CN109598430B

CN109598430B - Distribution range generation method, distribution range generation device, electronic equipment and storage medium

Info

Publication number: CN109598430B
Application number: CN201811434820.6A
Authority: CN
Inventors: 汤浅伟; 李淳敏
Original assignee: Rajax Network Technology Co Ltd
Current assignee: Rajax Network Technology Co Ltd
Priority date: 2018-11-28
Filing date: 2018-11-28
Publication date: 2021-01-08
Anticipated expiration: 2038-11-28
Also published as: CN109598430A

Abstract

The embodiment of the invention relates to a distribution range generation method, which comprises the following steps: the method comprises the steps of obtaining order characteristic parameters of all distribution units in a first distribution range corresponding to a target object, wherein the target object is located in an initial distribution unit in the first distribution range; taking the initial distribution unit as a center, and obtaining an expanded distribution unit from the distribution unit according to the distribution range model and the order characteristic parameters of the distribution unit; a second delivery range of the target object is generated based on the initial delivery unit and the expanded delivery unit. According to the invention, the intelligent distribution range optimized by the target object is automatically generated by constructing the distribution range model, the traditional method for confirming the distribution range by personal experience, common knowledge, geographic information around the target object and the conventional order analysis is replaced, time and labor are saved, and the distribution efficiency is improved.

Description

Distribution range generation method, distribution range generation device, electronic equipment and storage medium

Technical Field

The present invention relates to the field of communications technologies, and in particular, to a method and an apparatus for generating a distribution range, an electronic device, and a storage medium.

Background

The distribution range is very important for take-away merchants, the requirements for the distribution range are different for different types of merchants, and the current method for determining the distribution range of a merchant mainly comprises the steps of generating a vector according to each dimension value in the attribute information of the merchant, performing cluster analysis on the generated vector, and defining the distribution range corresponding to the merchant. Although the method can determine the distribution range of the merchant according to the defined grid, the method does not describe how to optimize and adjust the distribution range based on the existing distribution range.

Disclosure of Invention

An object of embodiments of the present invention is to provide a method and an apparatus for generating a distribution range, an electronic device, and a storage medium, which automatically define a preferred distribution range for a target object, reduce labor, and improve distribution efficiency.

To solve the above technical problem, an embodiment of the present invention provides a delivery range generating method, including: the method comprises the steps of obtaining order characteristic parameters of all distribution units in a first distribution range corresponding to a target object, wherein the target object is located in an initial distribution unit in the first distribution range; taking the initial distribution unit as a center, and obtaining an expanded distribution unit from the distribution unit according to the distribution range model and the order characteristic parameters of the distribution unit; a second delivery range of the target object is generated based on the initial delivery unit and the expanded delivery unit.

An embodiment of the present invention further provides a distribution range generation apparatus, including: the data acquisition module is used for acquiring order characteristic parameters of all distribution units in a first distribution range corresponding to a target object, wherein the target object is located in the initial distribution unit in the first distribution range; the grid expansion module is used for obtaining expanded distribution units from the distribution units by taking the initial distribution units as centers according to the distribution range model and the order characteristic parameters of the distribution units; and a range generation module for generating a second delivery range of the target object according to the initial delivery unit and the expanded delivery unit.

The embodiment of the invention also discloses electronic equipment, which comprises at least one processor; a memory communicatively coupled to the at least one processor; the communication component is respectively in communication connection with the processor and the memory and receives and transmits data under the control of the processor; wherein the memory stores instructions executable by the at least one processor to implement: the method comprises the steps of obtaining order characteristic parameters of all distribution units in a first distribution range corresponding to a target object, wherein the target object is located in an initial distribution unit in the first distribution range; taking the initial distribution unit as a center, and obtaining an expanded distribution unit from the distribution unit according to the distribution range model and the order characteristic parameters of the distribution unit; a second delivery range of the target object is generated based on the initial delivery unit and the expanded delivery unit.

The embodiment of the invention also discloses a nonvolatile storage medium for storing a computer readable program, and the computer readable program is used for causing a computer to execute the distribution range generation method.

Compared with the prior art, the implementation mode of the invention has the main differences and the effects that: according to the invention, by constructing the distribution range model and continuously training the distribution range model, when the data information of the target object is input, the optimized intelligent distribution range of the target object can be automatically generated, and the traditional method of confirming the distribution range by personal experience, common knowledge, geographic information around the target object and the conventional order analysis is replaced.

In addition, the method of the embodiment of the present invention further includes: constructing a distribution range model, wherein the distribution range model comprises a state value function, and the state value function takes the order characteristic parameters of the distribution units as parameters; and training a distribution range model according to a machine learning algorithm by using the historical order characteristic parameters of the object in the business circle where the target object is located.

In addition, the machine learning algorithm is a reinforcement learning algorithm, and taking the initial distribution unit as a center, and obtaining the expanded distribution unit from the distribution units according to the distribution range model specifically comprises the following steps: inputting order characteristic parameters of each distribution unit in a first distribution range of the target object in the state value function to obtain the value of the distribution unit; obtaining a complete state value matrix of the distribution units in the first distribution range according to the distribution range model and the values of the distribution units; and taking the initial distribution unit as a center, and judging the distribution unit according to the complete state value matrix to obtain the expanded distribution unit. The intelligent delivery range is generated through a reinforcement learning algorithm, the method is suitable for a simple scene determined by the initial delivery range of the target object, the calculation is accurate, and the most optimized intelligent delivery range can be obtained.

In addition, according to the distribution range model and the scores of the distribution units, a complete state value matrix of the distribution units in the first distribution range is obtained, which specifically comprises the following steps: obtaining an initial state value matrix according to the distribution range model and the values of the initial distribution units; taking the initial distribution unit and one or more distribution units adjacent to the initial distribution unit as a second distribution unit set, and obtaining a second state value matrix according to the distribution range model and the value sum of the distribution units of the second distribution unit set; taking the second distribution unit set and one or more distribution units adjacent to the second distribution unit set as a third distribution unit set, and obtaining a third state value matrix according to the distribution range model and the value sum of the distribution units of the third distribution unit set; and repeating the steps until the distribution units in the first distribution range are traversed to obtain a complete state value matrix. And finally, forming a complete state value matrix by taking the set as the state and the score sum of the set as a feedback value, so as to conveniently train the initial distribution range model through a reinforcement learning algorithm.

In addition, with the initial distribution unit as a center, the distribution unit is judged according to the complete state value matrix to obtain an expanded distribution unit, which specifically comprises: and inquiring the score and the highest distribution unit set in the complete state value matrix, and taking the distribution units except the initial distribution unit in the score and the highest distribution unit set as expansion distribution units. And selecting the distribution unit set with the highest score and the highest score as an intelligent distribution range, and obtaining an optimal distribution range model through training.

In addition, the machine learning algorithm is a deep learning algorithm, the initial distribution unit is used as a center, and the expansion distribution unit order characteristic parameters are obtained from the distribution unit according to the distribution range model, and the method specifically comprises the following steps: inputting order characteristic parameters of each distribution unit in a first distribution range of the target object in the state value function to obtain the value of the distribution unit; and taking the initial distribution unit as a center, and judging the distribution unit according to the score of the distribution unit to obtain the expanded distribution unit. The intelligent distribution range is generated through the depth learning algorithms, and the method can be suitable for complex application scenes with insufficient calculation power of the reinforcement learning algorithm, so that the optimal intelligent distribution range is obtained.

