Proposal system, proposal method, and computer program

Abstract

To provide a system for proposing a delivery destination on which an intention is reflected for each deliverer.SOLUTION: The proposal system includes a proposal unit configured to perform proposal processing for proposing a next delivery destination to be next delivered by a delivery person from a delivery destination candidate group, a storage unit configured to accumulate past data in which the delivery destination candidate group and a selected delivery destination that is actually selected as the next delivery destination by the delivery person from the delivery destination candidate group are associated with each other, and a learning unit configured to create learning data by using the accumulated past data. A creation unit configured to create a machine learning model that receives a delivery destination candidate group as an input and outputs a selected delivery destination by performing machine learning using learning data, and an acquisition unit configured to acquire the selected delivery destination output by inputting the delivery destination candidate group to the machine learning model, wherein in proposal processing after the machine learning model is created, the proposal unit sets the acquired selected delivery destination as the next delivery destination.SELECTED DRAWING: Figure 1

Description

[Technical Field]

[0001] The present disclosure relates to a suggestion system, a suggestion method, and a computer program. [Background technology]

[0002] There are conventional techniques for efficiently creating a route when visiting multiple destinations in order. For example, in a program disclosed in Patent Document 1, icons of candidate destinations are displayed on a display unit and the program accepts the selection of the next destination. [Prior art documents] [Patent documents]

[0003] [Patent Document 1] JP 2016-44980 A Summary of the Invention [Problem to be solved by the invention]

[0004] Incidentally, when a destination is a delivery destination where a delivery item is to be delivered, a delivery person may take into consideration information other than distance when deciding on the next delivery destination from among multiple candidate destinations. Information other than distance includes, for example, information about the delivery destination, such as the time the resident of the delivery destination is at home, and information about the delivery item, such as the size of the delivery item. Even if multiple candidate delivery destinations are the same, the next delivery destination may vary from delivery person to delivery person. In other words, the method for selecting the next delivery destination from multiple candidate delivery destinations may vary from delivery person to delivery person. Therefore, there is a need for a technology that can suggest a delivery destination that reflects the preferences of each delivery person. [Means for solving the problem]

[0005] The present disclosure can be realized in the following forms.

[0006] (1) According to a first aspect of the present disclosure, a proposal system is provided. The proposal system includes: a proposal unit that performs a proposal process to propose a next delivery destination for a delivery person from a group of candidate delivery destinations; a storage unit that accumulates past data correlating the group of candidate delivery destinations with selected delivery destinations actually selected by the delivery person as the next delivery destination; a creation unit that uses the accumulated past data to create learning data and performs machine learning using the learning data to create a machine learning model that inputs the group of candidate delivery destinations and outputs the selected delivery destinations; and an acquisition unit that inputs the group of candidate delivery destinations into the machine learning model and acquires the output selected delivery destinations. In the proposal process after the machine learning model is created, the proposal unit sets the acquired selected delivery destination as the next delivery destination. According to this aspect, the selected delivery destination output by the machine learning model, which was created using past data containing information on the next delivery destination actually selected by the delivery person, is set as the next delivery destination. Therefore, a next delivery destination that reflects the delivery person's intentions can be proposed. (2) In the proposed system of the above aspect, the learning data may include an identifier that identifies each of the candidate delivery destinations in the group of candidate delivery destinations and delivery destination data associated with the identifier, and the delivery destination data may include at least one of package data related to the package to be delivered and recipient data related to the recipient who will receive the package. According to this aspect, the delivery destination data may include at least one of package data and recipient data. This allows the machine learning model to reflect the actual situation in which a delivery person selects the next delivery destination, taking into account, for example, the size of the package and the recipient's home time. (3) In the proposed system of the above aspect, the storage unit may store the past data of the plurality of delivery personnel, the creation unit may create the learning data using the past data of the plurality of delivery personnel to create the common machine learning model for the plurality of delivery personnel, and the acquisition unit may use the common machine learning model. According to this aspect, a common machine learning model can be created using the past data of the plurality of delivery personnel. Furthermore, the acquisition unit may use the common machine learning model. (4) In the proposal system of the above aspect, the storage unit stores the past data in association with each of the multiple delivery personnel, and each of the multiple delivery personnel is assigned a predetermined attribute. The creation unit may classify the past data using the attributes, create the learning data for each of the classified past data, and perform machine learning to create the machine learning models for the multiple delivery personnel with the same attributes. The acquisition unit may use the machine learning models with the same attributes as the attributes of the target delivery personnel. According to this aspect, a machine learning model can be created for each attribute of the delivery personnel. By using selected delivery destinations output by machine learning models with the same attributes, it is possible to suggest next delivery destinations that are likely to reflect the delivery personnel's intentions. (5) In the proposal system of the above aspect, the proposal process may include a proposal process performed in a first time period before the machine learning model for a target delivery person among the plurality of delivery people is created, and a proposal process performed in a second time period after the machine learning model for the target delivery person is created, and the proposal unit may present the group of delivery destination candidates including the next delivery destination in a selectable manner in the proposal process, and in the proposal process in the first time period, the selected delivery destinations output by the machine learning model for the plurality of delivery people, which was created in advance using the past data for the plurality of delivery people excluding the target delivery person, may be set as the next delivery destination, and in the proposal process in the second time period, the selected delivery destinations output by the machine learning model for the target delivery person, which was created using the past data including the selected delivery destination selected by the target delivery person in the first time period, may be set as the next delivery destination. According to this aspect, the machine learning model created in advance can be used in the first time period, and the machine learning model for the target delivery person can be used in the second time period. (6) In the proposal system of the above aspect, the attribute may represent the length of time the delivery person has been engaged in delivery work. According to this aspect, it is possible to propose, as the next delivery destination, a selected delivery destination output by a machine learning model created using past data of delivery people who have been engaged in delivery work for a similar length of time. Therefore, it is possible to propose a next delivery destination that is likely to reflect the delivery person's intentions. (7) In the proposal system of the above embodiment, the attribute may represent the characteristics of the area where the delivery person performs delivery work. According to this embodiment, a selected delivery destination output by a machine learning model created using past data of delivery people with the same area characteristics can be proposed as the next delivery destination. Therefore, it is possible to propose a next delivery destination that is likely to reflect the delivery person's intentions. (8) In the proposed system of the above aspect, the delivery destination data may further include traffic information data to each of the delivery destinations in the group of delivery destination candidates. According to this aspect, the actual situation in which a delivery person selects the next delivery destination by taking into account the traffic information data can be reflected in the machine learning model. (9) According to a second aspect of the present disclosure, there is provided a computer program that causes a computer to implement the following functions: a function of proposing a next delivery destination for a delivery person to make a next delivery from a group of candidate delivery destinations; a function of transmitting the group of candidate delivery destinations and a selected delivery destination that the delivery person actually selected as the next delivery destination from the group of candidate delivery destinations to a server; a function of performing machine learning using training data created using past data stored on the server in which the candidate delivery destinations and the selected delivery destinations are associated, the machine learning model being created by performing machine learning using training data created using past data stored on the server in which the candidate delivery destinations and the selected delivery destinations are associated, the machine learning model taking the group of candidate delivery destinations as input and outputting the selected delivery destination; a function of receiving from the server the selected delivery destination output by inputting the group of candidate delivery destinations to the machine learning model; and a function of proposing the received selected delivery destination as the next delivery destination. In addition to the proposed system described above, the present disclosure can be realized in the form of a non-transitory tangible recording medium on which a providing method or program is recorded in a computer-readable manner. [Brief explanation of the drawings]

