CN114943407A

CN114943407A - Area planning method, device, equipment, readable storage medium and program product

Info

Publication number: CN114943407A
Application number: CN202210351294.7A
Authority: CN
Inventors: 秦锴; 丁雪涛; 毛禛; 何仁清; 柳星宇; 徐义尧; 易琴; 余锦斌; 张力夫
Original assignee: Beijing Sankuai Online Technology Co Ltd
Current assignee: Beijing Sankuai Online Technology Co Ltd
Priority date: 2022-04-02
Filing date: 2022-04-02
Publication date: 2022-08-26

Abstract

The application discloses a region planning method, a device, equipment, a readable storage medium and a program product, and relates to the field of machine learning. The method comprises the following steps: acquiring a geographical position set, wherein the geographical position set comprises geographical position coordinates corresponding to at least two target type entities in a candidate area range; determining environment data sets corresponding to at least two target type entities based on the geographic position coordinates; performing position analysis on at least two target type entities by combining entity service distribution and entity position relation to obtain candidate centers in a candidate area range; and performing region division on the candidate region range based on the candidate center to obtain region planning results of the at least two target type entities in the candidate region range. Namely, the accuracy and the planning efficiency of the area planning can be better improved by a mode of determining the area planning result by analyzing the position of the target type entity in combination with the entity service distribution.

Description

Area planning method, device, equipment, readable storage medium and program product

Technical Field

The present disclosure relates to the field of machine learning, and in particular, to a method, an apparatus, a device, a readable storage medium, and a program product for area planning.

Background

Region planning generally refers to a process of dividing a certain region into several non-overlapping sub-regions, such as: the urban area is divided into several non-overlapping sub-areas. In the dividing, usually, a plurality of central points are determined in the region, so that the corresponding sub-region range is determined according to each central point.

In the related technology, after n position points (n is more than or equal to 2) contained in a current area are determined, Clustering of the position points is carried out on the n position points contained in a city according to the corresponding geographical position distribution relation through a Noise-Based Density Clustering method (DBSCAN), a final Clustering result is obtained, and a subregion planning result is determined according to the Clustering result.

However, in the above method, since the method of clustering only depends on the geographical location distribution relationship of the location points, different service requirements cannot be met, and the accuracy of the area planning result is low.

Disclosure of Invention

The embodiment of the application provides a method, a device, equipment, a readable storage medium and a program product for area planning, which can improve the accuracy of area planning. The technical scheme is as follows:

in one aspect, a method for planning an area is provided, where the method includes:

acquiring a geographical position set, wherein the geographical position set comprises geographical position coordinates corresponding to at least two target type entities in a candidate area range;

determining environment data sets corresponding to the at least two target type entities based on the geographic position coordinates, wherein the environment data sets are used for indicating entity position relations and entity service distributions corresponding to the at least two target type entities, and the entity service distributions are used for indicating the distribution conditions of the services of the target type entities on space and time;

performing position analysis on the at least two target type entities by combining the entity service distribution and the entity position relation to obtain candidate centers in the candidate area range;

and performing region division on the candidate region range based on the candidate center to obtain region planning results of the at least two target type entities in the candidate region range.

In another aspect, an area planning apparatus is provided, the apparatus including:

the acquisition module is used for acquiring a geographical position set, and the geographical position set comprises geographical position coordinates corresponding to at least two target type entities in a candidate area range;

a determining module, configured to determine, based on the geographic position coordinates, environment data sets corresponding to the at least two target type entities, where the environment data sets are used to indicate entity position relationships and entity service distributions corresponding to the at least two target type entities, and the entity service distributions are used to indicate temporal and spatial distribution conditions of services of the target type entities;

the analysis module is used for carrying out position analysis on the at least two target type entities by combining the entity service distribution and the entity position relation to obtain a candidate center in the candidate area range;

and the dividing module is used for carrying out region division on the candidate region range based on the candidate center to obtain a region planning result of the at least two target type entities in the candidate region range.

In another aspect, a computer device is provided, which includes a processor and a memory, where at least one instruction, at least one program, a set of codes, or a set of instructions is stored in the memory, and the at least one instruction, the at least one program, the set of codes, or the set of instructions is loaded and executed by the processor to implement the area planning method according to any of the embodiments of the present application.

In another aspect, there is provided a computer readable storage medium having stored therein at least one instruction, at least one program, set of codes, or set of instructions, which is loaded and executed by a processor to implement a method of area planning as described in any of the embodiments of the present application.

In another aspect, a computer program product or computer program is provided, the computer program product or computer program comprising computer instructions stored in a computer readable storage medium. The processor of the computer device reads the computer instructions from the computer-readable storage medium, and the processor executes the computer instructions to cause the computer device to perform the area planning method described in any of the above embodiments.

The beneficial effects brought by the technical scheme provided by the embodiment of the application at least comprise:

the method comprises the steps of analyzing geographic position coordinates of target type entities in a geographic position set, determining entity position relations and entity service distributions corresponding to the target type entities, combining the entity service distributions and the entity position relations in the process of analyzing the positions of the target type entities, so that candidate centers for dividing candidate area ranges are obtained finally, and finally obtaining area planning results.

Drawings

In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings needed to be used in the description of the embodiments are briefly introduced below, and it is obvious that the drawings in the following description are only some embodiments of the present application, and it is obvious for those skilled in the art to obtain other drawings based on these drawings without creative efforts.

Fig. 1 is a schematic diagram of a region planning method provided in an exemplary embodiment of the present application;

FIG. 2 is a schematic illustration of an implementation environment provided by an exemplary embodiment of the present application;

FIG. 3 is a flow chart of area planning provided by an exemplary embodiment of the present application;

FIG. 4 is a flow chart of a method of area planning provided by another exemplary embodiment of the present application;

FIG. 5 is a schematic diagram of an actor network analysis process provided by an exemplary embodiment of the present application;

FIG. 6 is a schematic illustration of a result of a region planning provided by another exemplary embodiment of the present application;

FIG. 7 is a flow chart of a method for area planning provided by an exemplary embodiment of the present application;

FIG. 8 is a schematic diagram of a target strategy model training process provided by an exemplary embodiment of the present application;

fig. 9 is a flowchart of a method for area planning provided by another exemplary embodiment of the present application;

fig. 10 is a block diagram of an area planning apparatus provided in an exemplary embodiment of the present application;

fig. 11 is a block diagram of an area planning apparatus according to another exemplary embodiment of the present application;

fig. 12 is a block diagram of a computer device according to an embodiment of the present application.

Detailed Description

To make the objects, technical solutions and advantages of the present application more clear, embodiments of the present application will be described in further detail below with reference to the accompanying drawings.

First, a brief description is given of terms referred to in the embodiments of the present application:

markov Decision Process (MDP) is a mathematical model of Sequential Decision, which is mainly constructed by a set of interactive objects, namely an Agent and an Environment, and has elements including a State (State), an Action (Action), a strategy (Policy) and an expected value (Reward).

Actor Critic Algorithm (AC): a reinforcement learning algorithm that incorporates both policy-based and desired value-based algorithms, wherein the policy-based algorithms act as actor networks for selecting corresponding behaviors based on probability. And taking the algorithm based on the expected value as a critic network for evaluating the scores of the behaviors corresponding to the actors, adjusting the probability of the behaviors according to the scores of the critics by the actors, and finally training the algorithm based on the strategy to obtain the target model.