In addition, with the initial distribution unit as the center, the distribution unit is judged according to the score of each distribution unit to obtain an expanded distribution unit, which specifically comprises: selecting a distribution unit with the highest value in distribution units adjacent to the initial distribution unit as a first expanded distribution unit, wherein the initial distribution unit and the first expanded distribution unit jointly form a first expanded range; selecting the distribution unit with the highest value in the distribution units adjacent to the first expansion range as a second expansion distribution unit, wherein the first expansion range and the second expansion distribution unit jointly form a second expansion range; in the same way, obtaining the nth expanded distribution unit and the nth expanded range until the requirement of the second distribution range is met, wherein n is a positive integer; the first expansion distribution unit and the second expansion distribution unit are used as expansion distribution units. When the distribution units are expanded outwards, the distribution unit with the highest score is selected as the expanded distribution unit each time, so that the distribution unit set with the highest total score can be obtained conveniently, and the optimal intelligent distribution range can be obtained.

In addition, if the obtained (n-a) th expanded distribution unit comprises a plurality of (n-a) th expanded distribution units, the n-th expanded range with the highest total score is selected as the distribution range of the target object, wherein a is a positive integer which is larger than 1 and smaller than n.

In addition, the requirements for satisfying the second distribution range are specifically: the area of the second distribution range does not exceed a preset area threshold value, or the farthest distance between the initial distribution unit and the outermost expanded distribution unit does not exceed a preset distance threshold value. The purpose of setting the threshold is to automatically stop the expansion of the dispensing range in a proper area, and avoid the problem of insufficient computing power caused by infinite expansion.

In addition, the order characteristic parameter of the delivery unit is obtained by the following steps: extracting order data of a first distribution range from a database, wherein the order data comprises a waybill width table, an order width table, a target object width table and a POI (point of interest) interface of an electronic map; and screening the order data to obtain the order characteristic parameters of each delivery unit in the first delivery range. The order characteristic parameters are distribution data having the most influence on distribution quality or distribution efficiency of the target object, and are calculated according to the order characteristic parameters, so that the optimal intelligent distribution range of the target object is obtained.

The delivery range generation method further includes: acquiring a geo-fence comprising preset keywords; judging the distribution units in the second distribution range according to the geo-fence to obtain invalid distribution units, wherein the invalid sub-area units are located in the geo-fence or comprise the geo-fence; the invalid delivery units are deleted in the second delivery range. And the problem that the distribution efficiency is influenced by adding an invalid area in a distribution range is avoided.

Drawings

One or more embodiments are illustrated by way of example in the accompanying drawings, which correspond to the figures in which like reference numerals refer to similar elements and which are not to scale unless otherwise specified.

Fig. 1 is a flowchart of a delivery range generation method according to a first embodiment of the present invention;

FIG. 2-1 is a flowchart of a delivery range generation method according to a second embodiment of the present invention;

fig. 2-2 is a schematic diagram illustrating the generation of a distribution range of an object K1 in an M business district according to a second embodiment of the present invention;

FIGS. 2-3 are expanded delivery unit acquisition flow diagrams according to a second embodiment of the present invention;

FIGS. 2-4 are flow diagrams of a complete state value matrix acquisition in accordance with a second embodiment of the present invention;

FIGS. 2-5 are schematic diagrams illustrating the generation of a delivery range of a target object according to a second embodiment of the present invention;

FIG. 3-1 is a flowchart of a delivery range generation method according to a third embodiment of the present invention;

fig. 3-2 is a schematic diagram illustrating the generation of a distribution range of an object K1 in an M business district according to a third embodiment of the present invention;

3-3 are expanded delivery unit acquisition flow diagrams according to a third embodiment of the present invention;

FIGS. 3-4 are schematic diagrams of the generation of a delivery range of a target object according to a third embodiment of the present invention;

FIG. 4-1 is a flowchart of a delivery range generation method according to a fourth embodiment of the present invention;

fig. 4-2 is a schematic diagram of an intelligent delivery range generated by the delivery range generation method according to the fourth embodiment of the present invention;

FIG. 5-1 is a schematic diagram of a delivery range generating apparatus according to a fifth embodiment of the present invention;

FIG. 5-2 is a schematic view of a lattice expansion module according to a fifth embodiment of the present invention;

FIGS. 5-3 are schematic views of another lattice expansion module according to a fifth embodiment of the present invention;

fig. 6 is a schematic structural diagram of an electronic device according to a sixth embodiment of the invention.

Detailed Description

In order to make the objects, technical solutions and advantages of the embodiments of the present invention more apparent, embodiments of the present invention will be described in detail below with reference to the accompanying drawings. However, it will be appreciated by those of ordinary skill in the art that numerous technical details are set forth in order to provide a better understanding of the present application in various embodiments of the present invention. However, the technical solution claimed in the present application can be implemented without these technical details and various changes and modifications based on the following embodiments.

The first embodiment of the present invention relates to a distribution range generating method, and the present embodiment may be applied to a terminal side, such as a terminal device such as a mobile phone and a tablet computer, and may also be applied to a server on a network side. As shown in fig. 1, a delivery range generation method according to an embodiment of the present invention includes:

103, acquiring order characteristic parameters of each delivery unit in a first delivery range corresponding to a target object, wherein the target object is located in an initial delivery unit in the first delivery range;

step 104, taking the initial distribution unit as a center, and obtaining an expanded distribution unit from the distribution unit according to the distribution range model and the order characteristic parameters of the distribution unit;

step 105, generating a second delivery range of the target object according to the initial delivery unit and the expanded delivery unit.

Specifically, in step 103, a first delivery range (i.e., an initial delivery range) of a target object is formed by radiating a certain distance outwards with the position of the target object as a center, where the target object may be a delivery site or a business (e.g., a restaurant, a dessert store, a pharmacy store), etc.; for example, a geographical range is formed by radiating a certain distance outwards with longitude and latitude of a certain business as a center, and the geographical range is screened, for example, a regularly-shaped or irregularly-shaped geographical range formed after eliminating areas such as expressways, bridges and the like which cannot be used as order arrival points is used as an initial distribution range (i.e., a first distribution range) of the business. The first distribution range is then divided into a plurality of distribution units (each distribution unit may be a shape obtained by equally dividing the first area into uniform sizes, or may be obtained by dividing the first area according to a certain rule, such as dividing the first area according to a geographical location), and it should be specifically noted that each distribution unit may be a grid, such as a square grid of 140 meters × 140 meters, or a grid of other shapes and sizes.

In this embodiment, the order characteristic parameters are obtained according to the following method: extracting order data of a first distribution range from a database; and screening the order data to obtain the order characteristic parameters of each distribution unit.

Explaining by taking a target object as a merchant, collecting order data related to the merchant in the past in a database, and extracting order data of a first delivery range from the database, wherein the order data comprises a shipping bill wide table (a database table consisting of merchant coordinates, meal taking time, meal delivery time and the like, the wide table refers to a database table with more fields, and generally refers to a database table with related indexes, dimensions and attributes related to business topics), an order wide table (a database table consisting of restaurant categories, meal categories and the like of historical orders), a merchant wide table (a database table consisting of meal categories, merchant coordinates, merchant labels and the like), a POI interface of an electronic map (an Apache POI is an open source item of an Apache software foundation, and the POI provides an API for Java program to read and write documents in a Microsoft Office format); and after the order data are sorted, summarized and screened, the order characteristic parameters of all the distribution units in the first distribution range are obtained. The order characteristic parameters can be characteristic data which has influence on the delivery quality and efficiency of the order, and are convenient to calculate according to the order characteristic parameters subsequently to obtain an optimal delivery range.