[0007] [Figure 1] FIG. 1 is an explanatory diagram showing a schematic configuration of a proposed system. [Figure 2] FIG. 2 is a diagram illustrating package data and recipient data. [Figure 3] FIG. 10 is a diagram illustrating a proposal screen. [Figure 4] FIG. 1 is a diagram illustrating a general flow of creating a machine learning model. [Figure 5] FIG. 10 is a diagram illustrating learning data. [Figure 6] FIG. 10 is a diagram illustrating an outline of a second embodiment. [Figure 7] FIG. 11 is a diagram illustrating delivery destination data according to the third embodiment. DETAILED DESCRIPTION OF THE INVENTION

[0008] A. First embodiment: A1. System configuration: 1 is an explanatory diagram showing a schematic configuration of a proposed system 10 according to a first embodiment. The proposed system 10 is used, for example, in a delivery service that delivers packages mainly to building entrances using vehicles. The proposed system 10 includes a mobile terminal 100 and a server 200.

[0009] The mobile terminal 100 is used by a delivery person. Typically, the delivery person first loads a plurality of packages to be delivered that day into a vehicle at a collection point. Next, the delivery person moves to each delivery destination in order to deliver the packages. The proposal system 10 has a guidance function that provides guidance on the route to the delivery destination, a proposal function that suggests the next delivery destination after completing a delivery, and the like.

[0010] The mobile terminal 100 is, for example, a smartphone. The mobile terminal 100 can exchange information with a server 200 connected to the Internet 80 via a communication carrier 70 including a transmitting / receiving antenna, a wireless base station, and a switching center. The mobile terminal 100 is configured as a computer including a CPU 110 as a central processing unit, a storage unit 120 such as a ROM or RAM, a touch panel 130, a communication unit 131, and a positioning unit 132. The CPU 110 and other units are connected to each other via internal buses and interface circuits.

[0011] The communication unit 131 is a circuit for performing data communication with the server 200 via the communication carrier 70, wireless LAN communication, etc. The touch panel 130 functions as a display device that displays information and as an input device that accepts input by human operation such as touch. The positioning unit 132 includes a GNSS (Global Navigation Satellite System) receiver, and determines the latitude and longitude of the current location of the mobile terminal 100 based on radio waves received from artificial satellites that make up the GNSS.

[0012] The CPU 110 controls the various components of the mobile terminal 100. The storage unit 120 stores programs for providing the various functions provided in this embodiment. The CPU 110 has functional units such as a data reception unit 111, a proposal unit 112, and a route guidance unit 113. A delivery application program 121 is installed in the mobile terminal 100. The various functional units, such as the data reception unit 111, are realized by executing the delivery application program 121 stored in the storage unit 120.

[0013] The data receiving unit 111 receives package data 122 and recipient data 123, which will be described later, via the touch panel 130, for example, and stores the data in the storage unit 120. The suggestion unit 112 performs a suggestion process of suggesting the next delivery destination for the delivery person to make the next delivery from a group of delivery destination candidates.

[0014] The route guidance unit 113 acquires map data, route data, and guidance data from the server 200 during route guidance, which will be described later. Map data is data representing the shapes of roads and features. Features refer to objects that occupy land, whether natural or artificial, and include real-world entities such as rivers, mountains, plants, bridges, railways, and buildings. The map data, for example, stores information representing the latitude and longitude of features in association with the type of feature. The route guidance unit 113 acquires map data for the area around the current location, for example, in parcels divided into a mesh. Route data is data representing a route from a departure point to a destination. The route data may include data representing a route including intermediate stops on the way to the destination. Guidance data is data including images of guide signs, intersections, interchange entrances, and the like, around the route represented by the route data. The route guidance unit 113 acquires guidance data every time a predetermined distance is traveled along the route. The data acquired by the communication unit 131 is stored in the storage unit 120. The route guidance unit 113 acquires route data, for example, when generating a route or rerouting. Rerouting refers to the process of setting a new route when the current location deviates from the route. Route generation and rerouting are performed by either the server 200 or the mobile terminal 100.

[0015] The route guidance unit 113 controls the touch panel 130 to provide route guidance based on various data received from the server 200 and the current location acquired by the positioning unit 132. Specifically, the route guidance unit 113 acquires map data of the area around the current location from the server 200, displays the map on the touch panel 130, and displays the route represented by the route data acquired from the server 200 on the map.

[0016] The storage unit 120 stores a delivery application program 121, which is a computer program, package data 122, and recipient data 123. In the following description, the CPU 110 that executes the delivery application program 121 may be simply referred to as the "delivery application 121."

[0017] The server 200 includes a CPU 210 as a central processing unit, a storage unit 220, and a communication unit 202. The CPU 210 and other units are interconnected via a bus, an interface circuit, etc. The communication unit 202 is a circuit for communicating with the mobile terminal 100 via the Internet 80.

[0018] The CPU 210 controls all the components of the server 200. The storage unit 220 stores programs for providing the functions provided in this embodiment. The CPU 210 has functional units such as a route generation unit 211, a creation unit 212, and an acquisition unit 213. Each functional unit is realized by executing a program stored in the storage unit 220.