Machine Learning (ML): the method is a multi-field cross discipline and relates to a plurality of disciplines such as probability theory, statistics, approximation theory, convex analysis, algorithm complexity theory and the like. Machine learning is the core of artificial intelligence, is the fundamental approach for computers to have intelligence, and is applied to all fields of artificial intelligence. Machine learning and deep learning generally include techniques such as artificial neural networks, belief networks, reinforcement learning, transfer learning, inductive learning, and teaching learning.

Referring to fig. 1, schematically, a schematic diagram of a region planning method provided by an exemplary embodiment of the present application is shown, as shown in fig. 1, a candidate region range 100 is determined, and geographic position coordinates corresponding to at least two target type entities (represented by hollow circles in the drawing) are included in the candidate region range 100, where the target type entities are, for example: the method comprises the steps of determining an environment data set 110 corresponding to each target type entity according to geographic position coordinates, wherein the environment data set 110 comprises an entity position relation 111 and an entity service distribution 112 corresponding to at least two target type entities, performing position analysis on the target type entities by combining the entity position relation 111 and the entity service distribution 112 to obtain a candidate center 120 (shown by a black solid circle in figure 1) in a candidate area range 100, determining an area center 121 according to the candidate center 120, performing area division on the candidate area range through the area center 121, and determining an area planning result 130 in the candidate area range 100.

According to the area planning method, the entity position relation and the entity service distribution corresponding to the target type entity are determined, and the position analysis is performed on the entity position relation and the entity service distribution, so that the relevant entity service information of the target type entity except the geographical position coordinates can be combined in the process of determining the candidate center in the area planning, and the accuracy of the area planning is improved.

With reference to the above noun introduction, an application scenario of the embodiment of the present application is illustrated:

1.the method is applied to a distribution area planning scene.In the delivery area planning scenario, for example: express delivery transportation area planning, takeout distribution area planning and the like, wherein in the case of the takeout distribution area planning, geographical position coordinates corresponding to n merchants corresponding to a city are obtained, wherein n is a positive integer, the corresponding position relationship among the n merchants and the corresponding related service distribution (such as historical order data, expected values of the merchants for the order quantity, distribution difficulty of the merchants and the like, which are not limited) of the merchants are determined according to the geographical position coordinates corresponding to the n merchants, the n merchants are used as environment data sets corresponding to the n merchants, the position analysis is carried out on the merchants by combining the corresponding position relationship among the n merchants and the corresponding related service distribution of the merchants to obtain candidate merchants, the takeout distribution area planning corresponding to the city is carried out by carrying out area division according to the geographical positions of the candidate merchants, and finally a takeout distribution area division result is obtained;

2.be applied to road planning fieldAnd (5) landscape.When a road in a city is planned, obtaining geographic position coordinates corresponding to n city buildings contained in the city, determining a corresponding position relationship among the n city buildings and service distribution corresponding to the city buildings (such as traffic flow conditions, the number of residents and the like corresponding to the city buildings) according to the geographic position coordinates corresponding to the n city buildings, using the position relationship corresponding to the n city buildings and the service distribution corresponding to the city buildings as an environment data set corresponding to the n city buildings, performing position analysis on the n city buildings by combining the position relationship corresponding to the n city buildings and the service distribution corresponding to the city buildings, determining candidate city buildings, performing regional division on the city according to the candidate city buildings to obtain a division result, and designating a road planning scheme according to a contour corresponding to the divided region.

It should be noted that the area planning method provided in the embodiment of the present application may be implemented by a terminal, may also be implemented by a server, and may also be implemented by cooperation of the terminal and the server.

When the terminal and the server cooperatively implement the scheme provided by the embodiment of the present application, the terminal and the server may be directly or indirectly connected in a wired or wireless communication manner, which is not limited in the embodiment of the present application.

Referring to fig. 2, a schematic diagram of an implementation environment provided by an exemplary embodiment of the present application is shown, as shown in fig. 2, the implementation environment includes a terminal 210, a server 220, and the terminal 210 and the server 220 are connected through a communication network 230.

The terminal 210 is installed with an application program providing an area planning function, and when the terminal 210 runs the application program, the terminal 210 generates an area planning request by using geographic position coordinates corresponding to at least two target type entities corresponding to a candidate area range, and sends the area planning request to the server 220.

The target policy model 221 in the server 220 determines an environment data set corresponding to the terminal 210 according to the area planning area sent by the terminal 210, where the environment data set includes an entity location relationship and entity service distribution corresponding to at least two target type entities, the server further includes a target policy model 221, the environment data set is input into the target policy model and output to obtain a candidate center, the at least two target type entities are subjected to area division according to the candidate center, a final area planning result is determined, and the final area planning result is fed back to the terminal 210.

The target policy model 221 included in the server 220 is obtained by training through an AC framework, wherein an Actor network 223(Actor network) and a Critic network 224(Critic network) in the AC framework are trained through a sample data set 222.

The terminal 210 includes at least one of a smartphone, a tablet computer, a portable laptop computer, a desktop computer, a smart speaker, a smart wearable device, and a vehicle-mounted terminal.

It should be noted that the communication network 230 may be implemented as a wired network or a wireless network, and the communication network 230 may be implemented as any one of a local area network, a metropolitan area network, or a wide area network, which is not limited in the embodiment of the present application.

In some embodiments, the server 220 may be implemented as a stand-alone single server or may be implemented as a group of servers including a plurality of cooperating servers within a group of servers.

In some embodiments, the server 220 may be implemented as a cloud server in a cloud end, or may be implemented as a node in a blockchain system, which is not limited herein.

With reference to the above noun introduction and application scenarios, the area planning method provided in the embodiment of the present application is described, taking the method executed by the server as an example for performing the description, schematically referring to fig. 3, which shows a flowchart of the area planning method provided in an exemplary embodiment of the present application, where the method includes the following steps:

step 301, a set of geographic locations is obtained.

And the geographic position set comprises geographic position coordinates corresponding to at least two target type entities in the candidate area range.

Illustratively, the candidate region range refers to a specified region corresponding to the current region planning, such as: taking a city as a designated area, the candidate area range is a city area corresponding to the city, taking a park as a designated area, the candidate area range is a park area corresponding to the park, taking an area set corresponding to multiple cities as a designated area, the candidate area range is an area set corresponding to multiple cities, and the like, and the application is not limited herein.

In some embodiments, the target type entity includes an entity corresponding to a specified type within the candidate region range, such as: when the specified type is a merchant, the target type entity is a merchant entity (such as a shop corresponding to the merchant); when the designated type is a landmark, the target type entity is a building or the like corresponding to the landmark, and when the designated type is a traffic place, the target type entity is a related entity (such as a building or other reference object or the like) corresponding to the location where the traffic place is located, which is not limited herein.

When the target type entity is a merchant entity, the geographic position coordinate corresponding to the target type entity is a store position coordinate corresponding to the merchant; when the target type entity is a building, the geographic position coordinate corresponding to the target type entity is a position coordinate corresponding to the building; and when the target type entity is a traffic place, the geographic position coordinate corresponding to the target type entity is the position coordinate in the candidate area range corresponding to the traffic place.

Optionally, at least two target type entities in the candidate region range are entities of the same type, or entities of different types, which is not limited herein.

Step 302, determining an environment data set corresponding to at least two target type entities based on the geographic position coordinates.

The environment data set is used for indicating entity position relation and entity service distribution corresponding to at least two target type entities, and the entity service distribution is used for indicating the distribution situation of the services of the target type entities on the space-time.