In practical applications, the order characteristic parameters obtained after processing the order distribution data such as the waybill width table, the order width table, the merchant width table, the POI interface of the electronic map and the like may include a merchant ID, an ID of a distribution unit where the waybill is located, a walking distance from the merchant to the distribution unit, an average distribution time, an average AOI meal delivery time (a time when a rider arrives at a destination and goes upstairs or enters a cell to find a corresponding resident), an average meal delivery time, an effective completion order, a timeout order within a preset time, and the like. Take the take-away industry as an example, wherein an order for a single-finger rider to successfully deliver a meal to a user is effectively fulfilled; the overtime order is an order form of which the food delivery is overtime due to the fact that the rider does not deliver the food to the user within the specified time; and the order of the overtime meal delivery of the rider is given by the overtime single finger in the preset time length.

Further, the order distribution data is screened to obtain order characteristic parameters of each distribution unit in the first distribution range, specifically, orders of the merchants falling into each distribution unit (grid) are screened out through the merchant IDs and the IDs of the distribution units where the freight notes are located in the order distribution data, and then, relevant order data is screened out as the order characteristic parameters according to needs.

The expanded delivery unit of the present embodiment is another delivery unit in the first delivery range of the target object, which is added to the initial delivery unit to form the second delivery range, with the initial delivery unit in which the target object is located as the center, in order to obtain the more optimized second delivery range of the target object.

In step 104 of this embodiment, the expanded distribution units are obtained from the distribution units based on a distribution range model with the initial distribution unit as the center, as shown in fig. 1, wherein the distribution range model is obtained by the following method:

step 101, constructing an initial distribution range model, wherein the initial distribution range model comprises a state value function, and the state value function takes the order characteristic parameters of the distribution units as parameters;

and 102, training the initial distribution range model according to a machine learning algorithm by using the historical order characteristic parameters of the object of the business district where the target object is located, so as to obtain an optimized distribution range model.

The machine learning algorithm comprises a reinforcement learning algorithm or a deep learning algorithm, wherein the reinforcement learning is a self-correction and feedback machine learning mechanism, and a machine has self-learning and self-thinking capabilities, such as a Q-learning algorithm and an agent algorithm; the deep learning mainly focuses on feature extraction, and accurate feature extraction and identification are achieved by training a model through a large amount of labeled data.

In the prior art, a merchant distribution range is generally divided according to personal experience, common knowledge, geographic information around the merchant and past order analysis, so that the divided distribution range mainly depends on artificial subjective awareness and is likely to divide potential customers outside the distribution range to miss consumers and cause losses of the merchant, therefore, in order to solve the problem, a more optimal and more accurate distribution range is provided for the merchant, in the embodiment, a distribution range model is built and optimized, a large geographic range is divided into each small distribution unit, an expanded distribution unit of a target object is obtained by obtaining order characteristic parameters of the distribution unit in a first distribution range of the target object and according to the trained distribution range model and the order characteristic parameters of the distribution unit, and more expanded distribution units conforming to an expansion strategy are gradually added by taking an initial distribution unit as a starting point, and generating a second distribution range according to the initial distribution unit and the expanded distribution unit together, thereby automatically obtaining the optimized target object intelligent distribution range.

It should be noted that the distribution range model of the present embodiment may be determined according to the requirement of the target object, and is specifically embodied in different weight values of each parameter of the state value function in the distribution range model, for example, when a certain merchant wants more orders, the order quantity parameter in the state value function is given a larger weight value; if a merchant is interested in higher delivery quality and not order quantity, then the timeout quantity parameter in the state value function is given a higher weight value.

A second embodiment of the present invention discloses a distribution range generating method, as shown in fig. 2-1, including:

step 201, obtaining order characteristic parameters of each delivery unit in a first delivery range corresponding to a target object, wherein the target object is located in an initial delivery unit in the first delivery range;

step 202, inputting order characteristic parameters of each distribution unit in a first distribution range of the target object in a state value function to obtain the value of the distribution unit;

step 203, obtaining a complete state value matrix of the distribution units in the first distribution range according to the distribution range model and the scores of the distribution units;

step 204, taking the initial distribution unit as a center, and judging the distribution unit according to the complete state value matrix to obtain an expanded distribution unit;

in step 205, a second delivery range of the target object is generated based on the initial delivery unit and the expanded delivery unit.

In this embodiment, first, according to a reinforcement learning algorithm, a constructed distribution range model is trained with historical order characteristic parameters of one historical time period of all objects in a business district where a target object is located. Reinforcement Learning is a mature technology in the field of artificial intelligence, and Q-Learning is taken as an example for a brief explanation of the present embodiment.

The goal of Q-Learning is to learn a policy that tells the agent what action to take under what circumstances. It does not need environment model, and can deal with the problems of random conversion and reward without adjustment. For any Finite Markov Decision Process (FMDP), Q-Learning finds a strategy that is optimal because the expected value of the total reward for all successive steps is the maximum achievable, starting from the current state. Given an infinite exploration time and a partially random strategy, Q-Learning may determine the best action selection strategy for any given FMDP. "Q" is used to name the function that returns the reward for providing reinforcement, and can be said to represent the "quality" of the action taken at a given state. Q-learning comprises an agent, a set of states, and a set of actions. By performing an action, the agent' S state changes S1 → S2(Aa), and performing this action will result in the agent getting a numerical reward, and Q-learning uses a feedback function to calculate the reward value of (state, action), in short, Q-learning gets the best (highest quality) series of actions in the change process by changing state again and again, and records the score relationship between state and action in Q-Table.

Specifically, by using a Q-learning reinforcement learning algorithm, in the embodiment, taking an object K1 in an M business circle where a target object KI is located as an example, as shown in fig. 2-2, first, historical order characteristic parameters of each distribution unit in the M business circle in a historical time period are input in a state value function, so as to obtain a score of each distribution unit in the M business circle; taking the initial distribution unit A1 where the object K1 is located as an initial state, taking the score of the initial distribution unit A1 where the object K1 is located as a feedback value of the initial state, and obtaining an initial state value matrix according to the initial state and the feedback value of the initial state; a second distribution unit set formed by the initial distribution unit A1 and 4 distribution units B1-B4 adjacent to the initial distribution unit A1 is in a second state, namely the second distribution unit set can have four distribution unit sets of [ A1, B1 ], [ A1, B2 ], [ A1, B3 ], [ A1 and B4 ], and a second state value matrix is obtained according to the scores of the four distribution unit sets and feedback values of the second state; with a collection of four delivery units [ A1, B1 ], [ A1, B2 ], [ A1, B3 ], [ A1, B4 ], and a third distribution unit set of eight distribution units C1-C8 adjacent to the four distribution unit sets in a third state, that is, the third set of dispensing units may have twelve dispensing units [ a1, B1, C1 ], [ a1, B1, C2 ], [ a1, B1, C8 ], [ a1, B2, C3 ], [ a1, B2, C2 ], [ a1, B2, C3 ], [ a1, B2, C4 ], [ a1, B1, C1 ], [ a1, B1, C1 ], C1, [ a1, B1, C1 ], obtaining a second state value matrix according to the third state and the feedback value of the third state by taking the scores of the twelve distribution unit sets as the feedback value of the third state; and by analogy, all distribution units of the M business circles are guided to traverse, and finally a complete state matrix a corresponding to the object K1 is obtained.