[0019] When the route generation unit 211 receives a route generation request or a reroute request from the mobile terminal 100 via the communication unit 202, it performs route generation processing to generate a route in accordance with the request. When the route generation unit 211 receives a destination specified by the route generation request or reroute request from the mobile terminal 100, it searches for a route to the specified destination using road network data, which will be described later, and generates route data. Algorithms that can be used for route generation include well-known Dijkstra algorithms and A-Star algorithms. The route generation unit 211 transmits the generated route data, together with map data and guidance data, to the mobile terminal 100 via the communication unit 202.

[0020] The storage unit 220 is, for example, a RAM, a ROM, or a hard disk drive (HDD). The readable and writable area of ​​the storage unit 220 stores map data, guidance data, road network data, past data 223, delivery destination data 224, candidate delivery destination data 225, learning data 226, and a machine learning model 227.

[0021] Road network data is data that represents the state of road connections using link data that represent roads and node data that represent intersections, bends, and dead ends. The node data includes information that represents latitude and longitude. The link data includes information that indicates the nodes that make up the link data and cost information that represents the travel time required to pass through the road.

[0022] The past data 223 is data that associates a group of delivery destination candidates, which will be described later, with a selected delivery destination that a delivery person actually selected as the next delivery destination from the group of delivery destination candidates.

[0023] A2. Proposed function description: 2 is a diagram illustrating the package data 122 and the recipient data 123. The proposal function of the proposal system 10 uses the package data 122 and the recipient data 123 shown in FIG.

[0024] The package data 122 is data related to packages to be delivered. The package data 122 is typically created by a delivery person using the mobile terminal 100 when loading the packages into a vehicle. The package data 122 is input by specifying each of a plurality of packages. The contents of the package data 122 include, for example, the "delivery address," "designated delivery time period," "size," "weight," "fragile," and "perishable."

[0025] The data receiving unit 111 displays an input screen on the touch panel 130 and receives input of the package data 122. The data receiving unit 111 stores the input package data 122 in the storage unit 120, as will be described later.

[0026] The parcel data 122 includes items for which input options are pre-prepared on the input screen displayed by the data receiving unit 111. Items for which options are pre-prepared include "designated delivery time slot" and "size." For "designated delivery time slot," options for pre-defined time slots such as "morning," "12:00 to 14:00," and "14:00 to 16:00" are provided, as well as options such as "none," which indicates no designation. For "size," options such as "large," "medium," and "small" are provided. "Fragile" and "perishable" are options in themselves. The delivery person touches the appropriate option displayed on the input screen to select it. The data receiving unit 111 accepts the selected option as parcel data 122.

[0027] In order to accept input for items of the parcel data 122 for which no input options are prepared in advance, the input screen is provided with an input field for accepting input of characters, etc. The delivery person can freely enter text, etc. in the input field. Examples of text that can be entered as the parcel data 122 include "heavy," "light," "early delivery," "leave at delivery OK," and "fragile." The data accepting unit 111 accepts the information entered in the input field as the parcel data 122.

[0028] The data acceptance unit 111 has a function of accepting input of the "delivery address" by capturing an image of the delivery slip attached to the package using a camera (not shown) provided in the mobile terminal 100. The data acceptance unit 111 performs image analysis on the captured image of the delivery slip to acquire and accept the "delivery address." The data acceptance unit 111 can also accept the "delivery address" entered on the "delivery address" input screen.

[0029] The data receiving unit 111 associates the received package data 122 with an identifier uniquely assigned to the package and stores it in the storage unit 120 as package data 122. Specifically, when an option is selected and input, the data receiving unit 111 stores the information of the option in association with the item. When input into an input field, the data receiving unit 111 stores the input data as "other information." For example, a "slip number" can be used as the identifier assigned to the package data 122. The package data 122 is typically data associated with each of multiple packages scheduled to be delivered on that day. In other words, the multiple delivery destinations corresponding to these multiple packages are a group of candidate delivery destinations for delivery on that day.

[0030] The recipient data 123 is data related to the recipient who will receive the package. The recipient data 123 is typically created by the delivery person using the mobile terminal 100 while they are performing their delivery duties. The data accepting unit 111 displays an input screen on the touch panel 130 to accept input of the recipient data 123. The contents of the recipient data 123 include, for example, "often absent," "at home after 7 p.m.", "has a delivery box," and "has received delivery requests in the past." Note that, as with the package data 122, the recipient data 123 can also be entered using options prepared in advance.

[0031] The input field for receiving input of the package data 122 and the input field for receiving input of the recipient data 123 may be the same or different from each other.

[0032] The data receiving unit 111 associates the input recipient data 123 with the "delivery address" and stores it in the storage unit 120 as recipient data 123. Note that the information associated with the recipient data 123 is not limited to the "delivery address" and may be any information that can identify the delivery destination. For example, instead of the "delivery address", a symbol uniquely assigned to the delivery address may be used.

[0033] The contents of the package data 122 and recipient data 123 accepted by the data accepting unit 111 are not limited to those described above. There are no restrictions on the content of the text that the delivery person enters in the input field, and the delivery person can freely enter content in the input field on the input screen. Also, among the package data 122 and recipient data 123 entered on the input screen, there are cases where data other than the "delivery address" is not entered.

[0034] The delivery person can input the package data 122 and the recipient data 123 on the input screen that the data reception unit 111 displays on the touch panel 130 at any time, not just when loading the package into the vehicle.

[0035] The data receiving unit 111 transmits the created package data 122 and recipient data 123 to the server 200. The server 200 stores the received package data 122 and recipient data 123 in the storage unit 220. Upon receiving the package data 122 and recipient data 123, the server 200 stores the package data 122 and recipient data 123 as candidate delivery destination data 225. The candidate delivery destination data 225 is data in which information included in the package data 122 and recipient data 123 is associated with an identifier that is each delivery destination in the group of candidate delivery destinations.

[0036] It should be noted that the person who inputs the package data 122 is not limited to a delivery person. The package data 122 may be input, for example, by an administrator using a PC (Personal Computer) that can communicate with the server 200. In addition, information written on a delivery slip may be input during the delivery process, for example, from when the package is received from the sender to when it is delivered. In such a case, the server 200 may store the information input during the delivery process in the storage unit 220. When the proposed system 10 is used by a corporation, the package data 122 and the recipient data 123 may be shared among delivery people belonging to the same corporation.

