of each order within the target AOI into a delivery method discrimination model to obtain a result output by the delivery method discrimination model for characterizing the delivery method of the target AOI. By using this method, according to the Wi-Fi feature data, AOI feature data, and pedometer feature data of each order with completed deliveries within the target AOI, the order delivery method corresponding to the target AOI can be identified, thereby solving the problem in the related art that it is impossible to identify the specific method by which the delivery person completes the delivery to the customer stage.

Description

Technical Field

[0001] The present disclosure relates to the field of computer technologies, and in particular, to a delivery method discrimination method, apparatus, storage medium, and electronic device. Background Art

[0002] In the scenario of food delivery services, there are multiple delivery methods, and the delivery difficulties corresponding to different delivery methods are different. For example, a delivery person walks to the seventh floor to complete food delivery. Another example is that the delivery person takes an elevator upstairs to complete food delivery. Obviously, the delivery difficulties of these two delivery methods are different.

[0003] In related technologies, a delivery person is located through GPS to determine the real-time position of the delivery person. However, using GPS can only locate the coordinates of the delivery person on a two-dimensional plane, and it is impossible to determine the position of the delivery person in a three-dimensional space. Moreover, the reliability of GPS signals is relatively low near buildings and indoors. Therefore, currently, the GPS technology cannot identify the specific delivery method by which the delivery person completes the delivery to the customer stage. Summary of the Invention

[0004] The purpose of the present disclosure is to provide a delivery method discrimination method, apparatus, storage medium, and electronic device to solve the problems existing in related technologies.

[0005] To achieve the above purpose, according to the first aspect of the embodiments of the present disclosure, a delivery method discrimination method is provided, and the method includes:

[0006] Obtain order information of completed deliveries within a target AOI (Area Of Interest, also known as an area of interest, referring to a regional geographical entity in map data), where the order information includes Wi-Fi feature data, AOI feature data, and pedometer feature data of each order within the target AOI;

[0007] Input the Wi-Fi feature data, the AOI feature data, and the pedometer feature data of each order within the target AOI into a delivery method discrimination model, and obtain a result output by the delivery method discrimination model for characterizing the delivery method of the target AOI.

[0008] Optionally, the Wi-Fi feature data includes: the average value of the Wi-Fi similarities of each order within the target AOI, where the Wi-Fi similarity of each order is the maximum value among the similarities between the first Wi-Fi data list of the order and each list in the second Wi-Fi data list set;

[0009] Among them, the first Wi-Fi data list of each order is scanned by the user terminal when the user corresponding to the order places an order. Correspondingly, the set of the second Wi-Fi data lists of each order is a set of lists scanned in real time by the delivery terminal during the delivery process of the order corresponding to the order.

[0010] Optionally, the calculation formula for the Wi-Fi similarity of each order is:

[0011]

[0012] Among them, S represents the Wi-Fi similarity of the order, Mri represents the Wi-Fi data list at the i-th moment in the set of the second Wi-Fi data lists of the order, and Mu represents the first Wi-Fi data list of the order.

[0013] Optionally, the Wi-Fi feature data further includes: the average value of the Wi-Fi arrival times of each order within the target AOI, and the median of the Wi-Fi arrival times of each order; where the Wi-Fi arrival time is the difference between the arrival time and the departure time of the delivery terminal determined by the user terminal through Wi-Fi scanning; and / or,

[0014] The Wi-Fi feature data further includes: the average length value of the arrival and departure Wi-Fi lists of the target AOI, and the proportion of the number of sets in all the sets of the second Wi-Fi data lists within the AOI where the list length change is greater than a preset threshold;

[0015] Among them, the average length value of the arrival and departure Wi-Fi lists of the target AOI is calculated in the following manner:

[0016] For each order within the target AOI, the average value of the lengths of all the lists in the set of the second Wi-Fi data lists of the order that are between the arrival time and the departure time is used as the first average value of the order;

[0017] The average length value of the arrival and departure Wi-Fi lists of the target AOI is calculated based on the first average values of all the orders within the target AOI.

[0018] Optionally, the pedometer feature data includes: the average value of the steps collected by the delivery terminals corresponding to each order within the target AOI between the arrival time and the departure time.

[0019] Optionally, the AOI feature data includes: the preset quantiles of the delivery floors of each order within the target AOI, the proportion of orders without delivery floor numbers among the orders, the delivery address types of the orders, and the feature words in the delivery addresses, where the feature words are extracted by a deep learning model.

[0020] Optionally, the delivery method discrimination model is trained in the following manner:

[0021] Obtain the Wi-Fi feature data, AOI feature data, and pedometer feature data of historical orders within at least one AOI, where each historical order corresponds to barometer data;

[0022] According to the barometer data, add labels for characterizing the order delivery method to the Wi-Fi feature data, the AOI feature data, and the pedometer feature data of each historical order;

[0023] Construct model training samples based on the Wi-Fi feature data, the AOI feature data, and the pedometer feature data with the labels;

[0024] Train the delivery method discrimination model based on the model training samples.

[0025] Optionally, the constructing model training samples based on the Wi-Fi feature data, the AOI feature data, and the pedometer feature data with the labels includes:

[0026] Fuse the Wi-Fi feature data, the AOI feature data, and the pedometer feature data with the labels within each AOI to obtain the order delivery feature vector of the AOI, and use the order delivery feature vector as the model training sample;

[0027] The inputting the Wi-Fi feature data, the AOI feature data, and the pedometer feature data of each order within the target AOI into the delivery method discrimination model to obtain the result output by the delivery method discrimination model for characterizing the delivery method of the target AOI includes:

[0028] Generate the order delivery feature vector of the target AOI according to the Wi-Fi feature data, the AOI feature data, and the pedometer feature data of each order within the target AOI, and input the order delivery feature vector of the target AOI into the delivery method discrimination model to obtain the result output by the delivery method discrimination model for characterizing the delivery method of the target AOI.

[0029] Optionally, the loss function of the delivery method discrimination model is the cross-entropy loss function:

[0030]

[0031] Among them, s j is the value after the order delivery feature vector is mapped by the softmax function, which is used to represent the probability that the AOI corresponding to the order delivery feature vector belongs to the j-th delivery method. The value range of j is from 1 to T, and T is a natural number greater than 1, which is used to represent that there are T delivery methods. y j represents the label of the model training sample.

[0032] Optionally, the method further includes:

[0033] When generating a new order to be delivered, determining the target AOI to which the order to be delivered belongs according to the delivery address of the order to be delivered;

[0034] Determining the delivery parameters of the order to be delivered according to the delivery method of the target AOI to which the order to be delivered belongs.

[0035] Optionally, the method further includes:

[0036] During the process of the delivery person delivering the order to be delivered, detecting the location information of the delivery terminal of the delivery person;

[0037] In response to determining that the delivery person arrives at the target AOI to which the order to be delivered belongs through the location information, prompting the delivery method corresponding to the target AOI to the delivery person through the delivery terminal.