According to the Q-learning reinforcement learning algorithm, the distribution unit set with the highest score is searched in the complete state matrix a corresponding to the object K1, for example, the score sum corresponding to the distribution unit set [ a1, B3, C6, D5 · Z26 ] is 198, which is the highest value among all the state feedback values in the complete state matrix a, so that the state of the distribution unit set [ a1, B3, C6, D5 · Z26 ] is used as the optimal distribution range of the object a1, and each distribution unit B3, C6, D5 · Z26 ] included in the distribution unit set [ a1, B3, C6, D5 · Z26 ] is used as an expansion distribution unit of the target object K1.

Then, taking another object K2 in the M business district where the target object is located as an example, the expanded distribution unit of the object K2 and the optimal distribution range of the object K2 are obtained according to the above method. In the machine learning process, the expanded delivery unit and the optimal delivery range of an object in the business district with M business circles are obtained according to the same rule every time, and other rules are abandoned, so that the historical order characteristic parameters of all the delivery units in the business district where the target object is located in a historical time period are utilized to conduct continuous optimization training on the delivery range model, and when the order characteristic parameters of the target object are input into the delivery range model again, the expanded delivery unit and the optimal delivery range corresponding to the target object can be generated according to the same rule.

That is, after the training is completed, the expanded delivery unit is obtained from the delivery unit according to the delivery range model with the initial delivery unit where the target object is located as the center, as shown in fig. 2 to 3, which specifically includes the following steps:

step 2021, inputting order characteristic parameters of each distribution unit in the first distribution range of the target object in the state value function to obtain a value of the distribution unit;

step 2022, obtaining a complete state value matrix of the distribution unit in the first distribution range according to the distribution range model and the score of the distribution unit;

step 2023, taking the initial distribution unit as a center, and determining the distribution unit according to the complete state value matrix to obtain an expanded distribution unit.

Further, in step 2022, according to the distribution range model and the scores of the distribution units, a complete state value matrix of the distribution units in the first distribution range is obtained, as shown in fig. 2 to 4, which specifically includes:

step 2022a, obtaining an initial state value matrix according to the distribution range model and the scores of the initial distribution units;

step 2022b, using the initial distribution unit and one or more distribution units adjacent to the initial distribution unit as a second distribution unit set, and obtaining a second state value matrix according to the distribution range model and the score sum of the distribution units of the second distribution unit set;

step 2022c, taking the second delivery unit set and one or more delivery units adjacent to the second delivery unit set as a third delivery unit set, and obtaining a third state value matrix according to the delivery range model and the score sum of the delivery units of the third delivery unit set;

and 2022d, repeating the steps until the distribution units in the first distribution range are traversed to obtain a complete state value matrix.

For example, as shown in fig. 2 to 5, the four distribution units around the initial distribution unit P where the target object is located are P1(-2), P2(4), P3(-1), and P4(1) (the numbers in parentheses indicate the scores of the distribution units), the initial distribution unit and the adjacent distribution units are respectively configured into a second distribution unit set according to step 2022b, the status values (the sum of the scores of the second distribution unit sets) of which are shown as S1 in table 1 below, where Q1 in S1 indicates one second distribution unit set composed of the initial distribution unit P and the distribution unit P1, and the status values (the sum of the scores of the three distribution units) of which Q (S1, Q1) is-2; q2 in S1 represents another second delivery unit set composed of the initial delivery unit P and the delivery unit P2, and the state value Q (S2, Q2) is 4 · · and so on, and the state values of the initial delivery unit and the adjacent four delivery units are all listed in the table; each second distribution unit set is added with eight adjacent distribution units of P5(3), P6(15), P7(6), P8(-3), P9(-6), P10(10), P11(-3) and P12 (2); the second delivery unit sets and the neighboring delivery units form a third delivery unit set according to the step 2022c, and the status (sum of scores of the third delivery unit sets) is shown as S2 in table 1 below, where Q1 in S2 represents the third delivery unit set formed by the initial delivery unit P, the delivery unit P1, and the delivery unit P6, and the status value Q (S2, Q1) is 13; q2 in S2 indicates another third delivery unit set consisting of the initial delivery unit P, the delivery unit P1, and the delivery unit P7, and the state value Q (S2, Q2) is 4 · · and so on, and the state values of the second delivery unit set and the eight adjacent delivery units are listed in the table; until the distribution range requirement is met, the nth subregion set is obtained, and the state values (the scores of the nth subregion sets) of the nth subregion set are shown as Sn in the following table 1, wherein n is a positive integer.

As can be seen from table 1, since the score sum Q (S2, Q1) of the third delivery unit set constituted by the initial delivery unit P, the delivery unit P1, and the delivery unit P6 in common is the largest, the delivery unit P1 and the delivery unit P6 are added as the expanded delivery units to the initial delivery unit, and finally the delivery unit set constituted by the initial delivery unit P, the expanded delivery unit P1, and the expanded delivery unit P6 is the smart delivery range of the target object.

TABLE 1

	q1	q2	q3	q4	q5	q6	···
								S1	-2	4	-1	1
S2	13	4	2	-3	-1	1
								S3	4	-1	5	0	-2	3
S4	-3	3	-7	9	-4	0
								S5	2	10	1	-2	3	5
···	···	···	···	···	···	···	···
								Sn	N1	N2	N3	N4	N5	N6	···

It should be noted that, in the present embodiment, the order feature parameters may include an optimal amount of orders, an overtime amount of orders within a preset duration, and an overtime amount of orders outside the preset duration. Specifically, the optimal matching list is an order which is normally delivered, and the more the optimal matching list is, the higher the order delivery quality of the delivery unit is; the more overtime orders within the preset duration, the lower the order distribution efficiency of the distribution unit is; the number of orders beyond the preset duration is less, the order distribution efficiency of the distribution unit is higher, the three order characteristic parameters are used as the most concerned factors of the target object and have main influence on the distribution quality, the distribution efficiency and the like of order distribution, and therefore when the order distribution range is expanded, the number of best orders, the number of orders within the preset duration, the number of orders beyond the preset duration and the like carried by each distribution unit are used as training parameters to train a distribution range model.

Further, the state value function in this embodiment is a feedback function, and specifically, the value of the corresponding delivery unit is obtained by inputting 3 order characteristic parameters, i.e., the optimal quantity, the timeout quantity within the preset time length, and the timeout quantity outside the preset time length, into the feedback function, and the value of each delivery unit is recorded.

In the embodiment, the initial distribution range model is trained through a reinforcement learning algorithm, a set is used as a state, the value sum of the set is used as a feedback value, a complete state value matrix is finally formed, the value sum of the highest distribution unit set is selected as an intelligent distribution range, and the optimal distribution range can be obtained through training.

A third embodiment of the present invention discloses a distribution range generating method, as shown in fig. 3-1, including:

step 301, obtaining order characteristic parameters of each delivery unit in a first delivery range corresponding to a target object, wherein the target object is located in an initial delivery unit in the first delivery range;

step 302, inputting order characteristic parameters of each distribution unit in a first distribution range of the target object in the state value function to obtain the value of the distribution unit;

step 303, taking the initial distribution unit as a center, and judging the distribution unit according to the score of the distribution unit to obtain an expanded distribution unit;

in step 304, a second delivery range of the target object is generated based on the initial delivery unit and the expanded delivery unit.