Optionally, the entity location relationship comprises a route distance between two target type entities; or, the distribution relationship of the corresponding positions among the multiple target type entities, such as: the entity a, the entity b and the entity c are distributed in a ring shape; or, the position distribution state of a plurality of target type entities in the candidate area range, such as: the entity 1, the entity 2 and the entity 3 correspond to position points in a cluster, the entity 4 is a corresponding isolated point outside the cluster, and the position relation of the entities is determined by the geographic position coordinates corresponding to at least two target type entities.

Illustratively, the entity service distribution includes a temporal distribution situation or a spatial distribution situation of services of the target type entity, where when service types corresponding to the target type entity are different, the corresponding entity service distributions may be the same or different.

When the target type entity is a merchant entity, the service type corresponding to the merchant entity includes a distribution service, an operation service, and the like, where the distribution service is taken as an example for explanation, for example: when the distribution areas of a plurality of merchants in the designated area range are divided, the distribution of entity business distribution corresponding to merchant entities over time includes: the historical order quantity of the merchant in the historical time period, the business time corresponding to the merchant, the expected order quantity set by the merchant, the historical business turnover of the merchant and the like, wherein the distribution of the entity business corresponding to the merchant entity on the space comprises the following steps: the order amount corresponding to the region where the merchant is located, the picking and delivering difficulty corresponding to the merchant, the location profile corresponding to the location where the merchant is located, and the like, which are not limited. Notably, the entity service distribution and the entity location relationship corresponding to one or more target type entities are selected as the environment data set.

Optionally, the entity service distributions corresponding to the target type entities in the environment data set are the same or different, and are not limited herein.

And 303, analyzing the positions of at least two target type entities by combining entity service distribution and entity position relation to obtain a candidate center in a candidate area range.

In some embodiments, the candidate center refers to a position center corresponding to each divided region when the region is divided in the current region range.

Optionally, the candidate center includes one or more target type entities selected from at least two target type entities as the candidate center; or, determining at least one candidate center (or unselected candidate center) in the candidate area range according to the geographic positions corresponding to the at least two target type entities, that is, the candidate center is a certain target type entity or a certain geographic position between a plurality of target type entities, which is not limited herein.

In some embodiments, the location analysis refers to analyzing the adaptation results of at least two target type entities as candidate centers when performing region partitioning.

Illustratively, the position analysis means includes at least one of the following means:

1. establishing a target strategy model, inputting entity service distribution and an entity position relation into the target strategy model, and outputting to obtain a position analysis result corresponding to each target type entity so as to determine a candidate center, wherein the target strategy model is a model for performing position analysis on the target type entity;

2. the method comprises the steps of dividing a candidate region range in advance, determining region features corresponding to each candidate region, extracting position features of entity position relations corresponding to target type entities and parameter features corresponding to entity service distribution, fusing the position features and the parameter features corresponding to the target type entities to obtain fusion features corresponding to the target type entities, analyzing the association degree according to the fusion features and the region features corresponding to the target type entities to obtain the center association degree corresponding to each target type entity, and using the target type entity corresponding to the highest association degree with the candidate region as a candidate center according to the center association degree. And subsequently, the division of the candidate region is adjusted based on the candidate center.

It should be noted that the above-mentioned location analysis method is only an illustrative example, and the present application is not limited thereto.

In some embodiments, the environment data set is input into a target policy model, and a candidate center is obtained through output, wherein the target policy model is used for performing position analysis on at least two target type entities by combining entity service distribution and entity position relation.

Optionally, the manner of performing the location analysis by the target policy model includes at least one of the following manners:

1. after the environment data set is input into the target strategy model, the target strategy model performs position analysis according to entity service distribution corresponding to each target type entity to obtain a position analysis score corresponding to each target type entity, and at least one target type entity with the highest score in the position analysis scores is selected as a candidate center;

2. the parameter threshold is preset in the target strategy model, after the environment data set is input into the target strategy model, the target strategy model compares the entity service distribution corresponding to each target type entity with the parameter threshold, and the target type entity corresponding to the entity service distribution reaching the parameter threshold is used as a candidate center.

It should be noted that the above-mentioned manner for performing the location analysis by the objective policy model is only an illustrative example, and the present application is not limited thereto.

And 304, performing region division on the candidate region range based on the candidate center to obtain region planning results of at least two target type entities in the candidate region range.

In some embodiments, the area division refers to dividing the candidate area into at least one sub-area, and each sub-area includes an area center as a center position corresponding to the sub-area.

Illustratively, the region planning result is used to indicate a partitioning result corresponding to at least one sub-region within the candidate region range.

Optionally, the candidate center is used as a region center, and the candidate region range is divided according to the region center; or, according to the candidate center, determining a region center corresponding to the candidate region range, so as to perform region division on the candidate region range, that is, the candidate center may be a certain target type entity, or may be a certain geographic location coordinate, which is not limited herein.

Optionally, in the range of the candidate region, there is an overlapping portion between the planning regions corresponding to the region planning result, or the planning regions are closely attached without an overlapping portion, which is not limited in this respect.

It should be noted that information (including but not limited to user equipment information, user personal information, etc.), data (including but not limited to data for analysis, stored data, presented data, etc.), and signals referred to in this application are authorized by the user or sufficiently authorized by various parties, and the collection, use, and processing of the relevant data is required to comply with relevant laws and regulations and standards in relevant countries and regions. For example, the entity location relationship, the entity service distribution, and the data of the relevant entities such as the regional center corresponding to the target type entity referred to in the present application are all obtained under the condition of sufficient authorization.

To sum up, the area planning method provided in the embodiment of the present application determines the entity position relationship and the entity service distribution corresponding to each target type entity by analyzing the geographical position coordinates of the target type entities in the geographical position set, and combines the entity service distribution and the entity position relationship in the process of performing position analysis on the target type entities, so as to obtain a candidate center for dividing the candidate area range, and finally obtain an area planning result, that is, by determining the entity position relationship and the entity service distribution corresponding to the target type entities, so that the relevant service information corresponding to the target type entities can be combined in the process of determining the candidate center by area planning, an area planning scheme under different service requirements can be satisfied, and the accuracy of area planning can be improved.

In an alternative embodiment, please refer to fig. 4, which shows a flowchart of a method for planning an area according to an exemplary embodiment of the present application. This method is described as an example in which the server executes the method. That is, on the basis of the embodiment shown in fig. 3, step 303 further includes step 3031:

step 301, a geographical location set is obtained.

Illustratively, the candidate region range refers to a designated region corresponding to the current region planning.

Illustratively, the area plan includes at least one of the following plan contents:

1. and carrying out region division on the designated region. Such as: dividing a city into different regions with specified number according to business requirements (such as take-out distribution region planning, express delivery transportation region planning and the like) for carrying out business management on the different regions;

2. and planning a route of the designated area. Such as: planning a route in a certain park, dividing the park into a plurality of different areas, and determining a route laying scheme according to the area outline corresponding to each area;

3. and performing area aggregation on the designated area. Such as: the design of the transportation network between provinces and cities is carried out, and the set result corresponding to the cities is determined by planning a plurality of cities in a certain province and is used for constructing the transportation network.

It should be noted that the above planning content related to the area planning is only an illustrative example, and the embodiment of the present application does not limit this.