[0038] According to the second aspect of the embodiments of the present disclosure, a delivery method discrimination device is provided. The device includes:

[0039] An acquisition module, configured to acquire order information of completed deliveries within a target AOI, where the order information includes Wi-Fi feature data, AOI feature data, and pedometer feature data of each order within the target AOI;

[0040] An input module, configured to input the Wi-Fi feature data, the AOI feature data, and the pedometer feature data of each order within the target AOI into a delivery method discrimination model, and obtain a result output by the delivery method discrimination model for representing the delivery method of the target AOI.

[0041] Optionally, the Wi-Fi feature data includes: the average value of the Wi-Fi similarities of each order within the target AOI, where the Wi-Fi similarity of each order is the maximum value among the similarities between the first Wi-Fi data list of the order and each list in the second Wi-Fi data list set;

[0042] Among them, the first Wi-Fi data list of each order is scanned by the user terminal when the user corresponding to the order places an order. Correspondingly, the set of the second Wi-Fi data lists of each order is a set of lists scanned in real time by the delivery terminal during the process of delivering the order by the deliveryman corresponding to the order.

[0043] Optionally, the calculation formula for the Wi-Fi similarity of each order is:

[0044]

[0045] Among them, S represents the Wi-Fi similarity of the order, Mr i represents the Wi-Fi data list at the i-th moment in the set of the second Wi-Fi data lists of the order, and Mu represents the first Wi-Fi data list of the order.

[0046] Optionally, the average value of the Wi-Fi arrival times of each order within the target AOI, and the median of the Wi-Fi arrival times of each order; where the Wi-Fi arrival time is the difference between the arrival time and the departure time of the delivery terminal determined by the user terminal through Wi-Fi scanning; and / or,

[0047] The Wi-Fi feature data further includes: the average length value of the inbound and outbound Wi-Fi lists of the target AOI, and the proportion of the number of sets in which the list length changes by more than a preset threshold in all the sets of the second Wi-Fi data lists within the AOI;

[0048] Among them, the average length value of the inbound and outbound Wi-Fi lists of the target AOI is calculated in the following way:

[0049] For each order within the target AOI, the average value of the lengths of all the lists in the set of the second Wi-Fi data lists of the order that are between the arrival time and the departure time is used as the first average value of the order;

[0050] The average length value of the inbound and outbound Wi-Fi lists of the target AOI is calculated based on the first average values of all the orders within the target AOI.

[0051] Optionally, the average value of the number of steps collected by the delivery terminals corresponding to each order within the target AOI between the arrival time and the departure time.

[0052] Optionally, the AOI feature data includes: the preset quantiles of the delivery floors of each order within the target AOI, the proportion of orders without delivery floor numbers among the orders, the types of delivery addresses of the orders, and the feature words in the delivery addresses, where the feature words are extracted by a deep learning model.

[0053] Optionally, the delivery method discrimination model is trained in the following manner:

[0054] Obtain the Wi-Fi feature data, AOI feature data, and pedometer feature data of historical orders within at least one AOI, where each historical order corresponds to barometer data;

[0055] According to the barometer data, add labels for characterizing the order delivery method to the Wi-Fi feature data, the AOI feature data, and the pedometer feature data of each historical order;

[0056] Construct model training samples based on the Wi-Fi feature data, the AOI feature data, and the pedometer feature data with the labels;

[0057] Train the delivery method discrimination model based on the model training samples.

[0058] Optionally, the constructing model training samples based on the Wi-Fi feature data, the AOI feature data, and the pedometer feature data with the labels includes:

[0059] Fuse the Wi-Fi feature data, the AOI feature data, and the pedometer feature data with the labels within each AOI to obtain the order delivery feature vector of the AOI, and use the order delivery feature vector as the model training sample;

[0060] The inputting the Wi-Fi feature data, the AOI feature data, and the pedometer feature data of each order within the target AOI into the delivery method discrimination model to obtain the result for characterizing the delivery method of the target AOI output by the delivery method discrimination model includes:

[0061] Generate the order delivery feature vector of the target AOI according to the Wi-Fi feature data, the AOI feature data, and the pedometer feature data of each order within the target AOI, and input the order delivery feature vector of the target AOI into the delivery method discrimination model to obtain the result for characterizing the delivery method of the target AOI output by the delivery method discrimination model.

[0062] Optionally, the loss function of the delivery method discrimination model is the cross-entropy loss function:

[0063]

[0064] Among them, s j is the value after the order delivery feature vector is mapped by the softmax function, and is used to represent the probability that the AOI corresponding to the order delivery feature vector belongs to the j-th delivery method. The value range of j is from 1 to T, and T is a natural number greater than 1, which is used to represent that there are T delivery methods. y j represents the label of the model training sample.

[0065] Optionally, the device further includes:

[0066] A first determination module, configured to determine the target AOI to which the to-be-delivered order belongs according to the delivery address of the to-be-delivered order when generating a new to-be-delivered order;

[0067] A second determination module, configured to determine the delivery parameters of the to-be-delivered order according to the delivery method of the target AOI to which the to-be-delivered order belongs.

[0068] Optionally, the device further includes:

[0069] A detection module, configured to detect the location information of the delivery terminal of the deliveryman during the process of the deliveryman delivering the to-be-delivered order;

[0070] A prompt module, configured to, in response to determining that the deliveryman arrives at the target AOI to which the to-be-delivered order belongs through the location information, prompt the delivery method corresponding to the target AOI to the deliveryman through the delivery terminal.

[0071] According to the third aspect of the embodiments of the present disclosure, there is provided a computer-readable storage medium, on which a computer program is stored, and when the program is executed by a processor, the steps of any of the methods in the first aspect are implemented.

[0072] According to the fourth aspect of the embodiments of the present disclosure, there is provided an electronic device, including:

[0073] A memory, on which a computer program is stored;

[0074] A processor, configured to execute the computer program in the memory to implement the steps of any of the methods in the first aspect.

[0075] By adopting the above technical solutions, at least the following technical effects can be achieved:

[0076] By obtaining the order information of the completed deliveries within the target AOI, the order information includes Wi-Fi feature data, AOI feature data, and pedometer feature data of each order within the target AOI; inputting the Wi-Fi feature data, the AOI feature data, and the pedometer feature data of each order within the target AOI into the delivery mode discrimination model, and obtaining the result output by the delivery mode discrimination model for characterizing the delivery mode of the target AOI. Those of ordinary skill in the art should understand that the delivery modes of orders within the same AOI should be the same. For example, if a certain AOI corresponds to a community without an elevator, then the orders delivered to this community are all completed by walking upstairs. Therefore, by using this method of the present disclosure, according to the Wi-Fi feature data, AOI feature data, and pedometer feature data of each order within the target AOI, the order delivery mode corresponding to the target AOI can be identified.