In the embodiment, firstly, according to a deep learning algorithm, a constructed distribution range model is trained by using historical order characteristic parameters of one historical time period of all objects in a business circle where a target object is located. The training process is illustrated below:

for example, also taking an object K1 in an M business circle where the target object is located as an example, the object K1 is located in the initial distribution unit a1, as shown in fig. 3-2, first, historical order characteristic parameters of each distribution unit in the M business circle in a historical time period are input in a state value function, and a score of each distribution unit in the M business circle is obtained; the access method comprises the steps that an initial distribution unit A1 is taken as a center, a distribution unit (for example, B2) with the highest value in four distribution units B1-B4 adjacent to the initial distribution unit A1 is selected as a first expansion distribution unit, and the initial distribution unit A1 and the first expansion distribution unit B2 jointly form a first expansion range, namely a distribution unit set (A1, B2) shown in the figure; then, the distribution unit (for example, C4) with the highest score among the six distribution units B1, C2, C3, C4, B3 and B4 adjacent to the first expansion range is selected as a second expansion distribution unit, and the first expansion range and the second expansion distribution unit C4 jointly form a second expansion range, namely a distribution unit set [ A1, B2 and C4 ] shown in the figure; then, the distribution unit with the highest value among the seven distribution units adjacent to the second expanded range is selected as the third expanded distribution unit, and so on until the requirement of the second distribution range is met, the nth expanded distribution unit (for example, Zm) is obtained, the nth expanded range (n is a positive integer) is formed by the (n-1) th expanded range and the nth expanded distribution unit together, then, the nth expanded range is used as the optimal intelligent distribution range corresponding to the object K1 meeting the requirement of the second distribution range, and the distribution units B2 and C4 · Zm outside the initial distribution unit a1 included in the nth expanded range are used as the expanded distribution units.

That is, after the training is completed, the expanded delivery unit is obtained from the delivery unit according to the delivery range model with the initial delivery unit where the target object is located as the center, as shown in fig. 3-3, which specifically includes the following steps:

3031, selecting a distribution unit with the highest value in distribution units adjacent to the initial distribution unit as a first expansion distribution unit, wherein the initial distribution unit and the first expansion distribution unit jointly form a first expansion range;

step 3032, selecting the distribution unit with the highest value in the distribution units adjacent to the first expansion range as a second expansion distribution unit, wherein the first expansion range and the second expansion distribution unit jointly form a second expansion range;

step 3033, repeating the steps until the requirement of the second distribution range is met, and obtaining the nth expansion distribution unit and the nth expansion range, wherein n is a positive integer;

step 3034, the first expansion distribution unit and the second expansion distribution unit · nth expansion distribution unit are used as expansion distribution units.

Specifically, as shown in fig. 3-4, taking target object B as an example, and taking distribution unit B0 where target object B is located as a starting point, the scores (part numbers not identified in the figure) of four distribution units B1, B2, B3, and B4 around it are compared, the distribution unit with the highest score is selected as the first expanded distribution unit of target object B, assuming that the score of distribution unit B3 is highest, distribution unit B3 is used as the first expanded distribution unit, then initial distribution unit B0 and distribution unit B3 together form the first expanded range, taking the first expanded range as a starting point, the scores (part numbers not identified in the figure) of six adjacent distribution units (B5, B6, B7, B1, B2, and B4) around distribution unit B0 and distribution unit B3 are compared, assuming that distribution unit B7 is the highest score of distribution unit B7 is used as the second expanded distribution unit of target object B68568, and then, forming a second expanded range by using the area formed by the initial distribution unit B0, the distribution unit B3 and the distribution unit B7 together, comparing scores … … of a plurality of distribution units adjacent to the second expanded range by taking the second expanded range as a starting point, repeating the steps, and so on, and continuously adjusting and updating the distribution range until the requirement of the second distribution range is met so as to generate the optimal intelligent distribution range of the target object B.

It should be further noted that, if the (n-a) th expanded distribution unit includes a plurality of (n-a) th expanded distribution units, the nth expanded range with the highest total score is selected as the distribution range of the target object, where a is a positive integer greater than 1 and less than n. For example, in the above preferred embodiment taking the target object B as an example, if n is 3 and a is 1, then n-a is 2, and assuming that the scores of B6 and B7 are the same, i.e., two second expanded distribution units are obtained at the same time, then the total score of a third expanded range including B6 and B7 needs to be compared, for example, the third expanded range including B6 is composed of the initial distribution unit B0, the distribution unit B3, the distribution unit B6, and the distribution unit B11, and the total score is 15; the third expanded range including B7 is composed of the initial distribution unit B0, the distribution unit B3, the distribution unit B7, and the distribution unit B12, and the total score thereof is 10, and the former is selected as the second distribution range of the target object B, that is, the distribution unit B3, the distribution unit B6, and the distribution unit B11 are selected as the expanded distribution units.

The requirement for satisfying the distribution range may be that the area of the distribution range does not exceed an area threshold, or that the farthest distance between the initial distribution unit and the outermost expanded distribution unit does not exceed a distance threshold. The purpose of setting the threshold is to automatically stop the expansion of the distribution range in a proper area, and avoid adding distribution units with low distribution quality or low efficiency in the distribution range.

For example, the area threshold Z is set to 10 square meters, and when the total area of the second delivery range reaches 10 square meters, the expansion is stopped; or the distance threshold is set to be 5 km, and when the farthest distance between the initial distribution unit and the outermost expanded distribution unit reaches 5 km, the expansion should be stopped. In this embodiment, there are two preset thresholds, namely, an area threshold and a distance threshold, and in practical application, the corresponding preset threshold may be selected according to the specific situation and the need of the target object, so as to ensure that the generated optimal distribution range meets the demand of the target object.

In the present embodiment, the order characteristic parameters include a total order amount, an order amount of which the delivery time exceeds a preset time period (the preset time period may be 30 minutes or 60 minutes), and a predicted order amount of other target objects of the same brand. Specifically, in some scenarios, the larger the total order volume, the more the merchant tends to select the delivery unit; the more the order quantity of the distribution time exceeds the preset duration, the lower the order distribution efficiency of the distribution unit is; the more other target objects of the same brand predict the order quantity, the more the merchant tends to select the delivery unit, and the three order characteristic parameters have main influence on predicting the order quantity, the order delivery quality and the order delivery efficiency of the delivery unit, so the three order characteristic parameters are considered as preferable factors when expanding the delivery range.

When the distribution range is expanded outwards again in the embodiment, the distribution unit with the highest score is selected as the expansion distribution unit each time, and the expansion distribution unit automatically stops until the distribution unit is expanded to a proper range, so that the optimal intelligent distribution range can be obtained.

A fourth embodiment of the present invention provides a method for generating an intelligent distribution range, as shown in fig. 4, including:

step 401, obtaining order characteristic parameters of each delivery unit in a first delivery range corresponding to a target object, wherein the target object is located in an initial delivery unit in the first delivery range;

step 402, taking the initial distribution unit as a center, and obtaining an expanded distribution unit from the distribution unit according to the distribution range model and the order characteristic parameters of the distribution unit;

step 403, generating a second delivery range of the target object according to the initial delivery unit and the expanded delivery unit;

step 404, acquiring a geo-fence including preset keywords;

step 405, determining a distribution unit in a second distribution range according to the geofence to obtain an invalid distribution unit, wherein the invalid sub-area unit is located in the geofence or includes the geofence;

at step 406, the invalid delivery units are deleted in the second delivery scope.

In this embodiment, steps 401 to 403 are described in the above embodiments, and are not described herein. In step 404, the keywords may be rivers, lakes, seas, parks, cemeteries, etc., where no orders are present in the above areas, and therefore called invalid areas, and it is determined whether the distribution units and each sub-area unit in the target area in the first distribution range include rivers, seas, parks, cemeteries, or whether the distribution units and each sub-area unit in the target area in the first distribution range are entirely located in rivers, seas, parks, cemeteries, and if so, the invalid sub-area units including the invalid areas are removed from the second distribution range, and finally the intelligent distribution range shown in fig. 4-2 is generated.