[0037] In this embodiment, the data receiving unit 111 distinguishes between the package data 122 and the recipient data 123 before transmitting them to the server 200. In another embodiment, the data receiving unit 111 may transmit the package data 122 and the recipient data 123 to the server 200 without distinguishing between them. Specifically, the package data 122 and the recipient data 123 may be entered in input fields on an input screen. In this case, the content received in the input field may be transmitted to the server 200 as package data 122, for example, without distinguishing between content related to the package and content related to the recipient.

[0038] The delivery person selects the first delivery destination from the multiple delivery destination addresses included in the package data 122, i.e., the group of delivery destination candidates. The delivery application 121 displays the group of delivery destination candidates in a list on the touch panel 130. In detail, the delivery application 121 requests the server 200 to provide delivery destination candidate data 225, and uses the received delivery destination candidate data 225 to display the list of delivery destination candidate groups on the touch panel 130. For example, the delivery application 121 displays a map on the touch panel 130, and displays candidate pins IM102 (described below) indicating the delivery destination candidates on the map. The delivery person, for example, touches and selects the candidate pin IM102 corresponding to the desired delivery destination from the multiple pins displayed.

[0039] The route guidance unit 113 makes a route generation request to the server 200, with the delivery destination corresponding to the selected candidate pin IM102 as the destination. The route generation unit 211 of the server 200 performs a route generation process to create route data, and transmits the created route data together with map data and guidance data to the mobile terminal 100. The route guidance unit 113 of the mobile terminal 100 uses the received route data and the like to display on a map a travel route from the current location to the delivery destination corresponding to the selected pin.

[0040] When the delivery person completes the first delivery step, the delivery person uses the mobile terminal 100 to confirm the "delivery status" of the delivery destination for which the delivery step has been completed. In detail, the delivery application 121 displays a confirmation screen on the touch panel 130 for confirming the "delivery status" of the delivery destination. The confirmation screen displays selection buttons such as "Delivery" and "Absent." When the delivery person has completed delivery to the delivery destination, the delivery person touches and selects the "Delivery" button. On the other hand, when the delivery person was absent and was unable to deliver the package, the delivery person touches and selects the "Absent" button. The delivery application 121 transmits to the server 200 information on the "delivery status" of the delivery destination confirmed by the delivery person.

[0041] When the server 200 receives the "delivery status" information, it updates the delivery destination candidate data 225. Specifically, the delivery destination candidate data 225 has a "delivery status" field that is associated with an identifier. When the server 200 creates the delivery destination candidate data 225, it sets this "delivery status" to "undelivered." Then, when the server 200 receives the "delivery status" information, it updates the value of the "delivery status" to the received information, specifically, to "delivered" or "absent."

[0042] When the "Delivery" button or the "Absent" button is selected, the suggestion unit 112 displays the suggestion screen SC1 shown in Fig. 3. In detail, the delivery application 121 requests the server 200 to provide information on a group of candidate delivery destinations, including the next delivery destination. Upon receiving the request, the server 200 transmits the latest candidate delivery destination data 225 to the mobile terminal 100. The suggestion unit 112 uses the received candidate delivery destination data 225 to display the suggestion screen SC1 on the touch panel 130.

[0043] FIG. 3 is a diagram illustrating the proposal screen SC1. As shown in FIG. 3, the proposal screen SC1 displays a map image IM101, a candidate pin IM102, a current location pointer IM103, a candidate information image IM104, and a selection button IM105. The candidate pin IM102 indicates a delivery destination that the server 200 has extracted from the group of delivery destination candidates as a suitable next delivery destination for the delivery person. In other words, the candidate pin IM102 indicates a next candidate destination proposed by the proposal system 10. The process by which the proposal unit 112 displays the proposal screen SC1 on the touch panel 130 is also referred to as a proposal process. The method for extracting the next delivery destination candidate corresponding to the candidate pin IM102 will be described later.

[0044] A candidate pin IM102 and a current location pointer IM103 are displayed superimposed on a map image IM101. The candidate pin IM102 is a pin that indicates the position on the map of the next delivery destination candidate group. The candidate pin IM102 is displayed in a selectable manner. The current location pointer IM103 is a pointer that indicates the position of the mobile terminal 100, i.e., the current location of the delivery person. The candidate information image IM104 displays information about the delivery destinations, categorized by delivery destination. Specifically, information such as the delivery destination candidate data 225 of the delivery destination associated with the candidate pin IM102 is displayed as text information. The candidate information image IM104 is displayed in a scrolling manner. Note that the delivery destination candidate destinations displayed in the candidate information image IM104 may be limited to a limited number of delivery destination candidate destinations that correspond to the candidate pin IM102, or all delivery destination candidate destinations for which delivery has not yet been completed may be displayed. The delivery destination candidate destinations displayed in the candidate information image IM104 are displayed in a selectable manner.

[0045] The delivery person selects the next delivery destination by touching and selecting the candidate pin IM102 or the candidate information image IM104, and then touches and selects the selection button IM105. When the selection button IM105 is selected, the route guidance unit 113 displays route information to the selected delivery destination, which is the selected next delivery destination, in the same way as in the case of the first delivery destination.

[0046] In the initial setting, the server 200 extracts the top N (N is a natural number) delivery destinations that are closest in a straight line distance from the current location from among the group of delivery destination candidates, from among packages whose designated delivery time slots include the time slot that includes the current time, and packages whose designated delivery time slots are not specified. That is, in the initial setting, the server 200 extracts delivery destination candidates using the distance from the current location to the delivery destination and the current time. The server 200 then assigns numbers to the delivery destination candidates in ascending order of distance, with the closest delivery destination being assigned a number of "1." The server 200 then includes information on the extracted N delivery destinations in the delivery destination candidate data 225 that it sends to the mobile terminal 100.

[0047] The suggestion unit 112 displays N candidate pins IM102 corresponding to the N delivery destinations extracted by the server 200 on the touch panel 130. Fig. 3 shows a suggestion screen SC1 that is displayed when N is "5".

[0048] Here, the delivery person may not select the next delivery destination as suggested by the suggestion unit 112. There are various reasons why the delivery person may not select the next delivery destination as suggested. For example, the current time may be outside the "home hours" displayed in the candidate information image IM104, and the delivery person may determine that the recipient is likely not at home and remove the destination from the list of next delivery destination candidates. Also, there may be cases where the delivery person wants to prioritize delivery to a delivery destination with a parcel locker. Furthermore, if the package is heavy, the delivery person may want to prioritize delivery of the heavy package to reduce the weight of the vehicle. Furthermore, if the delivery destination is a recipient who has "received delivery requests in the past," the delivery person may want to deliver the package earlier in the specified time slot. Even in the same situation, the reasons for priority may differ depending on the delivery person. Therefore, in this embodiment, machine learning is performed using past data of delivery destinations that the delivery person actually selected as the next delivery destination from the group of delivery destination candidates, thereby making suggestions that reflect the delivery person's intentions.