[0077] Other features and advantages of the present disclosure will be described in detail in the following specific implementation section. BRIEF DESCRIPTION OF THE DRAWINGS

[0078] The drawings are used to provide a further understanding of the present disclosure and constitute a part of the specification. Together with the following specific implementation, they are used to explain the present disclosure, but do not constitute a limitation to the present disclosure. In the drawings:

[0079] Figure 1 is a flowchart of a method for discriminating a delivery mode shown according to an exemplary embodiment of the present disclosure.

[0080] Figure 2 is a schematic diagram of a method for extracting feature words shown according to an exemplary embodiment of the present disclosure.

[0081] Figure 3 is a flowchart of a method for training a delivery mode discrimination model shown according to an exemplary embodiment of the present disclosure.

[0082] Figure 4 is a block diagram of a delivery mode discrimination device shown according to an exemplary embodiment of the present disclosure.

[0083] Figure 5 is a block diagram of an electronic device shown according to an exemplary embodiment of the present disclosure. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0084] The following details the specific implementation of the present disclosure with reference to the drawings. It should be understood that the specific implementation described herein is only for explaining and understanding the present disclosure, and is not used to limit the present disclosure.

[0085] Exemplary embodiments will be described in detail herein, and examples thereof are shown in the accompanying drawings. When the following description refers to the accompanying drawings, unless otherwise indicated, the same numbers in different drawings represent the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with the present disclosure. On the contrary, they are merely examples of devices and methods consistent with some aspects of the present disclosure as detailed in the appended claims.

[0086] In the scenario of food delivery service, there are various delivery methods, and the delivery difficulties corresponding to different delivery methods are different. For example, the deliveryman walks to the seventh floor to complete the food delivery. Another example is that the deliveryman takes the elevator to the seventh floor to complete the food delivery. Obviously, the delivery difficulties of these two delivery methods are different.

[0087] In the related art, the deliveryman is located by GPS to determine the real-time position of the deliveryman. However, using GPS can only locate the coordinates of the deliveryman on a two-dimensional plane, and cannot determine the position of the deliveryman in three-dimensional space, and the reliability of GPS signals is relatively low near buildings and indoors. Therefore, currently, it is impossible to identify the specific way in which the deliveryman completes the delivery at the customer arrival stage using GPS technology.

[0088] Precisely because it is impossible to identify the specific way in which the deliveryman completes the delivery at the customer arrival stage, the pricing of the food delivery fee does not consider the difficulty of the delivery method, resulting in an unreasonable situation in the pricing of the food delivery fee. For example, if the delivery fee for the deliveryman to walk to the seventh floor to complete the food delivery is the same as that for the deliveryman to take the elevator to the seventh floor to complete the food delivery, then this is obviously unreasonable.

[0089] In view of this, the embodiments of the present disclosure provide a delivery method discrimination method, device, storage medium, and electronic device to achieve refined identification of the delivery method in the delivery scenario.

[0090] First, the application background of the present disclosure will be described.

[0091] The technical solution of the present disclosure is used to identify the specific delivery method in the delivery scenario. For example, the technical solution of the present disclosure can be used to identify the merchant delivery method in the pick-up stage of the food delivery service, and can also be used to identify the user delivery method in the customer arrival stage of the delivery service. That is to say, the technical solution of the present disclosure can be used to explore the delivery method of the merchant in addition to exploring the delivery method of the user.

[0092] By identifying the specific delivery methods in the delivery scenario, it is beneficial to depict the distribution process in the arrival stage and obtain a detailed delivery address portrait, that is, it is beneficial to the portrait production of the AOI delivery method. Further, if the AOI delivery method portrait data is applied in the delivery fee pricing system, the delivery fee pricing system can calculate a more reasonable delivery fee according to the specific order delivery method and the delivery difficulty corresponding to the delivery method. In addition, the AOI delivery method portrait can also be used as basic data in systems such as ETA and scheduling systems.

[0093] The technical solution of the present disclosure will be specifically described below according to the embodiments of the present disclosure.

[0094] Figure 1 It is a flowchart of a delivery method discrimination method shown according to an exemplary embodiment of the present disclosure. As Figure 1 shown, the delivery method discrimination method includes:

[0095] S11. Obtain the order information of the completed deliveries within the target area of interest (AOI), where the order information includes the Wi-Fi feature data, AOI feature data, and pedometer feature data of each order within the target AOI.

[0096] Among them, AOI (Area Of Interest) is the area of interest, also known as the information surface or the area of interest. AOI refers to the regional geographical entity in the map data. Obtaining the order information within the target AOI is to obtain the order information whose delivery address / pick-up address belongs to the area of interest.

[0097] Those of ordinary skill in the art should understand that each order corresponds to the Wi-Fi feature data, AOI feature data, and pedometer feature data of the order after the delivery is completed.

[0098] Specifically, the Wi-Fi feature data of each order includes the Wi-Fi data scanned by the user terminal of the corresponding user when placing the order, and the Wi-Fi data scanned in real time by the delivery terminal of the deliveryman during the delivery process of the order. The AOI feature data of each order includes the delivery address information of the order, such as a certain unit, a certain floor, a certain number in a certain community, etc. The pedometer feature data of each order includes the step information collected by the delivery terminal of the deliveryman during the delivery process of the order.

[0099] The order information includes the Wi-Fi feature data, AOI feature data, and pedometer feature data of each order within the target AOI. Specifically, the order information includes the Wi-Fi feature data, AOI feature data, and pedometer feature data of all orders in the target AOI. Among them, it should be noted that the present disclosure does not limit the number of orders within the target AOI.

[0100] S12. Input the Wi-Fi feature data, the AOI feature data, and the pedometer feature data of each order within the target AOI into a delivery method discrimination model, and obtain the result output by the delivery method discrimination model for characterizing the delivery method of the target AOI.

[0101] Those of ordinary skill in the art should understand that the delivery methods of orders within the same AOI are basically the same. For example, if the target AOI corresponds to a community without an elevator, then the orders delivered to this community may all be completed by walking upstairs. Another example, if the target AOI corresponds to a community where outsiders are prohibited from entering, then the orders delivered to this community may all be completed by the delivery person staying at the entrance of the community waiting for the user to pick up the meal. Another example, if the target AOI corresponds to a high-end office building, then the orders delivered to this office building may all be completed by the delivery person taking the elevator upstairs. Therefore, in a realizable implementation manner, inputting the Wi-Fi feature data, the AOI feature data, and the pedometer feature data of each order within the target AOI into the delivery method discrimination model can obtain the result output by the delivery method discrimination model for characterizing a delivery method corresponding to all orders within the target AOI.

[0102] Adopting this method of the present disclosure, based on the Wi-Fi feature data, the AOI feature data, and the pedometer feature data of each order within the target AOI, the delivery method corresponding to the target AOI can be accurately determined, thereby solving the problem in the related art that it is impossible to identify the specific method by which the delivery person completes the delivery at the stage of arriving at the customer.