In the present embodiment, the invalid delivery unit is deleted from the second delivery range, and the filtered second delivery range is used as the delivery range of the generated target object, thereby avoiding the influence on the delivery efficiency due to the addition of the invalid area to the delivery range.

A fifth embodiment of the present invention discloses a distribution range generation device, as shown in fig. 5-1, including:

the data obtaining module 503 is configured to obtain order characteristic parameters of each delivery unit in a first delivery range corresponding to a target object, where the target object is located in an initial delivery unit in the first delivery range;

a grid expansion module 504, which takes the initial distribution unit as a center and obtains an expanded distribution unit from the distribution units according to the distribution range model and the order characteristic parameters of the distribution units;

the range generation module 505 generates a second delivery range of the target object based on the initial delivery unit and the expanded delivery unit.

In addition, the distribution range generation device according to the embodiment of the present invention further includes:

the model construction module 501 is configured to construct a distribution range model, where the distribution range model includes a state value function, and the state value function takes the order characteristic parameters of the distribution units as parameters;

the model training module 502 trains the distribution range model according to a machine learning algorithm with the historical order characteristic parameters of one historical time period of all the objects in the business district where the target object is located.

Further, the machine learning algorithm is a reinforcement learning algorithm, and as shown in fig. 5-2, the mesh expansion module 504 specifically includes:

the score calculation unit 5041 is used for inputting the order characteristic parameters of each distribution unit in the first distribution range of the target object in the state value function to obtain the score of each distribution unit;

a matrix obtaining unit 5042, configured to obtain a complete state value matrix of the distribution unit in the first distribution range according to the distribution range model and the score of the distribution unit;

the expansion determination unit 5043 determines the delivery unit based on the complete state value matrix with the initial delivery unit as the center, and obtains an expanded delivery unit.

The intelligent delivery range is generated through the reinforcement learning algorithm, the method is suitable for a simple scene determined by the initial delivery range of the target object, the calculation is accurate, and the most optimized intelligent delivery range can be obtained.

In addition, the matrix obtaining unit 5042 obtains a complete state value matrix of the distribution unit in the first distribution range according to the distribution range model and the score of the distribution unit, specifically: obtaining an initial state value matrix according to the distribution range model and the values of the initial distribution units; taking the initial distribution unit and one or more distribution units adjacent to the initial distribution unit as a second distribution unit set, and obtaining a second state value matrix according to the distribution range model and the value sum of the distribution units of the second distribution unit set; taking the second distribution unit set and one or more distribution units adjacent to the second distribution unit set as a third distribution unit set, and obtaining a third state value matrix according to the distribution range model and the value sum of the distribution units of the third distribution unit set; and repeating the steps until the distribution units in the first distribution range are traversed to obtain a complete state value matrix. And finally, forming a complete state value matrix by taking the set as the state and the score sum of the set as a feedback value, so as to conveniently train the initial distribution range model through a reinforcement learning algorithm.

In addition, the expansion determination unit 5043 takes the initial distribution unit as a center, and determines the distribution unit according to the complete state value matrix to obtain an expansion distribution unit, which specifically is: and inquiring the score and the highest distribution unit set in the complete state value matrix, and taking the distribution units except the initial distribution unit in the score and the highest distribution unit set as expansion distribution units. And selecting the distribution unit set with the highest score and the highest score as an intelligent distribution range, and obtaining an optimal distribution range model through training.

In addition, the machine learning algorithm is a deep learning algorithm, as shown in fig. 5-3, the grid expansion module 504 may further include:

the score obtaining unit 5141 is configured to input order characteristic parameters of each distribution unit in the first distribution range of the target object in the state value function to obtain a score of the distribution unit;

the expansion acquisition unit 5142 determines the distribution units based on the scores of the distribution units, centering on the initial distribution unit, to obtain the expanded distribution units.

Therefore, the intelligent distribution range is generated through the deep learning algorithms, and the method can be suitable for complex application scenes with insufficient calculation power of the reinforcement learning algorithm, so that the optimal intelligent distribution range is obtained.

In addition, the expansion acquiring unit 5142 takes the initial distribution unit as the center, and determines the distribution unit according to the score of each distribution unit to obtain an expansion distribution unit, which specifically is: selecting a distribution unit with the highest value in distribution units adjacent to the initial distribution unit as a first expanded distribution unit, wherein the initial distribution unit and the first expanded distribution unit jointly form a first expanded range; selecting the distribution unit with the highest value in the distribution units adjacent to the first expansion range as a second expansion distribution unit, wherein the first expansion range and the second expansion distribution unit jointly form a second expansion range; in the same way, obtaining the nth expanded distribution unit and the nth expanded range until the requirement of the second distribution range is met, wherein n is a positive integer; the first expansion distribution unit and the second expansion distribution unit are used as expansion distribution units. When the distribution units are expanded outwards, the distribution unit with the highest score is selected as the expanded distribution unit each time, so that the distribution unit set with the highest total score can be obtained conveniently, and the optimal intelligent distribution range can be obtained.

In addition, if the (n-a) th expansion distribution unit obtained in the expansion obtaining unit 5142 includes a plurality of expansion distribution units, the nth expansion range with the highest total score is selected as the distribution range of the target object, where a is a positive integer greater than 1 and less than n.

In addition, the requirements for satisfying the second distribution range are specifically: the area of the second distribution range does not exceed a preset area threshold value, or the farthest distance between the initial distribution unit and the outermost expanded distribution unit does not exceed a preset distance threshold value. The purpose of setting the threshold is to automatically stop the expansion of the distribution range in an appropriate area, and to avoid the problem of insufficient computational power due to infinite expansion.

In addition, the order characteristic parameters of the delivery units in each module are obtained through the following steps: extracting order data of a first distribution range from a database, wherein the order data comprises a waybill width table, an order width table, a target object width table and a POI (point of interest) interface of an electronic map; and screening the order data to obtain the order characteristic parameters of each delivery unit in the first delivery range. The order characteristic parameters are distribution data having the most influence on distribution quality or distribution efficiency of the target object, and are calculated according to the order characteristic parameters, so that the optimal intelligent distribution range of the target object is obtained.

In addition, the distribution range generation device further includes:

a fence acquisition module 506, which acquires a geo-fence including preset keywords;

an invalid judgment module 507, which judges the distribution units in the second distribution range according to the geofence to obtain invalid distribution units, wherein the invalid sub-area units are located in the geofence or include the geofence;

the invalid delete module 508 deletes the invalid delivery unit in the second delivery range. By deleting invalid areas where logistics such as rivers, lakes and seas, scenic spots, parks, cemeteries and the like cannot be distributed, the order distribution range is further optimized.

The sixth embodiment of the present invention relates to an electronic device, and the electronic device of the present embodiment may be a terminal device, such as a mobile phone, a tablet computer, and the like, and may also be a server on a network side.