[0049] A3. Machine learning model creation and proposal processing: FIG. 4 is a diagram showing a general flow of creating the machine learning model 227. Step S1 is a step in which the next delivery destination is proposed using the initial settings as described above, and past data 223 for machine learning is accumulated. In step S1, past data 223 that associates a group of delivery destination candidates with selected delivery destinations that the delivery person actually selected as the next delivery destination from the group of delivery destination candidates is accumulated in the storage unit 220. Step S1 is performed for a predetermined period. The predetermined period is, for example, from several weeks to about a year. The period in which step S1 is performed is preferably varied depending on the amount of data collected in a unit period.

[0050] Step S2 is a step in which machine learning is performed using learning data 226 created using past data 223 accumulated in step S1, thereby creating a machine learning model 227. Step S3 is a step in which a proposal process is performed using the machine learning model 227 created in step S2, instead of initial settings. This makes it possible to propose next delivery destination candidates for each delivery person that reflect the delivery person's preferences. Steps S2 and S3 are described in detail below.

[0051] In step S2, first, the creation unit 212 creates learning data 226 for machine learning. Then, the creation unit 212 performs machine learning using the created learning data 226 to create a machine learning model 227 that inputs a group of delivery destination candidates and outputs selected delivery destinations. For example, supervised learning or reinforcement learning can be used as machine learning.

[0052] FIG. 5 is a diagram illustrating the training data 226. As shown in FIG. 5, the training data 226 includes multiple pieces of past data 223 corresponding to a single selection of a delivery destination by a delivery person. The past data 223 includes a group of delivery candidate destinations and a selection identifier that identifies the selected delivery destination selected by the delivery person from the group of delivery candidate destinations as the next delivery destination. The selection identifier is an identifier assigned to the package corresponding to the selected delivery destination. Each delivery candidate destination in the group of delivery candidate destinations is data that pairs the identifier assigned to the package with delivery destination data 224. The delivery destination data 224 includes status data, package data 122, and recipient data 123.

[0053] The creation unit 212 creates the past data 223 using the delivery destination candidate data 225, the delivery destination data 224, and information about the next delivery destination each time the delivery person selects a next delivery destination. The creation unit 212 can identify the next delivery destination selected by the delivery person, for example, when the server 200 receives a route generation request or when the server 200 receives "delivery status" information. Note that the timing when the creation unit 212 creates the past data 223 is not limited to when the delivery person selects the next delivery destination. For example, information about the delivery destination candidate group and the selected delivery destination associated with the delivery destination candidate group may be stored in the mobile terminal 100 or the server 200, and the past data 223 may be created when the machine learning model 227 is created, for example.

[0054] The status data items include "current location" and "current date and time." "Current location" is the location of the delivery person and the mobile terminal 100 at the time when the delivery person selected the corresponding delivery destination. "Current date and time" is the date and time when the delivery person selected the corresponding delivery destination.

[0055] Items in the package data 122 include "size," "weight," "designated delivery date and time," "fragile," and "perishable." The creation unit 212 converts the content of the package data 122 into content suitable for machine learning, i.e., quantifies it, to create the training data 226. When the package data 122 is input by selecting an option, the creation unit 212 creates the training data 226 using numerical values ​​previously associated with the option. As described above, when a delivery person inputs the package data 122, an input field may be used. Since there are no rules for the content to be input into the input field, the sentences input by the delivery person vary. Therefore, the creation unit 212 uses the input sentences to create data suitable for machine learning. For example, when a delivery person inputs the sentences "heavy" or "light" regarding "weight," the creation unit 212 quantifies the degree of "weight" and enters the value of the "weight" item.

[0056] For package data 122, when options are prepared in advance for an item, the delivery person typically selects an option, but there are also cases where the delivery person enters the same content in the input field without selecting an option. Therefore, the creation unit 212 creates learning data 226 using the input package data 122, regardless of whether it was selected from the options or entered in the input field.

[0057] Items of the recipient data 123 include "address," "hours at home," and "delivery box." "Address" is the address of the delivery destination. The value of "delivery box" is entered as information on whether or not the building at the delivery destination has a delivery box. The creation unit 212 identifies the delivery address associated with the package identifier and creates learning data 226 using the recipient data 123 stored in the memory unit 220.

[0058] There are cases where the input field for accepting input of package data 122 and the input field for accepting input of recipient data 123 are the same. There are also cases where a delivery person mistakenly enters recipient data 123 into the input field for accepting input of package data 122. Therefore, regardless of the form of input accepted by the data accepting unit 111 of the mobile terminal 100 as package data 122 and recipient data 123, the creation unit 212 uses delivery destination candidate data 225 to create learning data 226. Machine learning may be used to realize the functions of the creation unit 212.

[0059] For convenience of explanation, in this embodiment, the same symbols are assigned to the parcel data 122 and recipient data 123 stored in the mobile terminal 100 and the parcel data 122 and recipient data 123 in the learning data 226, respectively, but as mentioned above, there may be cases where the contents do not match.

[0060] When supervised learning is used as the machine learning, the selection identifier is used as the correct answer data in the task of selecting one delivery destination from a group of candidate delivery destinations. When reinforcement learning is used as the machine learning, whether the selected delivery destination is the same as the selection identifier is used as the reward in the task of selecting one delivery destination from a group of candidate delivery destinations. Specifically, for example, if the identifier is the same as the selection identifier, a positive reward is given, and if the identifier is different from the selection identifier, a negative reward is given. Note that the machine learning method is not limited to the above, and other machine learning methods, such as deep learning, or a combination of multiple machine learning methods may also be used.

[0061] In step S3 of FIG. 4, when the delivery person selects the "Delivery" button or the "Not at Home" button displayed on the touch panel 130 after completing delivery, the suggestion unit 112 requests the server 200 to extract next candidate destinations. When the server 200 receives this request, the creation unit 212 creates a group of candidate delivery destinations to be input into the machine learning model 227. The group of candidate delivery destinations is data similar to the past data 223 shown in FIG. 5. Specifically, the group of candidate delivery destinations is data in which "identifiers" are associated with "delivery destination data" for each candidate delivery destination. The current location and date and time of the mobile terminal 100 that sent the request are input into the "situation data." Note that the timing when the creation unit 212 creates the group of candidate delivery destinations is not limited to when the mobile terminal 100 requests the extraction of next candidate destinations; the creation unit 212 may have created part of the group of candidate delivery destinations before the request is made.