[0103] It should be noted here that the target AOI in the above method can correspond not only to the area of interest of the ordering user but also to the area of interest of the merchant. When the target AOI in the above method corresponds to the area of interest of the ordering user, by adopting the above method, the delivery method corresponding to the order within the target AOI at the stage of arriving at the customer can be determined. When the target AOI in the above method corresponds to the area of interest of the merchant, by adopting the above method, the delivery method corresponding to the order within the target AOI at the stage of picking up the goods can be determined. In the subsequent embodiments of the present disclosure, mainly taking the target AOI as the area of interest of the ordering user as an example for exemplary illustration.

[0104] Since the above method of the present disclosure can determine the delivery method corresponding to the target AOI. Therefore, based on the order information of the completed deliveries within a large number of different areas of interest, the delivery methods corresponding to different AOIs can be determined. And after determining the delivery method corresponding to each AOI, the delivery method of the new order to be delivered can be determined before or at the same time as generating the new order to be delivered.

[0105] Therefore, in an implementable embodiment, the method may further include the following steps:

[0106] When generating a new order to be delivered, determine the target AOI to which the order to be delivered belongs according to the delivery address of the order to be delivered; determine the delivery parameters of the order to be delivered according to the delivery method of the target AOI to which the order to be delivered belongs.

[0107] Among them, the delivery parameters may be parameters characterizing the delivery fee and the difficulty of delivery.

[0108] Specifically, when the server generates a new order to be delivered, the target AOI to which the order to be delivered belongs can be determined according to the delivery address of the order to be delivered, so that the delivery method of the order to be delivered can be determined. After determining the delivery method of the order to be delivered, relevant delivery parameters such as the difficulty of delivery and the delivery fee of the order can be determined according to the delivery method.

[0109] By adopting this method, the delivery method information of the new order can be added while the server generates a new order, so that the delivery information of the generated new order can be more complete and reasonable. For example, by adopting this method, the delivery fee of the order can be calculated more reasonably. More specifically, for example, the delivery fee for the deliveryman to walk to the seventh floor to complete the food delivery can be higher than the delivery fee for the deliveryman to take the elevator to the seventh floor to complete the food delivery.

[0110] In another implementable embodiment, the method may further include the following steps:

[0111] During the process of the deliveryman delivering the order to be delivered, detect the location information of the delivery terminal of the deliveryman; when it is determined that the deliveryman arrives at the target AOI to which the order to be delivered belongs through the location information, prompt the deliveryman with the delivery method corresponding to the target AOI through the delivery terminal.

[0112] Among them, the delivery method may be any one of the following: the deliveryman stays at the delivery point in the target AOI to wait for the user to complete the delivery, the deliveryman goes up / down the stairs on foot to complete the delivery, and the deliveryman uses a lifting device (such as an elevator) to go up / down the stairs to complete the delivery. The delivery point may be the designated delivery location of the target AOI, such as the west entrance of the community, the east entrance of the community, etc.

[0113] Specifically, during the process of a delivery person delivering a to-be-delivered order, the location information of the delivery terminal is detected through the GPS of the delivery person's mobile terminal. When it is determined through this location information that the delivery person has arrived at the target AOI area to which the to-be-delivered order belongs, the delivery method corresponding to the target AOI is prompted to the delivery person through the delivery terminal. For example, the delivery person can be prompted in a voice manner that the to-be-delivered order requires the delivery person to stay at the west gate of the target AOI and wait for the user to pick up the goods to complete the order delivery. For another example, text information can also be displayed on the delivery terminal to prompt the delivery person that the to-be-delivered order requires the delivery person to walk upstairs to complete the order delivery.

[0114] By adopting this method, when the delivery person arrives at the target AOI area to which the to-be-delivered order belongs, the delivery method of the to-be-delivered order can be prompted to the delivery person, which is convenient for the delivery person to complete the order delivery in a timely manner according to the delivery method.

[0115] An implementable implementation manner is that before inputting the Wi-Fi feature data, the AOI feature data, and the pedometer feature data of each order in the target AOI into the delivery method discrimination model, the method may further include: performing data cleaning on the Wi-Fi feature data, the AOI feature data, and the pedometer feature data to remove abnormal data.

[0116] Among them, abnormal data includes missing data, data with abnormal information parsing, and cached data, etc. For example, if no Wi-Fi data is scanned by the delivery terminal during a certain period when the delivery person is delivering an order, it is considered that the Wi-Fi data during that period is missing. For another example, in the related art, the delivery terminal scans the surrounding Wi-Fi signals every 15 seconds. If the mac address and intensity of the Wi-Fi scanned for the Nth time are the same as those of the Wi-Fi scanned for the (N - 1)th time, it is considered that the Wi-Fi data scanned for the Nth time is cached data. For another example, if the floor information parsed when parsing the delivery address is a null value or the parsed floor number is unreasonable, it is considered that the parsed data is abnormal.

[0117] Inputting the Wi-Fi feature data, the AOI feature data, and the pedometer feature data after removing abnormal data into the delivery method discrimination model can make the result output by the delivery method discrimination model more accurate.

[0118] Optionally, the Wi-Fi feature data includes: the average value of the Wi-Fi similarities of each order within the target AOI, where the Wi-Fi similarity of each order is the maximum value among the similarities between the first Wi-Fi data list of the order and each list in the set of second Wi-Fi data lists; the first Wi-Fi data list of each order is scanned by the user terminal when the user corresponding to the order places an order, and correspondingly, the set of second Wi-Fi data lists of each order is a set of lists scanned in real time by the delivery terminal during the delivery process of the order by the delivery person corresponding to the order.

[0119] It is not difficult to understand that in current society, Wi-Fi is widely used and almost covers every corner of the city. When a user places an order, the user terminal used by the user can scan the mac addresses and corresponding strength information of all Wi-Fi within the effective range nearby, so as to obtain the above-mentioned first Wi-Fi data list. Correspondingly, during the delivery process of each order, the delivery terminal can scan the Wi-Fi signals on the delivery route in real time, so as to obtain the above-mentioned set of second Wi-Fi data lists. Among them, the Wi-Fi data list can specifically be a Wi-Fi Mac address list.

[0120] For each order, the Wi-Fi similarity of the order can be obtained according to the first Wi-Fi data list and the set of second Wi-Fi data lists of the order. In an implementable embodiment, the Wi-Fi similarity of the order is calculated through the following formula:

[0121]

[0122] where S represents the Wi-Fi similarity of the order, Mr i represents the Wi-Fi data list at the i-th moment in the set of second Wi-Fi data lists of the order, Mu represents the first Wi-Fi data list of the order, and the |||| operator represents the length value of the list, specifically, it can be to find out how many Wi-Fi Mac addresses are in the list.

[0123] Exemplarily, if the first Wi-Fi data list of an order is list A and the set of second Wi-Fi data lists is [B, C, D], then the length value of the intersection of list A and list B can be calculated first, and this length value is used as the dividend and divided by the length value of list B to obtain the similarity between list A and list B. Similarly, the similarity between list A and list C is calculated, and the similarity between list A and list D is calculated. Then, the maximum value among the three similarities obtained is used as the Wi-Fi similarity of the order.