As shown in fig. 6, the electronic device: comprises at least one processor 601; and a memory 602 communicatively coupled to the at least one processor 601; and a communication component 603 communicatively coupled to the scanning device, the communication component 603 receiving and transmitting data under control of the processor 601; wherein the memory 602 stores instructions executable by the at least one processor 601, the instructions being executable by the at least one processor 601 to implement:

the method comprises the steps of obtaining order characteristic parameters of all distribution units in a first distribution range corresponding to a target object, wherein the target object is located in an initial distribution unit in the first distribution range; taking the initial distribution unit as a center, and obtaining an expanded distribution unit from the distribution unit according to the distribution range model and the order characteristic parameters of the distribution unit; generating a second delivery range of the target object according to the initial delivery unit and the expansion delivery unit;

acquiring a geo-fence comprising preset keywords; judging each distribution unit and sub-area unit in the second distribution range according to the geo-fence to obtain an invalid unit; the invalid unit is deleted in the second delivery range.

Specifically, fig. 6 illustrates a bus connection. The memory 602, which is a non-volatile computer-readable storage medium, may be used to store non-volatile software programs, non-volatile computer-executable programs, and modules. The processor 601 executes various functional applications and data processing of the device, that is, implements the above-described distribution range generation method, by executing nonvolatile software programs, instructions, and modules stored in the memory 602.

The memory 602 may include a storage program area and a storage data area, wherein the storage program area may store an operating system, an application program required for at least one function; the storage data area may store a list of options, etc. Further, the memory 602 may include high speed random access memory, and may also include non-volatile memory, such as at least one magnetic disk storage device, flash memory device, or other non-volatile solid state storage device. In some embodiments, the memory 602 may optionally include memory located remotely from the processor 601, which may be connected to an external device via a network. Examples of such networks include, but are not limited to, the internet, intranets, local area networks, mobile communication networks, and combinations thereof.

One or more modules are stored in the memory 602 and, when executed by the one or more processors 601, perform the method of generating a delivery range in any of the method embodiments described above.

The product can execute the method provided by the embodiment of the application, has corresponding functional modules and beneficial effects of the execution method, and can refer to the method provided by the embodiment of the application without detailed technical details in the embodiment.

In the embodiment, by constructing the distribution range model and continuously training the distribution range model, when the data information of the target object is input, the optimized intelligent distribution range of the target object can be automatically generated, and the traditional method for confirming the distribution range by personal experience, common knowledge, geographic information around the target object and the conventional order analysis is replaced.

A seventh embodiment of the invention relates to a non-volatile storage medium for storing a computer-readable program for causing a computer to perform some or all of the above method embodiments.

That is, as can be understood by those skilled in the art, all or part of the steps in the method for implementing the embodiments described above may be implemented by a program instructing related hardware, where the program is stored in a storage medium and includes several instructions to enable a device (which may be a single chip, a chip, or the like) or a processor (processor) to execute all or part of the steps of the method described in the embodiments of the present application. And the aforementioned storage medium includes: a U disk, a removable hard disk, a Read-only memory (ROM), a Random Access Memory (RAM), a magnetic disk or an optical disk, and other various media capable of storing program codes.

It will be understood by those of ordinary skill in the art that the foregoing embodiments are specific examples for carrying out the invention, and that various changes in form and details may be made therein without departing from the spirit and scope of the invention in practice.

The embodiment of the invention discloses A1. a method for generating a distribution range, which comprises the following steps:

the method comprises the steps of obtaining order characteristic parameters of all distribution units in a first distribution range corresponding to a target object, wherein the target object is located in an initial distribution unit in the first distribution range;

taking the initial distribution unit as a center, and obtaining an expanded distribution unit from the distribution unit according to a distribution range model and the order characteristic parameters of the distribution unit;

and generating a second delivery range of the target object according to the initial delivery unit and the expanded delivery unit.

A2. The delivery range generation method according to a1, the method further comprising:

constructing an initial distribution range model, wherein the initial distribution range model comprises a state value function, and the state value function takes the order characteristic parameters of the distribution units as parameters;

and training the distribution range model according to a machine learning algorithm by using the historical order characteristic parameters of the object in the business circle where the target object is located.

A3. The delivery range generation method according to A1, wherein the machine learning algorithm is a reinforcement learning algorithm,

the obtaining of the order characteristic parameters of the expanded distribution unit from the distribution unit according to the distribution range model by taking the initial distribution unit as the center specifically comprises:

inputting order characteristic parameters of each delivery unit in the first delivery range of the target object in the state value function to obtain the score of the delivery unit;

obtaining a complete state value matrix of the distribution units in the first distribution range according to the distribution range model and the values of the distribution units;

and with the initial distribution unit as a center, judging the distribution unit according to the complete state value matrix to obtain the expanded distribution unit.

A4. According to the distribution range generating method described in a3, the obtaining a complete state value matrix of the distribution unit in the first distribution range according to the distribution range model and the score of the distribution unit specifically includes:

obtaining an initial state value matrix according to the distribution range model and the value of the initial distribution unit;

taking the initial distribution unit and one or more distribution units adjacent to the initial distribution unit as a second distribution unit set, and obtaining a second state value matrix according to the distribution range model and the value sum of the distribution units of the second distribution unit set;

taking the second distribution unit set and one or more distribution units adjacent to the second distribution unit set as a third distribution unit set, and obtaining a third state value matrix according to the distribution range model and the value sum of the distribution units of the third distribution unit set;

and repeating the steps until the distribution units in the first distribution range are traversed to obtain the complete state value matrix.

A5. According to the distribution range generating method described in a4, the determining, with the initial distribution unit as a center, the distribution unit according to the complete state value matrix to obtain the expanded distribution unit specifically includes:

inquiring the score and the highest distribution unit set in the complete state value matrix, and taking the distribution units except the initial distribution unit in the score and the highest distribution unit set as the expansion distribution units.

A6. The delivery range generation method of A1, wherein the machine learning algorithm is a deep learning algorithm,

and with the initial distribution unit as a center, judging the distribution unit according to the score of the distribution unit to obtain the expanded distribution unit.

A7. The distribution range generating method according to a6, wherein the determining the distribution units based on the scores of the distribution units with the initial distribution unit as a center to obtain expanded distribution units includes:

selecting a distribution unit with the highest value in distribution units adjacent to the initial distribution unit as a first expanded distribution unit, wherein the initial distribution unit and the first expanded distribution unit jointly form a first expanded range;

selecting a distribution unit with the highest value in distribution units adjacent to the first expansion range as a second expansion distribution unit, wherein the first expansion range and the second expansion distribution unit jointly form a second expansion range;

repeating the steps until the requirement of the second distribution range is met, and obtaining an nth expansion distribution unit and an nth expansion range, wherein n is a positive integer;

the first expansion distribution unit and the second expansion distribution unit are the n-th expansion distribution unit.

A8. The delivery range generation method according to a7, the delivery range generation method further comprising:

if the obtained (n-a) th expanded distribution unit comprises a plurality of units, selecting the nth expanded range with the highest total score as the distribution range of the target object, wherein a is a positive integer which is greater than 1 and less than n.

A9. The distribution range generation method according to a7, wherein the requirement for satisfying the first distribution range is specifically: the area of the second distribution range does not exceed a preset area threshold, or the farthest distance between the initial distribution unit and the outermost expanded distribution unit does not exceed a preset distance threshold.

A10. According to the distribution range generation method of any one of A1-A9, the order characteristic parameters of the distribution units are obtained through the following steps:

extracting order data of the first distribution range from a database, wherein the order data comprises a waybill width table, an order width table, a target object width table and a POI (point of interest) interface of an electronic map;

and screening the order data to obtain order characteristic parameters of each distribution unit in the first distribution range.