[0062] The acquisition unit 213 inputs the group of delivery destination candidates created by the creation unit 212 into the machine learning model 227 created in step S2 and acquires the selected delivery destination output. The acquisition unit 213 transmits the acquired selected delivery destination to the mobile terminal 100. In the proposal process, the proposal unit 112 sets the received selected delivery destination as the next delivery destination. That is, the proposal unit 112 displays the received selected delivery destination as the next delivery destination on the proposal screen SC1 and proposes it as the next delivery destination. Note that the method of proposing the next delivery destination is the same as above, and a candidate pin IM102 indicating the next delivery destination is displayed, and information about the next delivery destination is displayed in the candidate information image IM104 to propose it.

[0063] The proposal process may propose multiple next delivery destinations or only one. When multiple next delivery destinations are proposed in the proposal process, for example, the acquisition unit 213 configures the machine learning model 227 to output multiple selected delivery destinations and acquires the multiple selected delivery destinations. Specifically, when acquiring the output of the machine learning model 227, the acquisition unit 213 configures the multiple selected delivery destinations to be output in descending order of evaluation value. The evaluation value is an index that indicates a higher value the closer to the "correct answer." In the case of deep learning, it refers to an error calculated using, for example, a loss function. When proposing multiple next delivery destinations, the acquisition unit 213 may, for example, set one selected delivery destination output by the machine learning model 227 as the top next delivery destination, and set the second and subsequent next delivery destinations to next delivery destinations extracted by initial setting. When only one next delivery destination is proposed, the one selected delivery destination output by the machine learning model 227 is set as the next delivery destination.

[0064] The proposal process performed by the proposal unit 112 is also referred to as a proposal step. The processing step in which the memory unit 220 accumulates the past data 223 is also referred to as a storage step. Step S2 is also referred to as a creation step. The processing step in which the acquisition unit 213 inputs a group of delivery destination candidates into the machine learning model 227 to acquire the selected delivery destination that is output is also referred to as an acquisition step.

[0065] According to the first embodiment described above, the proposal system 10 includes a proposal unit 112, a storage unit 220, a creation unit 212, and an acquisition unit 213. The creation unit 212 creates learning data 226 using accumulated past data 223. The creation unit 212 then performs machine learning using the learning data 226 to create a machine learning model 227 that inputs a group of delivery destination candidates and outputs a selected delivery destination. In the proposal process after the machine learning model 227 is created, the proposal unit 112 sets the selected delivery destination acquired by the acquisition unit 213 as the next delivery destination. As a result, the selected delivery destination output by the machine learning model 227, which is created using the past data 223 of next delivery destinations actually selected by the delivery person, is set as the next delivery destination. Therefore, the proposal system 10 can propose a next delivery destination that reflects the delivery person's intentions.

[0066] The learning data 226 also includes an identifier that identifies each delivery destination candidate in the group of delivery destination candidates, and delivery destination data 224 associated with the identifier. The delivery destination data 224 includes package data 122 related to the package to be delivered, and recipient data 123 related to the recipient who will receive the package. This allows the machine learning model 227 to reflect the actual situation in which a delivery person selects the next delivery destination, taking into consideration, for example, the size of the package and the recipient's home time.

[0067] B. Second embodiment: FIG. 6 is a diagram illustrating an overview of the second embodiment. In the first embodiment, in step S1 shown in FIG. 4 for storing past data 223, the next delivery destination is extracted from a group of delivery destination candidates using the initial settings of date, time, and distance. In contrast, in step S1 of this embodiment, a machine learning model 227 created in advance is used. As a result, even in step S1, the proposal system 10 can propose a next delivery destination that is close to the delivery person's intention. The same configurations and processing steps as in the first embodiment are assigned the same reference numerals, and detailed explanations will be omitted as appropriate.

[0068] The proposal system 10 of this embodiment is used, for example, by multiple delivery personnel belonging to the same corporation that performs delivery operations. In step S2, before one of the delivery personnel for whom the machine learning model 227 is to be created engages in delivery operations, the machine learning model 227 is created using past data 223 of other delivery personnel excluding the target delivery personnel. Then, as shown in FIG. 6 , immediately after the target delivery personnel begins using the proposal system 10, a proposal process for the target delivery personnel is performed using the created machine learning model 227 created using past data 223 of multiple delivery personnel excluding the target delivery personnel. During this period, the past data 223 of the target delivery personnel is accumulated. This period is referred to as a first period.

[0069] In step S2, a machine learning model 227 of the target delivery person is created using past data 223 of the target delivery person. Then, in the proposal process after the machine learning model 227 of the target delivery person is created, the proposal process is performed using the created machine learning model 227 of the target delivery person. This period is referred to as the second period. Note that the accumulation of past data 223 continues during this second period. Then, at each predetermined period, the machine learning model 227 is created using past data 223 with an increased amount of data compared to when the initial machine learning model 227 of the target delivery person was created, and the machine learning model 227 stored in the storage unit 220 is overwritten with the latest machine learning model 227. As a result, as the target delivery person continues to use the proposal system 10, it is possible to propose next delivery candidates with greater accuracy that more accurately reflect the intentions of the target delivery person.

[0070] In the above example, the machine learning model 227 used in the first period is a common machine learning model 227 that uses the past data 223 of multiple delivery personnel excluding the target delivery personnel. In another embodiment, each delivery personnel is assigned a predetermined attribute, and the creation unit 212 may classify the past data 223 using the attribute, create learning data 226 for each of the past data 223 for the multiple classified delivery personnel, and perform machine learning to create a machine learning model 227 for each attribute. Then, the suggestion unit 112 may use a machine learning model 227 with the same attribute as the target delivery personnel in the proposal process for the first period.

[0071] Attributes can include the length of time a delivery person has been engaged in delivery work and the characteristics of the area where the delivery person works. Specifically, predetermined categories are used for the "length of time engaged in delivery work," such as "less than three years" for beginners and "more than three years" for experts. When the attribute is the length of time engaged in delivery work, the criteria for selecting the next delivery destination may differ between beginners and experts. Furthermore, beginners and experts may have similar criteria for selecting the next delivery destination from a group of candidate delivery destinations. Specifically, predetermined categories are used for the "area characteristics," such as "urban" with dense residential areas and "rural" with sparse residential areas. Because the average distance to the next delivery destination and traffic conditions differ between "urban" and "rural," the criteria for selecting the next delivery destination may differ. Therefore, by creating a machine learning model 227 for each attribute and creating a machine learning model 227 for the attributes of the target delivery person, the proposal system 10 can suggest a next delivery destination that is close to the delivery person's intentions.