[0124] After calculating the Wi-Fi similarities of each order within the target AOI, the average value of the Wi-Fi similarities of each order within the target AOI can be obtained based on the Wi-Fi similarities of each order within the target AOI and the number of orders within the target AOI.

[0125] It is not difficult to understand that when the first Wi-Fi data list is basically the same as a certain list in the second Wi-Fi data list set, it indicates that the delivery person delivers the order items to the location where the user placed the order, that is, delivers to the home. Therefore, the magnitude of the Wi-Fi similarity value of each order can be used to estimate whether the delivery person delivers to the home or stays at a location such as the entrance of the community waiting for the user to pick up the goods. Correspondingly, based on the average value of the Wi-Fi similarities of each order within the target AOI, the delivery method of the target AOI can be estimated.

[0126] Optionally, the Wi-Fi feature data further includes: the average value of the Wi-Fi arrival times of each order within the target AOI, and the median of the Wi-Fi arrival times of each order; wherein, the Wi-Fi arrival time is the difference between the arrival time and the departure time of the delivery terminal determined by the user terminal through Wi-Fi scanning.

[0127] An implementable implementation manner is that the user terminal can scan the wireless signal of the delivery terminal in real time to determine the arrival time and departure time of the delivery terminal. Similarly, the delivery terminal can also scan the wireless signal of the user terminal in real time to determine the arrival time and departure time of the delivery terminal. The specific implementation manner is similar to the method of determining the arrival time and departure time of the delivery person based on the Wi-Fi signal in the related art, and will not be described in detail here.

[0128] For each order, after determining the arrival time and departure time of the delivery terminal based on the Wi-Fi signal, the Wi-Fi arrival time of the order can be further determined. Based on the Wi-Fi arrival times of each order within the target AOI and the number of orders within the AOI, the average value of the Wi-Fi arrival times of each order within the target AOI and the median of the Wi-Fi arrival times of each order within the target AOI can be determined. By way of example, assuming that the Wi-Fi arrival times of each order within the target AOI are 13 minutes, 12 minutes, 10 minutes, 7 minutes, and 6 minutes respectively, then the average value of the Wi-Fi arrival times of each order within the target AOI can be obtained as 9.6 minutes, and the median of the Wi-Fi arrival times of each order within the target AOI is 10 minutes.

[0129] It is not difficult to understand that different delivery methods have different corresponding delivery difficulties, and thus the time spent by the delivery staff in the delivery stage is different. For example, the time spent in the elevator-upstairs delivery method is less than that in the walking-upstairs delivery method. Therefore, based on the average value of the Wi-Fi arrival times of each order within the target AOI and the median of the Wi-Fi arrival times of each order, the time spent by the delivery staff in the delivery stage within this AOI can be determined, and based on this spent time, the delivery method of the delivery staff for the orders within this AOI can be roughly estimated.

[0130] Optionally, the Wi-Fi feature data further includes: the average length value of the arrival and departure Wi-Fi lists of the target AOI, and the proportion of the number of sets in all the second Wi-Fi data list sets within this AOI where the list length change is greater than a preset threshold;

[0131] Among them, the average length value of the arrival and departure Wi-Fi lists of the target AOI is calculated in the following manner:

[0132] For each order within the target AOI, the average value of the lengths of all the lists within the second Wi-Fi data list set of this order that are between the arrival time and the departure time is used as the first average value of this order; the average length value of the arrival and departure Wi-Fi lists of the target AOI is calculated based on the first average values of all the orders within the target AOI.

[0133] Exemplarily, assume that the second Wi-Fi data list set of a certain order is [V, W, X, Y, Z], and all the lists within the second Wi-Fi data list set of this order that are between the arrival time and the departure time of this order are [X, Y, Z]. Then the first average value of this order is the average value of the lengths of the X list, the Y list, and the Z list. If the length of the X list is 22, the length of the Y list is 20, and the length of the Z list is 18, then the first average value of this order is 20. Further, the average length value of the arrival and departure Wi-Fi lists of the target AOI is calculated based on the first average values of all the orders within the target AOI. Exemplarily, assume that the first average values of each order within the target AOI are 20, 15, 17, 18 respectively. Then the average length value of the arrival and departure Wi-Fi lists of the target AOI is 17.5.

[0134] The proportion of the number of sets in all the second Wi-Fi data list sets within the target AOI where the change in list length is greater than a preset threshold. Here, the preset threshold is set according to actual requirements. For example, the preset threshold can be set to 80%. Specifically, for each order within the target AOI, if the Wi-Fi list length at the Nth moment in the second Wi-Fi data list set of this order is 80% greater than the Wi-Fi list length at the (N - 1)th moment, it is considered that there is a mutation in the list length of this second Wi-Fi data list set. Further, through the above method, the proportion of the number of sets with list length mutation in the second Wi-Fi data list sets of all orders within the target AOI can be determined.

[0135] It should be noted that the reason for the mutation in the list length of the second Wi-Fi data list set is that the delivery person may pass through areas with weak or no signal such as elevators, stairs, and basements during the delivery process. Therefore, by determining the proportion of the number of sets in all the second Wi-Fi data list sets within the target AOI where the change in list length is greater than the preset threshold, it can be estimated whether the delivery person passes through areas such as elevators within this AOI during order delivery, and thus the order delivery method within this AOI can be further determined.

[0136] Optionally, the pedometer feature data includes: the average number of steps collected by the delivery terminals corresponding to each order within the target AOI between the arrival time and the departure time.

[0137] It should be understood that for different delivery methods, the number of steps taken by the delivery person during the delivery stage varies greatly. For example, the number of steps generated by the delivery method of waiting for the user to pick up the meal at the community entrance is significantly less than the number of steps generated by the delivery method of the delivery person walking upstairs to complete the delivery. Therefore, according to the average number of steps collected by the delivery terminals corresponding to each order within the target AOI between the arrival time and the departure time, the delivery methods of each order within the target AOI can be determined.

[0138] Optionally, the AOI feature data includes: the preset quantiles of the delivery floors of each order within the target AOI, the proportion of orders without delivery floor numbers among all orders, the types of delivery addresses of each order, and the feature words in the delivery addresses, where the feature words are extracted by a deep learning model.

[0139] It should be noted that the delivery address of each order within the target AOI includes the name of the target AOI. For example, if the delivery address of an order is Room 506, Unit 4, Building 2, Sunshine Community, then the corresponding AOI name for this order can be Sunshine Community.

[0140] Among them, the above preset quantile can be set to the 0.8 quantile. According to the preset quantile of the delivery floors of each order within the target AOI, the number of building floors within this AOI can be estimated. For example, if the 0.8 quantile of the delivery floors of each order within the target AOI is 3, then it can be basically determined that the buildings within this target AOI are low-rise buildings, and low-rise buildings are very likely not equipped with facilities such as elevators.