Optionally, the geographic position coordinate of the target type entity is a plane coordinate, or the geographic position coordinate of the target type entity is a three-dimensional coordinate; the geographic position coordinates of the target type entity include, but are not limited to, longitude and latitude coordinates determined according to the candidate position area range, or the geographic position coordinates of the target type entity are longitude and latitude coordinates corresponding to the entity.

Illustratively, the determining manner of the entity service distribution includes at least one of the following manners:

1. acquiring historical data corresponding to a target type entity as entity service distribution of the target type entity;

2. presetting a time range, and acquiring historical data corresponding to a target type entity in the time range as entity service distribution;

3. after the parameter type of the target type entity is determined, a parameter threshold is preset for the parameter, and the parameter is used as entity service distribution corresponding to the target type entity, namely, the parameter threshold is an expected value corresponding to the parameter type of the target type entity.

It should be noted that the above determination manner regarding the entity service distribution is only an illustrative example, and the embodiment of the present application does not limit this.

And step 3031, inputting the environment data set into the target strategy model, and outputting to obtain the candidate center.

The target strategy model is used for carrying out position analysis on at least two target type entities by combining entity service distribution and entity position relation.

In some embodiments, the at least two target type entities include merchant entities, the environmental data set is used to indicate entity location relationships of the merchant entities, and business distributions of the merchant.

Optionally, in an implementable scenario, the target type entity is implemented as a merchant entity, such as: the environmental data set includes the entity location relationship corresponding to the merchant entity and the business distribution corresponding to the merchant.

The entity position relation corresponding to the merchant entity comprises a corresponding route distance between the two merchant entities; alternatively, the location distribution relationship of the plurality of business entities within the candidate area range is not limited herein.

The service distribution corresponding to the merchant refers to relevant service distribution information contained in the merchant under the specified service type, such as: when the merchant is in a take-away business scene, the business distribution corresponding to the merchant comprises at least one of the take-away order quantity of the merchant, the meal taking difficulty corresponding to the merchant, the meal taking difficulty of the corresponding area in the designated area where the merchant is located, the area outline corresponding to the position where the merchant is located, the business distribution information such as the order expectation preset by the merchant manager for the merchant, and the like.

In some embodiments, based on the environmental data set, determining a historical region center of the candidate region range, the historical region center comprising merchant entities that are divided region centers within a historical time period; and analyzing the historical region center and the environment data set through the target strategy model, and outputting to obtain a candidate center.

Illustratively, in the environment data set, the corresponding region center selection states of at least two merchant entities in the historical time period are further included, that is, each merchant entity includes one region center selection state: selected as the region center or not selected as the region center.

The selected area center is selected once or multiple times as the area center in the process that the business entity performs area planning on the candidate area range in the historical time period; the non-selected area center means that the merchant entity does not select the history record as the area center in the process of area planning of the candidate area range in the historical time period. Therefore, the target type entity selected as the area center is taken as the history area center.

In some embodiments, the historical time period includes a certain historical time range, or includes at least one historical region planning process, which is not limited herein.

Optionally, one or more historical region centers may be determined from the environment data set during the current region planning process, or no historical region center may exist, which is not limited herein.

In this embodiment, the environment data set is used as environment, and the history area center is a State corresponding to the merchant entity, that is, the current merchant entity is correspondingly included in the history time period and is used as the State of the area center.

And inputting the environment data sets corresponding to at least two merchant entities and the historical region center into the target strategy model to perform position analysis on the merchant entities, and outputting to obtain the merchant entities serving as candidate centers.

In some embodiments, the target strategy model comprises an actor network in an reinforcement learning architecture, the actor network is used for performing center prediction, and the actor network comprises an encoder and a decoder in a coding and decoding architecture.

In this embodiment, the reinforcement learning framework includes an actor network, the center prediction refers to analyzing whether a target type entity can be used as a candidate center, and the actor network is configured to perform prediction analysis on data related to the target type entity of the input target policy model to determine the target type entity serving as the candidate center, where the actor network implements a center prediction function through an encoder and a decoder in the encoding and decoding framework.

In some embodiments, based on the environment data set, determining a historical region center of the candidate region range, the historical region center including a target type entity as a divided region center within a historical time period; inputting the historical region center and the environment data set into an encoder of an actor network, and outputting to obtain an encoding characteristic and a center characteristic set; decoding and analyzing the coding characteristics through a decoder to obtain decoding characteristics; and obtaining a candidate center based on the decoding characteristic and the center characteristic set.

In this embodiment, the center selection state of each target type entity in the historical time range is determined by the environment data set, and the target type entity whose center selection state is selected as the candidate center is determined as the historical region center. And inputting the environment data set and the historical region center into a target strategy model, and performing position analysis on at least two target type entities by combining the entity service distribution corresponding to the target type entities in the environment data set and the historical region center to determine the target type entities which are finally used as candidate centers.

Referring to fig. 5, which is a schematic diagram illustrating an actor network analysis process according to an exemplary embodiment of the present application, as shown in fig. 5, a center selection state 510 and a geographic location coordinate 520 corresponding to a target type entity are input into an encoder 530, where the center selection state 510 is used to indicate whether the target type entity is selected as a candidate center in a historical time period, that is, the center selection state 510 includes two states of selecting and not selecting, a value range of the center selection state 510 is [0, t ], 0 indicates that the target type entity is not selected as a candidate center in the historical time period, t indicates that the target type entity is selected as a candidate center in a tth selection process, and the geographic location coordinate 520 is a two-dimensional plane coordinate (ln, lat) of the target type entity in a candidate area range.

The encoder 530 includes two neural networks (531 and 532), the neural network 531 is configured to extract a state feature 541 corresponding to the center selection 510, the neural network 532 is configured to extract a position feature 542 corresponding to the geographic position coordinate 520, the center selection state 510 and the geographic position coordinate 520 are simultaneously input to the neural network 531 and the neural network 532, the state feature 541 and the position feature 542 corresponding to each target type entity are obtained, and the state feature representation 541 and the position feature representation 542 corresponding to each target type entity are used as a center feature set 540.

The target policy model also includes a decoder, which includes a neural network 550. Determining an encoding feature 543 corresponding to the center of the history region through the neural network 532, further determining a service feature 544 corresponding to the center of the history region according to the entity position relationship and the entity service distribution corresponding to the target type entity, inputting the encoding feature 543 and the service feature 544 into the neural network 550, outputting to obtain a decoding feature 551, wherein the decoding feature 551 is used for indicating a feature representation corresponding to the candidate center 560, and determining the final candidate center 560 through an Attention mechanism (Attention mechanism) by using the center feature set 540 and the decoding feature 551.

In this embodiment, the Neural Networks 531 and 532 are implemented as a network with a Convolutional Neural Network (CNN) structure, and the Neural network 550 is implemented as a network with a Long Short-Term Memory (LSTM) structure, which is not limited in this application.

And 304, performing region division on the candidate region range based on the candidate center to obtain region planning results of the at least two target type entities in the candidate region range.

Illustratively, in the process of determining the center of the area according to the candidate centers, the determining method includes:

1. taking a candidate center obtained by the output of the target strategy model as a region center;

2. when m candidate centers are included in the same planning region, connecting position points corresponding to the m candidate centers, and taking an intersection obtained by connecting the positions as a region center, wherein m is a positive integer;

3. and after the target strategy model obtains the candidate center, traversing the candidate center by using a weighted centroid algorithm to determine the final area center.

It should be noted that the above determination manner regarding the center of the area is only an illustrative example, and the present embodiment does not limit this.