[0072] The proposed system 10 can acquire attributes of delivery personnel by, for example, having the delivery personnel or other personnel input the attributes. In addition, if the attribute is "length of time engaged in delivery work," the proposed system 10 can calculate the "length of time engaged in delivery work" from the total usage time that the delivery personnel uses the delivery application 121 without relying on human input.

[0073] In the above example, the suggestion unit 112 uses a machine learning model 227 with the same attributes as the attributes of the target delivery person in the suggestion process for the first period. In another embodiment, if the attribute is "length of time engaged in delivery work," the machine learning model 227 created using past data 223 of an expert may be used for a novice delivery person. An expert may select the next delivery destination to create a more efficient travel route. Therefore, by using the machine learning model 227 of an expert for a novice delivery person, the novice can view the next delivery destination suggested in the suggestion process and perform a more efficient delivery work.

[0074] According to the second embodiment described above, the storage unit 220 accumulates past data 223 of multiple delivery personnel. The creation unit 212 creates learning data 226 using the past data 223 of the multiple delivery personnel, and creates a machine learning model 227 for the multiple delivery personnel. This allows the creation unit 212 to create the machine learning model 227 using the past data 223 of the multiple delivery personnel. By using the machine learning model 227 created in advance in the proposal process for the first period, it is possible to propose a next delivery destination that is likely to reflect the intentions of the delivery personnel.

[0075] C. Third embodiment: 7 is a diagram illustrating the delivery destination data 224 according to the third embodiment. As shown in FIG. 7, the delivery destination data 224 according to this embodiment includes "traffic information data" in addition to "status data," "package data," and "recipient data." The "traffic information data" is data indicating traffic information to a candidate delivery destination at the time of delivery.

[0076] In the proposal process of this embodiment, traffic information for the candidate delivery destinations is also presented to the delivery person. There are various methods for presenting traffic information. Examples of presentation methods include presenting information on whether there are traffic jams, accidents, or road closures along the travel route, and presenting a predicted arrival time. Traffic information may be displayed on the proposal screen SC1 or on a route guidance screen. The delivery person selects the next delivery destination based on the displayed traffic information. Therefore, in this embodiment, traffic information data is included in the delivery destination data 224, thereby creating a machine learning model 227 suited to the delivery person. The traffic information data is created by the creation unit 212, similar to the parcel data 122. For example, if a predicted installation time is presented, the creation unit 212 quantifies the predicted travel time. For example, if the presence or absence of traffic congestion is presented, the creation unit 212 quantifies the presence or absence of traffic congestion. In this case, the creation unit 212 may quantify the degree of traffic congestion, specifically, the vehicle's progress or the length of the congested road.

[0077] According to the third embodiment described above, the delivery destination data 224 further includes traffic information data to each delivery destination in the group of delivery destination candidates. This allows the machine learning model 227 to reflect the actual situation in which the delivery person selects the next delivery destination taking into account the traffic information data.

[0078] D. Other Embodiments: (D1) In the third embodiment, a case was described in which traffic information is presented when the delivery person selects the next delivery destination. In another embodiment, if traffic information data is not included in the past data 223, current traffic information may be reflected in the proposal process of step S3. For example, the machine learning model 227 outputs N delivery destination candidates. If there is current congestion on the travel route of each of the output N delivery destination candidates, the current traffic information can be reflected by performing a process to lower the ranking. Alternatively, the estimated travel time for each delivery destination is calculated, and for delivery destinations with extremely long estimated travel times, the current traffic information can be reflected by performing a process to lower the ranking.

[0079] (D2) In the first embodiment described above, initial settings are applied to the proposal process in step S1 shown in FIG. 4, and the created machine learning model 227 is used in the proposal process in step S3. As another embodiment, in the proposal process in step S3, both a proposal based on the initial settings and a proposal based on the machine learning model 227 may be made. In this embodiment, the proposal screen shown in FIG. 3 includes (first display form), (second display form), and (third display form). In (first display form), either a "proposal screen based on initial settings" or a "proposal screen using the machine learning model 227" is displayed in a switchable manner. For each of the "proposal screen based on initial settings" and the "proposal screen using the machine learning model 227," a label indicating which setting is being used is displayed on the screen. For example, the words "distance extraction" are used as a label indicating the "proposal screen based on initial settings." For example, the words "machine learning" are used as a label indicating the "proposal screen using the machine learning model 227." Instead of characters, the label may be a display format, such as the background color of the screen, or both characters and a display format may be used as the label.

[0080] In (D2: second display form), the next candidate destination extracted by the initial setting and the next candidate destination output by the machine learning model 227 are displayed together on the same display screen of the touch panel 130. In this case, each next candidate destination is displayed with a label similar to that described above so that it is clear whether the initial setting or the machine learning model 227 has been used. The next candidate destination extracted by the initial setting and the next candidate destination output by the machine learning model 227 may be displayed in different display areas, or both may be displayed in a list in the same display area. Next candidate destinations that overlap between the next candidate destination extracted by the initial setting and the next candidate destination extracted by the machine learning model 227 are displayed with a label indicating that they overlap. For example, either "distance extraction" or "machine learning" may be displayed as this label, or instead of these two, a word such as "overlap" may be displayed.

[0081] In (D2: third display form), the first proposal screen displayed that day is a proposal screen to which initial settings are applied, and from the second time onwards, either a "proposal screen with initial settings" or a "proposal screen using machine learning model 227" is displayed depending on the input operation performed on touch panel 130, for example. In other words, from the second time onwards, the proposal screen may be a "proposal screen with initial settings" or may be switched to a "proposal screen using machine learning model 227". This embodiment (D2) may be applied to the second and third embodiments.

[0082] (D3) In the first embodiment, the learning data 226 includes both the package data 122 and the recipient data 123, but it is not necessary to include either the package data 122 or the recipient data 123. By including at least either the package data 122 or the recipient data 123, the learning data 226 can propose next delivery destination candidates that reflect the intentions of the delivery person.