[0141] Furthermore, according to the proportion of orders without delivery floor numbers among the orders within the target AOI, it can be determined whether this target AOI area is a region where outsiders are prohibited from entering. It is not difficult to understand that when the delivery address in the order does not include detailed information such as floor numbers and room numbers, it indicates that the delivery area corresponding to this order is very likely not accessible to the delivery staff. Further, for such areas where delivery staff are not allowed to enter, the delivery method for orders within such areas is very likely to be the method of staying and waiting for the user to pick up the meal.

[0142] Still further, the AOI feature data can also include the delivery address types of each order and the feature words in the delivery address, where the feature words are extracted by a deep learning model.

[0143] Among them, the delivery address types include residential area type, government agency type, entertainment venue type, medical unit type, enterprise & office building type, hotel type, school type, store type, etc. It should be understood that different types of delivery addresses have different restrictions on the delivery methods in the arrival customer stage. For example, areas such as government agencies and schools may not allow delivery staff to enter. While areas such as residential areas and entertainment venues may allow delivery staff to enter.

[0144] The feature words in the delivery address include college, campus, profession, limited company, university, technology, technique, new village, community, family courtyard, street, number, numbered courtyard, lane, international, square, store, grand hotel, phase, etc. It is worth noting that the delivery address includes the name of the target AOI, and the feature words in the delivery address are exactly extracted from the name of the target AOI.

[0145] With the development of society, there are certain style changes in the naming of new buildings. Therefore, according to the feature words in the delivery address, the completion year information of the building corresponding to this delivery address can be estimated, and according to the completion year information of the building, it can be estimated whether this building area is equipped with facilities such as elevators. Further, the type of delivery method for the AOI area corresponding to this building can be estimated. A feasible implementation method is through Figure 2Extract the feature words in the delivery address in the manner shown. Perform char embedding encoding on the delivery address (AOI name) corresponding to the order to convert it into a vector, and then input it into an LSTM (Long Short-Term Memory) model for feature word extraction to obtain a feature word vector. Exemplarily, assume that the feature words include college, campus, occupation, limited company, university, technology, technique, new village, community, family compound, street, number, courtyard, lane, international, square, store, grand hotel, phase. Then, a 19-dimensional 0-1 vector can be used to represent the feature words in the AOI name. Specifically, mark the corresponding positions of the above-mentioned feature words that appear in the AOI name as 1, and mark the corresponding positions of the feature words that do not appear as 0, so as to obtain a 19-dimensional 0-1 vector representing the feature words in the AOI name.

[0146] Optionally, refer to Figure 3 , the delivery method discrimination model is trained in the following manner:

[0147] S31. Obtain Wi-Fi feature data, AOI feature data, and pedometer feature data of historical orders within at least one AOI, where each of the historical orders corresponds to barometer data;

[0148] S32. According to the barometer data, add labels for characterizing the order delivery method to the Wi-Fi feature data, the AOI feature data, and the pedometer feature data of each of the historical orders;

[0149] S33. Construct a model training sample according to the Wi-Fi feature data, the AOI feature data, and the pedometer feature data with the labels;

[0150] S34. Train the delivery method discrimination model according to the model training sample.

[0151] In the related art, a barometric pressure sensor is configured on the delivery terminals of some deliverymen. The delivery terminal with the barometric pressure sensor can obtain barometer data during the process of the deliveryman delivering an order. The movement state of the deliveryman can be distinguished according to the barometer data. Specifically, the movement state of the deliveryman can be distinguished as a staying state, a walking upstairs state, or a taking the elevator state according to the magnitude of the barometric variance and the rate of barometric change during the delivery process of the deliveryman.

[0152] However, the coverage rate of distribution terminals with barometric pressure sensors is less than 5%, and it is costly to configure barometric pressure sensors for each distribution terminal. Therefore, in the present disclosure, a small amount of historical order information with barometer data is used as training sample data to train a delivery method discrimination model. Specifically, Wi-Fi feature data, AOI feature data, and pedometer feature data of historical orders within at least one AOI are obtained, where each of the historical orders corresponds to barometer data; according to the barometer data, labels for characterizing the order delivery method are added to the Wi-Fi feature data, the AOI feature data, and the pedometer feature data of each of the historical orders; a model training sample is constructed based on the Wi-Fi feature data, the AOI feature data, and the pedometer feature data with the labels; and the delivery method discrimination model is trained based on the model training sample.

[0153] The delivery method discrimination model trained based on a small amount of historical order information with barometer data can discriminate a large amount of historical order information without barometer data to obtain the delivery methods corresponding to a large number of AOIs. In this way, it is not necessary to configure barometer sensors on the distribution terminals of each deliveryman, greatly reducing the cost of identifying the order delivery method.

[0154] Optionally, the constructing a model training sample based on the Wi-Fi feature data, the AOI feature data, and the pedometer feature data with the labels includes:

[0155] Fusing the Wi-Fi feature data, the AOI feature data, and the pedometer feature data with the labels within each AOI to obtain an order delivery feature vector for the AOI, and using the order delivery feature vector as the model training sample;

[0156] The inputting the Wi-Fi feature data, the AOI feature data, and the pedometer feature data of each order within the target AOI into the delivery method discrimination model to obtain the result output by the delivery method discrimination model for characterizing the delivery method of the target AOI includes:

[0157] Generating the order delivery feature vector of the target AOI based on the Wi-Fi feature data, the AOI feature data, and the pedometer feature data of each order within the target AOI, and inputting the order delivery feature vector of the target AOI into the delivery method discrimination model to obtain the result output by the delivery method discrimination model for characterizing the delivery method of the target AOI.

[0158] In this way, for the Wi-Fi feature data, AOI feature data, and pedometer feature data with tags for characterizing the delivery method within each AOI, fusion is performed to obtain the order delivery feature vector of the AOI. The obtained order delivery feature vector of the AOI also has a tag for characterizing the delivery method. Using the order delivery feature vector with the tag as a model training sample to perform supervised learning training on the above-mentioned delivery method discrimination model, a trained delivery method discrimination model can be obtained.

[0159] Optionally, the loss function of the delivery method discrimination model is a cross-entropy loss function:

[0160]

[0161] where s j is the value after the order delivery feature vector is mapped by the softmax function, used to represent the probability that the AOI corresponding to the order delivery feature vector belongs to the j-th delivery method. The value range of j is from 1 to T, and T is a natural number greater than 1, used to represent that there are T delivery methods. y j represents the label of the model training sample.

[0162] Using the above method, the above-mentioned delivery method discrimination model is trained through a small amount of order information with barometer data. According to the trained delivery method discrimination model, the delivery methods of a large amount of order information without barometer data are discriminated regionally, so as to obtain the delivery method corresponding to each AOI region. After determining the delivery method corresponding to each AOI, the delivery method corresponding to each AOI can be used as basic data in the delivery fee pricing system, ETA, and scheduling system. A feasible implementation manner is that when generating a new order, the delivery method of the order can be determined according to the AOI corresponding to the delivery address of the order. Further, the delivery fee pricing system can generate the delivery fee of the order more reasonably according to the delivery method of the order, thereby improving the experience of the delivery staff.