Illustratively, the determining manner of the area planning result includes at least one of the following manners:

1. presetting a designated number, and taking the area center as a reference after determining the area center, and taking the area range corresponding to the target type entities with the designated number as a planned area;

2. after the center of the area is determined, determining a planning area corresponding to the center of the area according to the preset area contour by taking the center of the area as a reference;

3. determining a region planning result according to the service requirement, such as: when regional planning is carried out on merchants, merchants operating the same type of content are distributed in different planning regions, and merchants having business contacts (such as movie theaters and restaurants) are placed in the same planning region so as to meet the final business requirements.

It should be noted that the above-mentioned contents related to the area planning result are only illustrative examples, and the present application embodiment is not limited thereto.

Illustratively, the region planning result includes at least one of the following results:

1. the area planning result is to display each planning area divided according to the area center in the candidate area range, namely, each planning area in the candidate area range;

2. the area planning result is to display the area outline corresponding to each planning area in the candidate area range, namely, to display the graphic shape and the corresponding contour line corresponding to each planning area in the candidate area range;

3. the area planning result is a clustering result corresponding to at least two target type entities displayed in the candidate area range, that is, at least one clustering result corresponding to the target type entities displayed in the candidate area range.

It should be noted that the above-mentioned contents related to the area planning result are only illustrative examples, and the present application does not limit this.

In some embodiments, performing centroid localization on the candidate center to obtain a region center corresponding to the candidate region range; determining entity clustering results corresponding to at least two target type entities in the candidate region range based on the region center; and taking the entity clustering result as an area planning result.

Schematically, after candidate centers corresponding to at least two target type entities are obtained, at least one area center is determined from the candidate centers through a weighted centroid algorithm, the target type entities and the area centers closest to the target type entities are clustered through an Expectation-maximization (EM) algorithm to obtain entity clustering results corresponding to the target type entities in a candidate area range, wherein one target type entity is clustered with one or more area centers, and a final area planning result is determined through a graph outsourcing algorithm according to the entity clustering results, wherein the graph outsourcing algorithm is used for determining a polygonal profile corresponding to the geographical position where the target type entity is located as a basic unit, the outsourcing profile of a planning area corresponding to each area center is determined through the basic unit, namely, at least one outsourcing graph is generated in the current candidate area range, as a result of the area planning.

Referring to fig. 6, schematically, a schematic diagram of a region planning result provided by an exemplary embodiment of the present application is shown, as shown in fig. 6, fig. 6 is shown as a region planning result interface 600, where the region planning result interface 600 includes at least one planning region 601 (illustrated by a region depicted by a bold line in fig. 6), and an outsourcing contour of the planning region 601 refers to a contour line corresponding to a range where the planning region 601 is located.

In this embodiment, the historical region center is determined by the environment data set, the historical region center and the environment data set are input to the target policy model together for performing the position analysis on the target type entity, and the analysis accuracy of the target policy can be improved by performing the analysis in combination with the center selection state of the target type entity, so that the accuracy of the region planning is improved.

In an alternative embodiment, please refer to fig. 7, which shows a flowchart of a method for area planning according to an exemplary embodiment of the present application. This embodiment is used to explain the training side of the area planning method, and this method is executed by a server as an example. That is, on the basis of the embodiment shown in fig. 3, step 304 further includes the following steps:

step 701, acquiring a sample data set.

The sample data set is used for indicating the sample position relation and the sample service distribution corresponding to at least two sample type entities.

In some embodiments, the sample type entity comprises: a business, a building, a traffic location, etc., wherein at least two sample type entities are located within a sample candidate area.

When the sample type entity is a merchant, the geographic position coordinate corresponding to the sample type entity is a shop position coordinate corresponding to the merchant; when the sample type entity is a building, the geographic position coordinate corresponding to the sample type entity is a position coordinate corresponding to the building; when the sample type entity is a traffic place, the geographic position coordinate corresponding to the sample type entity is the position coordinate in the sample candidate area range corresponding to the traffic place.

Optionally, the sample location relationship comprises a route distance between two sample type entities; or, the distribution relation of the positions corresponding to the multiple sample type entities; or, the position distribution state of the plurality of sample type entities in the sample candidate area range is not limited herein, wherein the sample position relationship is determined by the geographic position coordinates corresponding to at least two sample type entities.

In some embodiments, the sample service distribution refers to a relevant service parameter corresponding to a single target type entity, and the sample service distribution is different according to different service requirements, for example: when a commodity distribution area is planned for a merchant in a city, according to the demand of distribution business, the sample business distribution includes the historical number of orders of the merchant in a historical time period, the business type and the business time of the merchant, the goods taking and distribution difficulty of the merchant, the number of orders expected by the merchant, the area outlines corresponding to areas where a plurality of merchants are located, and the like, which is not limited herein. And selecting one or more sample service distributions as a sample data set in the process of planning the region.

Illustratively, the geographic position coordinate corresponding to the sample type entity is a plane position coordinate.

Step 702, predicting the sample data set through the actor network to obtain sample candidate centers of at least two sample type entities.

Schematically, a sample history center corresponding to a sample data set is determined, wherein before first training, the number of centers is preset, a sample type entity corresponding to the number of the preset centers in the sample data set is determined as a sample candidate center by a random selection method and is used as a sample history center for inputting in an actor network in the first training process, then in the first training process, the sample data set and the sample history center are input in the actor network, and a sample candidate center generated correspondingly in the first training process is output and obtained and is used as the sample history center in the second training process.

And starting from the second training, and taking the sample candidate center obtained in the previous training as a corresponding sample history center in the current training process in each training process.

And 703, evaluating the sample candidate center through a critic network in the reinforcement learning architecture to obtain the expected deviation.

Wherein the expected deviation is used to indicate the predicted effect of the sample candidate center.

In this embodiment, the reinforcement learning architecture further includes a critic network, and the critic network is used for generating a predicted value of a result corresponding to the estimated sample candidate center.

First, substituting the sample candidate centers into a preset expectation function to obtain first expectation values corresponding to the sample candidate centers, wherein the first expectation values are used for indicating aggregation relations among the sample candidate centers.

Illustratively, the preset expectation function is a preset function determined according to the sample service type corresponding to the sample type entity, and is used for determining a first expectation value corresponding to the sample candidate center.

In this embodiment, in a scenario of planning a distribution area of a merchant, a first expected value is determined in a manner referring to formula one:

the formula I is as follows: rt is alpha, sum of single amount in a circle with a certain radius + beta, distance between two merchants at the center point + gamma (1-area average distribution difficulty) + delta (KL (area preset single amount distribution, merchant estimated single amount distribution)

The method comprises the following steps that alpha, beta, gamma and delta are correspondingly different adjustable weights, the sum of the single quantities in a circle with a certain radius refers to the sum of the number of orders of a sample type entity (a merchant) in a sample candidate planning area (a circle with a certain radius), the distance between every two merchants at the center point refers to the distance between corresponding routes of two sample candidate centers, the area average distribution difficulty refers to the average value of the distribution difficulty corresponding to the sample candidate planning area, KL (area preset single quantity distribution, merchant estimated single quantity distribution) refers to the relative entropy between the merchant manager preset single quantity distribution and the merchant estimated single quantity distribution, wherein the merchant manager preset single quantity distribution refers to the distribution condition of the preset single quantities corresponding to each planning area corresponding to an area planning result, and the merchant estimated single quantity distribution refers to the single quantity expected quantity estimated by each merchant.