[0083] (D4) In the third embodiment, the delivery destination data 224 includes traffic information data. The delivery destination data 224 may also include other information, such as the weather or the delivery person's physical condition on that day, as situation data. By including many types of information in the delivery destination data 224, the accuracy of the next delivery destination suggestion can be improved.

[0084] (D5) In the first embodiment, when accepting the package data 122, characters are entered into the input field on the input screen. As another example, an image may be entered into the input field. Then, the package data 122 may be entered by performing image analysis on the entered image.

[0085] (D6) In the third embodiment, the attribute used is, for example, "length of time engaged in delivery work." Other attributes may include, for example, the delivery person's work style, age, or gender. An example of a work style is working only in the morning, or working three days a week. Delivery people with the same attributes tend to have similar criteria for selecting the next delivery destination. Therefore, it is advisable to create a machine learning model 227 for each attribute with similar selection criteria. By using the machine learning model 227 for each attribute during a period when sufficient past data 223 cannot be accumulated to create a useful machine learning model 227, it is possible to suggest a next delivery destination that is likely to reflect the delivery person's intentions.

[0086] The present disclosure is not limited to the above-described embodiments and can be realized in various configurations without departing from the spirit thereof. For example, the technical features of the embodiments corresponding to the technical features in each aspect described in the Summary of the Invention section can be appropriately replaced or combined to solve some or all of the above-described problems or achieve some or all of the above-described effects. Furthermore, if a technical feature is not described as essential in this specification, it can be appropriately deleted. [Explanation of symbols]

[0087] 10...Proposal system, 70...Communication carrier, 80...Internet, 100...Mobile terminal, 110...CPU, 111...Data acceptance unit, 112...Proposal unit, 113...Route guidance unit, 120...Memory unit, 121...Delivery application program, 122...Parcel data, 123...Recipient data, 130...Touch panel, 131...Communication unit, 132...Positioning unit, 200...Server, 202...Communication unit, 210...CPU, 211...Route generation unit, 212...Creation unit, 213...Acquisition unit, 220...Memory unit, 223...Past data, 224...Delivery destination data, 225...Delivery destination candidate data, 226...Learning data, 227...Machine learning model, SC1...Proposal screen

Claims

1. A proposed system, a proposal unit that performs a proposal process to propose a next delivery destination for the delivery person from the group of delivery destination candidates; a storage unit that stores past data that associates the group of delivery destination candidates with selected delivery destinations that the delivery person actually selected as the next delivery destination from the group of delivery destination candidates; a creation unit that creates learning data using the accumulated past data and performs machine learning using the learning data to create a machine learning model that takes the group of delivery destination candidates as input and outputs the selected delivery destination; an acquisition unit that inputs the group of delivery destination candidates into the machine learning model and acquires the selected delivery destinations output, A proposal system in which, in the proposal process after the machine learning model is created, the proposal unit sets the acquired selected delivery destination as the next delivery destination.

2. The proposal system according to claim 1, the learning data includes an identifier for identifying each of the delivery destination candidates in the group of delivery destination candidates, and delivery destination data associated with the identifier; The delivery destination data includes at least one of package data relating to the package to be delivered and recipient data relating to the recipient who will receive the package.

3. The proposal system according to claim 1, The storage unit accumulates the past data of the plurality of delivery personnel, the creation unit creates the learning data using the past data for the plurality of delivery personnel to create the common machine learning model for the plurality of delivery personnel; A proposal system in which the acquisition unit uses a common machine learning model.

4. The proposal system according to claim 1, The storage unit stores the past data in association with each of the plurality of delivery persons, Each of the plurality of delivery persons is assigned a predetermined attribute, the creation unit classifies the past data using the attributes, creates the learning data for each of the classified past data, and performs machine learning to create the machine learning model for the plurality of delivery persons having the same attributes; A proposal system in which the acquisition unit uses the machine learning model with the same attributes as the attributes of the target delivery person.

5. 5. The proposal system according to claim 3 or 4, The proposal process includes a proposal process that is performed in a first period before the machine learning model of a target delivery person among the plurality of delivery people is created, and a proposal process that is performed in a second period after the machine learning model of the target delivery person is created, the proposal unit presents the group of delivery destination candidates including the next delivery destination in a selectable manner in the proposal processing; In the proposal process for the first period, the selected delivery destination output by the machine learning model for the plurality of delivery members, which was created in advance using the past data for the plurality of delivery members excluding the target delivery member, is set as the next delivery destination; In the proposal process for the second period, the selected delivery destination output by the machine learning model for the target delivery person, which was created using past data including the selected delivery destination selected by the target delivery person in the first period, is set as the next delivery destination, a proposal system.

6. 5. The proposal system according to claim 4, A proposal system, wherein the attribute represents the length of time the delivery person has been engaged in delivery work.

7. 5. The proposal system according to claim 4, A proposal system in which the attributes represent the characteristics of the area in which the delivery person performs delivery operations.

8. The proposal system according to claim 2, The delivery destination data further includes traffic information data to each of the delivery destinations in the group of delivery destination candidates.

9. A proposed method, a proposal step of proposing a next delivery destination for the delivery person from the group of delivery destination candidates; a storage step of storing past data that associates the group of delivery destination candidates with selected delivery destinations that the delivery person actually selected as the next delivery destination from the group of delivery destination candidates; a creation step of creating learning data using the accumulated past data, and performing machine learning using the learning data to create a machine learning model that inputs the group of delivery destination candidates and outputs the selected delivery destinations; an acquisition step of inputting the group of delivery destination candidates into the machine learning model and acquiring the selected delivery destinations outputted, A proposal method in which, in the proposal step after the machine learning model is created, the acquired selected delivery destination is set as the next delivery destination.

10. A computer program comprising: A function to suggest the next delivery destination for the delivery person from a group of delivery destination candidates, a function of transmitting the group of delivery destination candidates and a selected delivery destination that the delivery person actually selected as the next delivery destination from the group of delivery destination candidates to a server; a function of transmitting the group of delivery destination candidates to the server for new proposals for the next delivery destination after the machine learning model is created by performing machine learning using learning data created using past data stored in the server in which the group of delivery destination candidates and the selected delivery destinations are associated, the machine learning model inputting the group of delivery destination candidates and outputting the selected delivery destinations; a function of receiving from the server the selected delivery destination output by inputting the group of delivery destination candidates into the machine learning model; a computer program that causes a computer to realize a function of proposing the received selected delivery destination as the next delivery destination.

Images