[0163] Based on the same inventive concept, an embodiment of the present disclosure also provides a delivery method discrimination device, as Figure 4 shown. The device 400 includes:

[0164] An acquisition module 410, configured to acquire order information of completed deliveries within a target AOI, where the order information includes Wi-Fi feature data, AOI feature data, and pedometer feature data of each order within the target AOI;

[0165] An input module 420, configured to input the Wi-Fi feature data, the AOI feature data, and the pedometer feature data of each order within the target AOI into a delivery method discrimination model, and obtain a result output by the delivery method discrimination model for characterizing the delivery method of the target AOI.

[0166] By using such a device, by obtaining order information within a target AOI, the order information including the Wi-Fi feature data, the AOI feature data, and the pedometer feature data of each order within the target AOI; inputting the Wi-Fi feature data, the AOI feature data, and the pedometer feature data of each order within the target AOI into a delivery method discrimination model, and obtaining a result output by the delivery method discrimination model for characterizing the delivery method of the target AOI. Therefore, by using the method of the present disclosure, according to the Wi-Fi feature data, the AOI feature data, and the pedometer feature data of each order within the target AOI, the order delivery method corresponding to the target AOI can be identified, thereby solving the problem in the related art that it is impossible to identify the specific method by which the delivery person completes the delivery to the customer stage.

[0167] Optionally, the Wi-Fi feature data includes: the average value of the Wi-Fi similarities of each order within the target AOI, where the Wi-Fi similarity of each order is the maximum value among the similarities between the first Wi-Fi data list of the order and each list in the second Wi-Fi data list set;

[0168] Wherein, the first Wi-Fi data list of each order is scanned by the user terminal when the user corresponding to the order places an order. Correspondingly, the second Wi-Fi data list set of each order is a list set scanned in real time by the delivery terminal when the delivery person corresponding to the order is delivering the order.

[0169] Optionally, the calculation formula for the Wi-Fi similarity of each order is:

[0170]

[0171] Wherein, S represents the Wi-Fi similarity of the order, Mr i represents the Wi-Fi data list at the i-th moment in the second Wi-Fi data list set of the order, and Mu represents the first Wi-Fi data list of the order.

[0172] Optionally, the average value of the Wi-Fi arrival times of each order within the target AOI, and the median value of the Wi-Fi arrival times of each order; wherein, the Wi-Fi arrival time is the difference between the arrival time and the departure time of the delivery terminal determined by the user terminal through Wi-Fi scanning; and / or,

[0173] The Wi-Fi feature data further includes: the average length value of the arrival / departure Wi-Fi list of the target AOI, and the proportion of the number of sets in all the second Wi-Fi data list sets within the AOI where the list length change is greater than a preset threshold;

[0174] Wherein, the average length value of the arrival / departure Wi-Fi list of the target AOI is calculated by the following method:

[0175] For each order within the target AOI, the average value of the lengths of all the lists within the second Wi-Fi data list set of this order that are between the arrival time and the departure time is used as the first average value of this order;

[0176] The average length value of the arrival / departure Wi-Fi list of the target AOI is calculated based on the first average values of all the orders within the target AOI.

[0177] Optionally, the average value of the number of steps collected by the delivery terminal corresponding to each order within the target AOI between the arrival time and the departure time.

[0178] Optionally, the AOI feature data includes: the preset quantiles of the delivery floors of each order within the target AOI, the proportion of orders without delivery floor numbers among each order, the delivery address types of each order, and the feature words in the delivery address, wherein the feature words are extracted by a deep learning model.

[0179] Optionally, the delivery method discrimination model is trained by the following method:

[0180] Obtain the Wi-Fi feature data, AOI feature data, and pedometer feature data of historical orders within at least one AOI, wherein each historical order corresponds to barometer data;

[0181] According to the barometer data, add labels for characterizing the order delivery method to the Wi-Fi feature data, the AOI feature data, and the pedometer feature data of each historical order;

[0182] Construct model training samples based on the Wi-Fi feature data, the AOI feature data, and the pedometer feature data with the labels;

[0183] The delivery mode discrimination model is trained based on the model training samples.

[0184] Optionally, the constructing the model training samples according to the Wi-Fi feature data, the AOI feature data, and the pedometer feature data with the label includes:

[0185] Fusing the Wi-Fi feature data, the AOI feature data, and the pedometer feature data with the label within each AOI to obtain an order delivery feature vector for the AOI, and using the order delivery feature vector as the model training sample;

[0186] The inputting the Wi-Fi feature data, the AOI feature data, and the pedometer feature data of each order within the target AOI into the delivery mode discrimination model to obtain the result output by the delivery mode discrimination model for characterizing the delivery mode of the target AOI includes:

[0187] Generating the order delivery feature vector of the target AOI according to the Wi-Fi feature data, the AOI feature data, and the pedometer feature data of each order within the target AOI, and inputting the order delivery feature vector of the target AOI into the delivery mode discrimination model to obtain the result output by the delivery mode discrimination model for characterizing the delivery mode of the target AOI.

[0188] Optionally, the loss function of the delivery mode discrimination model is a cross-entropy loss function:

[0189]

[0190] where s j is the value after the order delivery feature vector is mapped by the softmax function, and is used to characterize the probability that the AOI corresponding to the order delivery feature vector belongs to the j-th delivery mode, the value range of j is from 1 to T, and T is a natural number greater than 1, which is used to characterize that there are T delivery modes, and y j characterizes the label of the model training sample.

[0191] Optionally, the apparatus 400 further includes:

[0192] A first determination module configured to determine the target AOI to which the to-be-delivered order belongs according to the delivery address of the to-be-delivered order when generating a new to-be-delivered order;

[0193] A second determination module configured to determine the delivery parameters of the to-be-delivered order according to the delivery mode of the target AOI to which the to-be-delivered order belongs.

[0194] Optionally, the device 400 further includes:

[0195] a detection module configured to detect the location information of the delivery terminal of the delivery person during the process of the delivery person delivering the order to be delivered;

[0196] a prompt module configured to, in response to determining that the delivery person arrives at the target AOI to which the order to be delivered belongs based on the location information, prompt the delivery person through the delivery terminal with the delivery method corresponding to the target AOI.

[0197] Regarding the device in the above embodiments, the specific manners in which each module performs operations have been described in detail in the embodiments related to the method, and will not be elaborated herein.

[0198] Figure 5 is a block diagram of an electronic device 1900 shown according to an exemplary embodiment. For example, the electronic device 1900 may be provided as a server. Referring to Figure 5 , the electronic device 1900 includes a processor 1922, the number of which may be one or more, and a memory 1932 for storing computer programs executable by the processor 1922. The computer programs stored in the memory 1932 may include one or more modules each corresponding to a set of instructions. In addition, the processor 1922 may be configured to execute the computer program to perform the above-described delivery method discrimination method.