Different weights are respectively configured for each sample service distribution and sample position relation corresponding to each sample candidate center, and a weighted sum result of each sample service distribution and sample position relation is calculated to serve as a first expected value corresponding to a sample candidate center cluster, wherein the sample candidate center cluster refers to a center set corresponding to the sample candidate center.

It should be noted that the parameters used in the first formula for determining the first desired value are only illustrative examples, and different parameters may be set according to actual requirements during the application process.

Illustratively, the aggregation relation refers to the degree of position dispersion of the sample candidate center within the area planning range.

And secondly, performing expected analysis on the sample data set through a critic network to obtain a second expected value corresponding to the sample data set, wherein the second expected value is used for indicating the pre-estimated aggregation relation corresponding to the sample data set.

In some embodiments, the sample data set is input into a candidate pre-estimation model, the candidate pre-estimation model comprises a critic network, and the candidate pre-estimation model is a machine learning model obtained through pre-training; and (4) performing expected estimation on the sample data set through a critic network, and outputting to obtain a second expected value.

In this embodiment, a candidate prediction model including a critic network is constructed, and a second expected value corresponding to the sample data set is obtained through output of the critic network, where the critic network in the candidate prediction model may be implemented as a network structure including two layers of CNN networks and two layers of full-connected structures, and an output result is a scalar network structure used for representing a predicted value corresponding to the first expected value, that is, a prediction result of an aggregation relationship corresponding to the sample data set.

Finally, an expected deviation is obtained based on the first expected value and the second expected value.

In some embodiments, a gradient relationship is determined that corresponds to the first expected value and the second expected value; a gradient relationship between the first desired value and the second desired value is determined as the desired deviation.

Illustratively, the parameter adjustment direction of the actor network is determined by analyzing a gradient relationship between the first expected value and the second expected value, wherein the gradient relationship may be determined in a manner specifically referring to a formula two:

the formula II is as follows:

wherein, ω (Z) ⁱ | G) refers to a first expected value, b (G) refers to a second expected value,

refers to the gradient weight.

Step 704, training the actor network based on the expected deviation to obtain a target strategy model.

In this embodiment, the expected deviation corresponding to the first expected value and the second expected value is determined by the second formula, when the first expected value is closer to the second expected value, it indicates that the accuracy of the parameters of the actor network is higher, and when the first expected value is greater than the second expected value, the training direction of the actor network is adjusted by taking the first expected value as the positive direction of the gradient weight; and when the first expectation value is smaller than the second expectation value, adjusting the training direction of the actor network by taking the gradient weight as a negative direction and finally obtaining the target strategy model.

Step 705, based on the first expected value and the second expected value, adjusting model parameters of the candidate estimation model to obtain an expected estimation model.

Wherein the expectation prediction model is used to determine a second expectation value.

Illustratively, model parameters of the candidate prediction model are adjusted by determining a mean square error value corresponding to the first expected value and the second expected value, wherein the mean square error value can be calculated in a manner specifically referring to a formula three:

the formula III is as follows:

wherein, ω (Z) ⁱ | G) denotes a first desired value, b (G) denotes a second desired value, L _c And expressing the mean square error value, wherein the smaller the mean square error value is, the higher the model parameter precision of the candidate prediction model is.

Notably, during the training of the actor network, the critic network is also being trained, and only the actor network is involved in the final objective strategy model.

Referring to fig. 8, which is a schematic diagram illustrating a target policy model training process provided in an exemplary embodiment of the present application, as shown in fig. 8, a sample data set 810 is obtained, where the sample data set 810 includes geographic position coordinates corresponding to at least two sample type entities (t sample data sets required for t times of training are shown in the drawing), a sample history center corresponding to the sample type entity is determined according to the sample data set 810, and a center selection state 820 corresponding to each sample type entity in the sample data set 810 is determined according to the sample history center, that is, the center selection state 820 includes a sample type entity selected as the sample history center and a sample type entity unselected as the sample history center, where the selected state and the unselected state correspond to two different states.

Extracting parameter characteristics corresponding to sample business distribution of sample type entities in the sample data set 810, inputting the sample position relationship, the parameter characteristics and the center selection state 820 in the sample data set 810 into the actor network 830 (the input process of the sample position relationship and the parameter characteristics is not shown), and outputting to obtain a sample candidate center 840.

Aggregating the sample candidate centers output by the actor network to generate a sample candidate cluster, analyzing the aggregation degree of the sample candidate cluster by combining the sample business distribution and the sample position relation corresponding to the sample candidate center, and acquiring an aggregation analysis result corresponding to the sample candidate cluster as a first expected value. As shown in fig. 8, a corresponding sample candidate cluster 850 is determined by the sample candidate center 840, and a first expected value thereof is determined from the sample candidate cluster 850.

And inputting the sample data set 810 into the candidate prediction model 860, and outputting to obtain a second expected value, wherein the sample data set 810 is expected to be predicted through a critic network contained in the candidate prediction model, and the obtained prediction result is output as the second expected value. And determining an expected deviation through the first expected value and the second expected value, and using the expected deviation to train the actor network 830 to finally obtain a target strategy model.

In summary, the area planning method provided in the embodiment of the present application determines the entity position relationship and the entity service distribution corresponding to each target type entity by analyzing the geographical position coordinates of the target type entities in the geographical position set, and combines the entity service distribution and the entity position relationship in the process of performing position analysis on the target type entities, so as to obtain a candidate center for dividing the candidate area range, and finally obtain an area planning result, that is, by determining the entity position relationship and the entity service distribution corresponding to the target type entity, so as to combine the relevant service information corresponding to the target type entity in the process of determining the candidate center by area planning, and can satisfy the area planning scheme under different service requirements, thereby improving the accuracy of area planning.

In the embodiment, the candidate pre-estimation model is added in the training process to train the critic network, and the critic network is also trained in the training process of the actor network, so that the parameter precision of the actor network can be improved, and the accuracy of the obtained target strategy model is higher.

Referring to fig. 9, a flow chart of a method for planning an area according to an exemplary embodiment of the present application is schematically shown. That is, fig. 9 is a detailed description of the region planning method for training in the merchant distribution scenario, and the method includes the following steps:

and 910, preparing data.

In this embodiment, each merchant set in one city is used as a location point set on a two-dimensional plane, that is, the training process is to perform area planning on merchants in one city, and a single merchant is used as a sample type entity to obtain geographic location coordinates corresponding to each merchant.

Schematically, a sample data set corresponding to each merchant in a city is determined according to a geographic position coordinate corresponding to each merchant, wherein the sample data set comprises a sample geographic position relationship corresponding to the merchant and sample business distribution, the sample geographic position relationship comprises a route distance corresponding to two merchants, the number of orders of each merchant in a historical time period in the sample business distribution is determined, an outsourcing outline finally corresponding to a planning area in an area planning range determined according to distribution business, distribution difficulty corresponding to the merchant, expected number of orders preset by a management layer for the merchant and the like are determined without limitation, and feature representation corresponding to each data contained in the sample data set is determined to serve as an environment feature representation set (G).

The corresponding sample number B in the single training process and the training times E are preset so as to determine the number N corresponding to the planning area when the city is planned.

And 920, initializing the model.

In this embodiment, in the process of starting training, various parameters are initialized, including initializing the environment feature set (G), and initializing the actor network (θ) and the critic network (θ) _c ) Initializing, and determining the parameter characteristics corresponding to the merchants, the number of the planning areas and second expected values in different states. The critic network is a network contained in the expected estimation model.