A11. The delivery range generation method according to a10, further comprising:

acquiring a geo-fence comprising preset keywords;

judging each distribution unit in the second distribution range according to the geo-fence to obtain an invalid distribution unit, wherein the invalid sub-area unit is located in the geo-fence or comprises the geo-fence;

deleting the invalid delivery units in the second delivery range.

The embodiment of the present invention further provides a b12. a distribution range generation apparatus, including:

the data acquisition module is used for acquiring order characteristic parameters of all distribution units in a first distribution range corresponding to a target object, wherein the target object is located in an initial distribution unit in the first distribution range;

the grid expansion module is used for obtaining an expanded distribution unit from the distribution units by taking the initial distribution unit as a center according to a distribution range model and the order characteristic parameters of the distribution units;

a range generation module that generates a second delivery range of the target object based on the initial delivery unit and the expanded delivery unit.

An embodiment of the present invention further provides c13. an electronic device, including at least one processor; and the number of the first and second groups,

a memory and a communication unit communicatively coupled to the at least one processor;

wherein the memory stores instructions executable by the at least one processor to implement:

the method comprises the steps of obtaining order characteristic parameters of each delivery unit in a first delivery range corresponding to a target object, wherein the target object is located in an initial delivery unit in the first delivery range;

C14. The electronic device of C13, the processor further configured to perform:

constructing a distribution range model, wherein the distribution range model comprises a state value function, and the state value function takes the order characteristic parameters of the distribution units as parameters;

C15. The electronic device of claim 13, wherein the machine learning algorithm is a reinforcement learning algorithm,

the processor executes the steps of taking the initial distribution unit as a center, and obtaining order characteristic parameters of the expanded distribution unit from the distribution unit according to a distribution range model, wherein the steps are as follows:

C16. According to the electronic device of C15, the processor executes the score according to the distribution range model and the distribution unit to obtain a complete state value matrix of the distribution unit set in the first distribution range, specifically:

C17. According to the electronic device of C16, the processor executes the determining, centering on the initial distribution unit, on the distribution unit according to the complete state value matrix, to obtain an expanded distribution unit, specifically:

C18. The electronic device of C13, wherein the machine learning algorithm is a depth learning algorithm,

inputting historical order characteristic parameters of the distribution units in the first distribution range of the target object in the state value function to obtain the scores of the distribution units;

C19. According to the electronic device of C18, the processor executes the determining of the distribution units based on the scores of the distribution units centered on the initial distribution unit to obtain expanded distribution units, specifically:

C20. The electronic device of C19, the processor further configured to perform:

C21. According to the electronic device of C19, the requirement for satisfying the first distribution range is specifically: the area of the second distribution range does not exceed a preset area threshold, or the farthest distance between the initial distribution unit and the outermost expanded distribution unit does not exceed a preset distance threshold.

C22. The electronic device of any of C13-C21, the order characteristic parameters of the delivery unit obtained by:

C23. The electronic device of C22, the processor further configured to perform:

acquiring a geo-fence comprising preset keywords;

judging each distribution unit in the second distribution range according to the geofence to obtain an invalid distribution unit, wherein the invalid sub-area unit is located in the geofence or comprises the geofence;

deleting the invalid delivery units in the second delivery range.

An embodiment of the present invention also provides a non-volatile storage medium storing a computer-readable program for causing a computer to execute the distribution range generating method according to any one of a1 to a11.

Claims

1. A method for generating a delivery range, comprising:

after a state value function included in a distribution range model determines the score of the distribution unit by using the order characteristic parameter of the distribution unit, determining a complete state value matrix of the distribution unit in the first distribution range according to the distribution range model and the score of the distribution unit, and expanding outwards by using the initial distribution unit as a center by using the complete state value matrix to obtain an expanded distribution unit; the distribution range model is obtained by training historical order characteristic parameters of objects in a business district where the target object is located, and the state value function in the distribution range model takes the order characteristic parameters of a distribution unit as input parameters;

2. The delivery range generation method according to claim 1, further comprising:

3. The delivery range generation method according to claim 2, wherein the machine learning algorithm is a reinforcement learning algorithm.

4. The delivery range generation method of claim 3, wherein the complete state value matrix is determined by:

5. The delivery range generation method according to claim 4, wherein the expanded delivery unit is determined by:

6. The delivery range generation method according to claim 5, wherein the expanded delivery unit is determined by:

7. The delivery range generation method according to claim 6, wherein the delivery range generation method further comprises:

if the obtained (n-a) th expanded distribution unit comprises a plurality of (n-a) th expanded distribution units, selecting the nth expanded range with the highest total score as the distribution range of the target object, wherein a is a positive integer which is greater than 1 and less than n.

8. The method for generating delivery ranges according to claim 6, wherein the requirement for satisfying the first delivery range is specifically: the area of the second distribution range does not exceed a preset area threshold, or the farthest distance between the initial distribution unit and the outermost expanded distribution unit does not exceed a preset distance threshold.

9. The method for generating a delivery range according to any one of claims 1 to 8, wherein the order characteristic parameter of the delivery unit is obtained by:

10. The delivery range generation method according to claim 9, further comprising:

acquiring a geo-fence comprising preset keywords;

judging each distribution unit in the second distribution range according to the geo-fence to obtain an invalid distribution unit, wherein the invalid distribution unit is located in the geo-fence or comprises the geo-fence;

deleting the invalid delivery units in the second delivery range.

11. An apparatus for generating a delivery range, comprising:

the grid expansion module is used for determining a score of the distribution unit by using an order characteristic parameter of the distribution unit through a state value function included in a distribution range model, determining a complete state value matrix of the distribution unit in the first distribution range according to the distribution range model and the score of the distribution unit, and expanding outwards by using the initial distribution unit as a center through the complete state value matrix to obtain an expanded distribution unit; the distribution range model is obtained by training historical order characteristic parameters of objects in a business district where the target object is located, and the state value function in the distribution range model takes the order characteristic parameters of a distribution unit as input parameters;

12. An electronic device comprising at least one processor; and the number of the first and second groups,

after a state value function included in a distribution range model determines the score of the distribution unit by using the order characteristic parameter of the distribution unit, determining a complete state value matrix of the distribution unit in the first distribution range according to the distribution range model and the score of the distribution unit, and expanding the initial distribution unit outwards by using the complete state value matrix as a center to obtain an expanded distribution unit; the distribution range model is obtained by training historical order characteristic parameters of objects in a business district where the target object is located, and the state value function in the distribution range model takes the order characteristic parameters of a distribution unit as input parameters;

13. The electronic device of claim 12, wherein the processor is further configured to perform:

14. The electronic device of claim 13, wherein the machine learning algorithm is a reinforcement learning algorithm.

15. The electronic device of claim 14, wherein the complete matrix of state values is determined by:

16. The electronic device of claim 15, wherein the expanded dispensing unit is determined by:

17. The electronic device of claim 16, wherein the expanded dispensing unit is determined by:

18. The electronic device of claim 17, wherein the processor is further configured to perform:

19. The electronic device according to claim 18, wherein the requirement for satisfying the first delivery range is specifically: the area of the second distribution range does not exceed a preset area threshold, or the farthest distance between the initial distribution unit and the outermost expanded distribution unit does not exceed a preset distance threshold.

20. Electronic device according to any of claims 12 to 19, wherein the order characterizing parameters of the delivery unit are obtained by:

21. The electronic device of claim 20, wherein the processor is further configured to perform:

acquiring a geo-fence comprising preset keywords;

deleting the invalid delivery units in the second delivery range.

22. A non-volatile storage medium storing a computer-readable program for causing a computer to execute the distribution range generation method according to any one of claims 1 to 10.