[0199] In addition, the electronic device 1900 may further include a power supply component 1926 and a communication component 1950. The power supply component 1926 may be configured to perform power management of the electronic device 1900, and the communication component 1950 may be configured to implement communication of the electronic device 1900, for example, wired or wireless communication. In addition, the electronic device 1900 may further include an input / output (I / O) interface 1958. The electronic device 1900 may operate based on an operating system stored in the memory 1932, such as Windows Server TM , Mac OS X TM , Unix TM , Linux TM and so on.

[0200] In another exemplary embodiment, there is also provided a computer-readable storage medium including program instructions, which when executed by a processor implement the steps of the above-described delivery method discrimination method. For example, the computer-readable storage medium may be the above-mentioned memory 1932 including program instructions, and the above program instructions may be executed by the processor 1922 of the electronic device 1900 to complete the above-described delivery method discrimination method.

[0201] In another exemplary embodiment, a computer program product is also provided, which includes a computer program executable by a programmable device, and the computer program has a code portion for performing the above-described delivery method discrimination method when executed by the programmable device.

[0202] The preferred embodiments of the present disclosure have been described in detail above in conjunction with the accompanying drawings. However, the present disclosure is not limited to the specific details in the above embodiments. Within the scope of the technical concept of the present disclosure, various simple modifications can be made to the technical solutions of the present disclosure, and these simple modifications all fall within the protection scope of the present disclosure.

[0203] In addition, it should be noted that, among the various specific technical features described in the above specific embodiments, they can be combined in any suitable manner without conflict. To avoid unnecessary repetition, the present disclosure does not separately describe various possible combination methods.

[0204] Furthermore, any combination can be made among various different embodiments of the present disclosure, as long as it does not violate the idea of the present disclosure, and it should also be regarded as the content disclosed by the present disclosure.

Claims

1. A method for discriminating delivery modes, characterized in that, The method includes: Obtaining order information of completed deliveries within a target area of interest (AOI), where the order information includes Wi-Fi feature data, AOI feature data, and pedometer feature data of each order within the target AOI; Inputting the Wi-Fi feature data, the AOI feature data, and the pedometer feature data of each order within the target AOI into a delivery method discrimination model to obtain a result output by the delivery method discrimination model for characterizing the delivery method of the target AOI; The delivery method discrimination model is trained through the following method: Obtaining Wi-Fi feature data, AOI feature data, and pedometer feature data of historical orders within at least one AOI, where each historical order corresponds to barometer data; According to the barometer data, adding a label for characterizing the order delivery method to the Wi-Fi feature data, the AOI feature data, and the pedometer feature data of each historical order; Constructing a model training sample based on the Wi-Fi feature data, the AOI feature data, and the pedometer feature data with the label; Training the delivery method discrimination model according to the model training sample; Wherein, constructing the model training sample based on the Wi-Fi feature data, the AOI feature data, and the pedometer feature data with the label includes: Fusing the Wi-Fi feature data, the AOI feature data, and the pedometer feature data with the label within each AOI to obtain an order delivery feature vector for that AOI, and using the order delivery feature vector as the model training sample; The step of inputting the Wi-Fi feature data, the AOI feature data, and the pedometer feature data of each order within the target AOI into the delivery method discrimination model to obtain a result output by the delivery method discrimination model for characterizing the delivery method of the target AOI includes: Generating the order delivery feature vector of the target AOI according to the Wi-Fi feature data, the AOI feature data, and the pedometer feature data of each order within the target AOI, and inputting the order delivery feature vector of the target AOI into the delivery method discrimination model to obtain a result output by the delivery method discrimination model for characterizing the delivery method of the target AOI.

2. The method according to claim 1, wherein The Wi-Fi feature data includes: the average value of the Wi-Fi similarities of each order within the target AOI, where the Wi-Fi similarity of each order is the maximum value among the similarities between the first Wi-Fi data list of the order and each list in the second Wi-Fi data list set; Wherein, the first Wi-Fi data list of each order is scanned by the user terminal when the user corresponding to the order places the order, and correspondingly, the second Wi-Fi data list set of each order is a list set scanned in real time by the delivery terminal when the delivery person corresponding to the order is delivering the order.

3. The method according to claim 2, wherein The calculation formula for the Wi-Fi similarity of each order is as follows: ， Among them, characterizes the Wi-Fi similarity of the order, represents the Wi-Fi data list at the th moment in the set of the second Wi-Fi data lists of the order, represents the first Wi-Fi data list of the order.

4. The method according to claim 2, wherein The Wi-Fi feature data further includes: the average value of the Wi-Fi arrival time of each order within the target AOI, and the median of the Wi-Fi arrival time of each order; wherein, the Wi-Fi arrival time is the difference between the arrival time and the departure time of the delivery terminal determined by the user terminal through Wi-Fi scanning; and / or, The Wi-Fi feature data further includes: the average length value of the arrival / departure Wi-Fi list of the target AOI, and the proportion of the number of sets with a list length change greater than a preset threshold in all the second Wi-Fi data list sets within this AOI; Among them, the average length value of the arrival / departure Wi-Fi list of the target AOI is calculated through the following method: For each order within the target AOI, the mean value of the lengths of all the lists within the second Wi-Fi data list set of this order that are between the arrival time and the departure time is used as the first average value of this order; The average length value of the arrival / departure Wi-Fi list of the target AOI is calculated based on the first average values of all the orders within the target AOI.

5. The method according to claim 4, wherein The pedometer feature data includes: the average value of the number of steps collected by the delivery terminal corresponding to each order within the target AOI between the arrival time and the departure time.

6. The method according to any one of claims 1-5, characterized in that, The AOI feature data includes: the preset quantiles of the delivery floors of each order within the target AOI, the proportion of orders without a delivery floor number among each order, the delivery address type of each order, and the feature words in the delivery address, where the feature words are extracted through a deep learning model.

7. The method according to claim 6, characterized in that, The loss function of the delivery method discrimination model is a cross-entropy loss function: ； Among them, is the value after the order delivery feature vector is mapped by the softmax function, which is used to represent the probability that the AOI corresponding to the order delivery feature vector belongs to the j-th delivery method. The value range of j is from 1 to T, and T is a natural number greater than 1, which is used to represent that there are T delivery methods. represents the label of the model training sample.

8. The method according to claim 1, wherein The method further includes: When generating a new order to be delivered, determining the target AOI to which the order to be delivered belongs according to the delivery address of the order to be delivered; Determining the delivery parameters of the order to be delivered according to the delivery method of the target AOI to which the order to be delivered belongs.

9. The method according to claim 8, wherein The method further includes: During the process of the delivery person delivering the order to be delivered, detecting the location information of the delivery terminal of the delivery person; In response to determining that the delivery person arrives at the target AOI to which the order to be delivered belongs through the location information, prompting the delivery method corresponding to this target AOI to the delivery person through the delivery terminal.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When this program is executed by a processor, it implements the steps of the method described in any one of claims 1-9.

11. An electronic device, characterized in that, It includes: A memory, on which a computer program is stored; A processor, configured to execute the computer program in the memory to implement the steps of the method described in any one of claims 1-9.

Images