The actor network comprises two convolutional neural networks and a long-term and short-term memory network, and the actor network comprises two convolutional neural networks and two full-connected layers.

930, the sample candidate center and the first expected value are determined.

In this embodiment, the training process includes determining a sample candidate center corresponding to each sample in each training process. And determining the corresponding merchants serving as the sample history center in the city, marking the merchants as selected states, and marking the remaining merchants as unselected states as center features.

The center feature and the environment feature are expressed and input into the actor network, and the sample candidate center (Z) is output ⁱ | G), determining a first expected value (ω (Z) corresponding to the sample candidate center according to the formula I ⁱ |G))。

And storing the sample candidate center and a first expected value corresponding to the sample candidate center.

940, a second expected value corresponding to the sample data set is determined.

And inputting the sample data set corresponding to each merchant into the critic network, and outputting to obtain a second expected value (b | G) corresponding to the sample data set.

950, model parameters are adjusted and updated.

After the first expected value and the second expected value are determined, the gradient relation corresponding to the first expected value and the second expected value is determined according to a formula II, the actor network is adjusted according to the gradient relation, and the parameters are updated to obtain the target strategy model.

And determining a mean square error value corresponding to the first expectation value and the second expectation value according to a formula III, performing parameter adjustment on the candidate prediction model, and updating parameters.

The beneficial effect that this scheme brought still includes:

1. the method comprises the steps of converting a mode of clustering entities to determine candidate centers into a sequence decision scheme, and judging a training result only according to a historical state;

2. the mode of determining the second expected value can integrate different service requirements, can better combine actual conditions and adapt to the development of services;

3. the model training efficiency can be improved by constructing a training model.

Fig. 10 is a block diagram of a structure of an area planning apparatus according to an exemplary embodiment of the present application, and as shown in fig. 10, the apparatus includes:

an obtaining module 1010, configured to obtain a geographic location set, where the geographic location set includes geographic location coordinates corresponding to at least two target type entities within a candidate area range:

a determining module 1020, configured to determine, based on the geographic position coordinates, environment data sets corresponding to the at least two target type entities, where the environment data sets are used to indicate entity position relationships corresponding to the at least two target type entities and entity service distribution, and the entity service distribution is used to indicate a distribution situation of services of the target type entities in space-time;

an analysis module 1030, configured to perform location analysis on the at least two target type entities according to the entity service distribution and the entity location relationship, so as to obtain a candidate center in the candidate area range;

the dividing module 1040 is further configured to perform area division on the candidate area range based on the candidate center, so as to obtain an area planning result of the at least two target type entities in the candidate area range.

In an alternative embodiment, the analyzing module 1030 includes:

an output unit 1031, configured to input the environment data set into a target policy model, and output to obtain the candidate center, where the target policy model is configured to perform location analysis on the at least two target type entities by combining the entity service distribution and the entity location relationship.

In an alternative embodiment, the at least two target type entities include a merchant entity, and the environmental data set is used to indicate an entity location relationship of the merchant entity and a business distribution of the merchant;

the output unit 1031 is further configured to determine, based on the environment data set, a historical region center of the candidate region range, where the historical region center includes a merchant entity serving as a divided region center in a historical time period; and analyzing the historical region center and the environment data set through the target strategy model, and outputting to obtain the candidate center.

In an optional embodiment, the target policy model comprises an actor network in a reinforcement learning architecture, the actor network is used for center prediction, and the actor network comprises an encoder and a decoder in a coding and decoding architecture;

the output unit 1031 is further configured to determine, based on the environment data set, a history area center of the candidate area range, where the history area center includes a target type entity serving as a divided area center in a history time period; inputting the historical region center and the environment data set into the encoder of the actor network, and outputting to obtain an encoding feature and a center feature set; decoding and analyzing the coding characteristics through the decoder to obtain decoding characteristics; and obtaining the candidate center based on the decoding characteristics and the center characteristic set.

In an optional embodiment, the target strategy model comprises an actor network in a reinforcement learning architecture, and the actor network is used for center prediction;

the device further comprises:

the obtaining module 1010 is further configured to obtain a sample data set, where the sample data set is used to indicate a sample position relationship and sample service distribution corresponding to at least two sample type entities;

the prediction module 1050 is configured to predict the sample data set through the actor network to obtain sample candidate centers of the at least two sample type entities;

an evaluation module 1060, configured to evaluate the sample candidate center through a critic network in the reinforcement learning architecture to obtain an expected deviation, where the expected deviation is used to indicate a prediction effect of the sample candidate center;

a training module 1070, configured to train the actor network based on the expected deviation to obtain the target strategy model.

In an optional embodiment, the evaluation module 1060 is further configured to substitute the sample candidate centers into a preset expectation function to obtain a first expectation value corresponding to the sample candidate centers, where the first expectation value is used to indicate an aggregation relationship between the sample candidate centers; performing expected analysis on the sample data set through the critic network to obtain a second expected value corresponding to the sample data set, wherein the second expected value is used for indicating the pre-estimated aggregation relation corresponding to the sample data set; the expected deviation is derived based on the first expected value and the second expected value.

In an optional embodiment, the evaluation module 1060 is further configured to determine a gradient relationship between the first expected value and the second expected value; determining a gradient relationship between the first expected value and the second expected value as the expected deviation.

In an optional embodiment, the evaluation module 1060 is further configured to input the sample data set into a candidate prediction model, where the candidate prediction model includes the critic network, and the candidate prediction model is a machine learning model obtained through pre-training; and performing expectation prediction on the sample data set through the critic network, and outputting to obtain the second expectation value.

In an optional embodiment, the apparatus further comprises:

an adjusting module 1080, configured to adjust a model parameter of the candidate prediction model based on the first expectation value and the second expectation value to obtain an expected prediction model, where the expected prediction model is used to determine the second expectation value.

In an optional embodiment, the analysis module 1030 is further configured to perform centroid location on the candidate center to obtain a region center corresponding to the candidate region range; determining entity clustering results corresponding to the at least two target type entities in the candidate region range based on the region center; and taking the entity clustering result as the area planning result.

To sum up, the area planning apparatus provided in this embodiment determines the entity location relationship and the entity service distribution corresponding to each target type entity by analyzing the geographical location coordinates of the target type entities in the geographical location set, and combines the entity service distribution and the entity location relationship in the process of performing location analysis on the target type entities, thereby obtaining a candidate center for dividing the candidate area range, and finally obtaining an area planning result, that is, by determining the entity location relationship and the entity service distribution corresponding to the target type entities, it is possible to combine other parameter information of the target type entities except the geographical location coordinates in the process of determining the candidate center by area planning, thereby improving the accuracy of area planning.

It should be noted that: the area planning apparatus provided in the foregoing embodiment is only illustrated by dividing the functional modules, and in practical applications, the functions may be allocated to different functional modules according to needs, that is, the internal structure of the device is divided into different functional modules, so as to complete all or part of the functions described above. In addition, the area planning apparatus and the area planning method provided by the above embodiments belong to the same concept, and specific implementation processes thereof are described in the method embodiments and are not described herein again.

Referring to fig. 12, a block diagram of a computer device 1200 according to an embodiment of the present application is shown. The computer device 1200 may be an electronic device as described above for implementing the above-described area planning method.

Generally, computer device 1200 includes: a processor 1201 and a memory 1202.

The processor 1201 may include one or more processing cores, such as a 4-core processor, an 8-core processor, or the like. The processor 1201 may be implemented in at least one hardware form of a DSP (Digital Signal Processing), an FPGA (Field Programmable Gate Array), and a PLA (Programmable Logic Array). The processor 1201 may also include a main processor and a coprocessor, where the main processor is a processor for Processing data in an awake state, and is also called a Central Processing Unit (CPU); a coprocessor is a low power processor for processing data in a standby state. In some embodiments, the processor 1201 may be integrated with a GPU (Graphics Processing Unit) that is responsible for rendering and drawing content that the display screen needs to display. In some embodiments, the processor 1201 may further include an AI (Artificial Intelligence) processor for processing a computing operation related to machine learning.

Memory 1202 may include one or more computer-readable storage media, which may be non-transitory. Memory 1202 may also include high-speed random access memory, as well as non-volatile memory, such as one or more magnetic disk storage devices, flash memory storage devices.

Those skilled in the art will appreciate that the configuration shown in FIG. 12 is not intended to be limiting of the computer device 1200 and may include more or fewer components than those shown, or some components may be combined, or a different arrangement of components may be used.

In an example embodiment, there is also provided a computer device comprising a processor and a memory, the memory having stored therein a computer program. The computer program is configured to be executed by one or more processors to implement the above-described area planning method.

In an exemplary embodiment, a computer-readable storage medium is also provided, in which a computer program is stored which, when being executed by a processor of a computer device, carries out the above-mentioned area planning method.

Alternatively, the computer-readable storage medium may be a ROM (Read-Only Memory), a RAM (Random Access Memory), a CD-ROM (Compact Disc Read-Only Memory), a magnetic tape, a floppy disk, an optical data storage device, and the like.

In an exemplary embodiment, a computer program product is also provided, which, when run on a computer device, causes the computer device to perform the above-mentioned area planning method.

It should be understood that reference to "a plurality" herein means two or more. "and/or" describes the association relationship of the associated object, indicating that there may be three relationships, for example, a and/or B, which may indicate: a exists alone, A and B exist simultaneously, and B exists alone. The character "/" generally indicates that the former and latter associated objects are in an "or" relationship. In addition, the step numbers described herein only exemplarily show one possible execution sequence among the steps, and in some other embodiments, the steps may also be executed out of the numbering sequence, for example, two steps with different numbers are executed simultaneously, or two steps with different numbers are executed in a reverse order to the order shown in the figure, which is not limited by the embodiment of the present application.

The above description is only exemplary of the present application and should not be taken as limiting the present application, and any modifications, equivalents, improvements and the like that are made within the spirit and principle of the present application should be included in the protection scope of the present application.

Claims

1. A method of area planning, the method comprising:

performing position analysis on the at least two target type entities by combining the entity service distribution and the entity position relation to obtain a candidate center in the candidate area range;

2. The method of claim 1, wherein the performing location analysis on the at least two target type entities in combination with the entity service distribution and the entity location relationship comprises:

and inputting the environment data set into a target strategy model, and outputting to obtain the candidate center, wherein the target strategy model is used for carrying out position analysis on the at least two target type entities by combining the entity service distribution and the entity position relation.

3. The method of claim 2, wherein the at least two target type entities comprise merchant entities, the set of environmental data is indicative of entity location relationships of the merchant entities, and business distributions of the merchant;

the inputting the environment data set into a target strategy model and outputting to obtain a candidate center comprises:

determining historical region centers of the candidate region ranges based on the environment data set, wherein the historical region centers comprise merchant entities which are divided into region centers in historical time periods;

and analyzing the historical region center and the environment data set through the target strategy model, and outputting to obtain the candidate center.

4. The method of claim 2, wherein the target strategy model comprises an actor network in a reinforcement learning architecture, the actor network being used for center prediction, the actor network comprising an encoder and a decoder in a codec architecture;

the inputting the environment data set into a target strategy model and outputting to obtain the candidate center comprises:

determining a historical region center of the candidate region range based on the environment data set, wherein the historical region center comprises a target type entity which is used as a divided region center in a historical time period;

inputting the historical region center and the environment data set into the encoder of the actor network, and outputting to obtain an encoding feature and a center feature set;

decoding and analyzing the coding characteristics through the decoder to obtain decoding characteristics;

and obtaining the candidate center based on the decoding characteristics and the center characteristic set.

5. The method of claim 2, wherein the target strategy model comprises an actor network in a reinforcement learning architecture, the actor network being used for center prediction;

before inputting the environment data set into a target policy model and outputting to obtain the candidate center, the method further includes:

acquiring a sample data set, wherein the sample data set is used for indicating the sample position relationship and the sample service distribution corresponding to at least two sample type entities;

predicting the sample data set through the actor network to obtain sample candidate centers of the at least two sample type entities;

evaluating the sample candidate center through a critic network in the reinforcement learning architecture to obtain an expected deviation, wherein the expected deviation is used for indicating the prediction effect of the sample candidate center;

and training the actor network based on the expected deviation to obtain the target strategy model.

6. The method of claim 5, wherein said evaluating said prediction center through a network of critics for expected deviation comprises:

substituting the sample candidate centers into a preset expectation function to obtain first expectation values corresponding to the sample candidate centers, wherein the first expectation values are used for indicating the aggregation relation among the sample candidate centers;

performing expected analysis on the sample data set through the critic network to obtain a second expected value corresponding to the sample data set, wherein the second expected value is used for indicating a pre-estimated aggregation relation corresponding to the sample data set;

the expected deviation is derived based on the first expected value and the second expected value.

7. The method of claim 6, wherein said deriving the expected deviation based on the first expected value and the second expected value comprises:

determining a gradient relation corresponding to the first expected value and the second expected value;

determining a gradient relationship between the first expected value and the second expected value as the expected deviation.

8. The method of claim 6, wherein said performing a desired analysis on the sample data set via the critic network to obtain a second desired value for the sample data set comprises:

inputting the sample data set into a candidate pre-estimation model, wherein the candidate pre-estimation model comprises the critic network and is a machine learning model obtained through pre-training;

and performing expected prediction on the sample data set through the critic network, and outputting to obtain the second expected value.

9. The method of claim 8, further comprising:

and adjusting model parameters of the candidate prediction model based on the first expectation value and the second expectation value to obtain an expected prediction model, wherein the expected prediction model is used for determining the second expectation value.

10. The method according to any one of claims 1 to 9, wherein the performing the area division on the candidate area range based on the candidate center to obtain the area planning result of the at least two target type entities in the candidate area range comprises:

carrying out mass center positioning on the candidate center to obtain a region center corresponding to the candidate region range;

determining entity clustering results corresponding to the at least two target type entities in the candidate region range based on the region center;

and taking the entity clustering result as the area planning result.

11. An area planning apparatus, the apparatus comprising:

12. A computer device comprising a processor and a memory, the memory having stored therein at least one instruction, at least one program, a set of codes, or a set of instructions, which is loaded and executed by the processor to implement a region planning method according to any one of claims 1 to 10.

13. A computer readable storage medium having stored therein at least one instruction, at least one program, a set of codes, or a set of instructions, which is loaded and executed by a processor to implement a method of area planning as claimed in any of claims 1 to 10.

14. A computer program product comprising a computer program or instructions which, when executed by a processor, implements the area planning method according to any one of claims 1 to 10.