Shop site selection method and device, and storage medium

Abstract

The invention discloses a shop site selection method and device and a storage medium, and relates to the technical field of data processing, and the method comprises the steps: obtaining at least one piece of site selection data based on an execution module; transmitting the site selection data as input data to the intelligent agent, and acquiring decision action data output by the intelligent agent based on the input data; and determining a target shop point location selected by the intelligent agent based on the decision action data, and feeding back the target shop point location to the user side. Therefore, shop site selection can be carried out through the trained intelligent agent, and the accuracy of shop site selection decision is improved.

Description

Technical Field

[0001] The present application relates to the field of data processing technology, and in particular to a store site selection method, device and storage medium. Background Art

[0002] Currently, in the process of store location selection, business personnel usually analyze the available store locations based on simple statistical analysis. This analysis process mostly relies on personal experience and intuition to evaluate the pros and cons of potential locations, lacks the ability to learn and iterate independently, and cannot adapt to the complex and changing market environment, which leads to low accuracy in store location decision-making.

[0003] The above contents are only used to assist in understanding the technical solution of the present application and do not constitute an admission that the above contents are prior art. Summary of the invention

[0004] The present application provides a store site selection method, device and storage medium, aiming to solve the problem of low accuracy of store site selection decisions.

[0005] To achieve the above-mentioned purpose, the present application provides a store location selection method, the store location selection system includes an execution module, an agent and a training module, and the store location selection method includes the following steps:

[0006] Acquire at least one site selection data based on the execution module;

[0007] Passing the site selection data as input data to the intelligent agent, and obtaining decision action data output by the intelligent agent based on the input data;

[0008] The target store location selected by the agent is determined based on the decision action data, and the target store location is fed back to the user end.

[0009] In one embodiment, before the step of transmitting the site selection data as input data to the agent and obtaining the decision action data output by the agent based on the input data, the step includes:

[0010] Acquiring scene state data based on the training module;

[0011] Passing the scene state data as input data to the agent, and obtaining decision action data output by the agent based on the input data;

[0012] Calculate the Q value based on the scene state data and the decision action data and based on a preset formula;

[0013] Determine a reward value according to the scene state data and the reward function, and determine a target Q value according to the reward value and the Q value;

[0014] When the target Q value and the difference between the Q value are continuously iterated for a target number of times and are both less than a preset threshold, the agent is output.

[0015] In one embodiment, when the target Q value and the difference between the Q value are continuously iterated for a target number of times and are both less than a preset threshold, before the agent is output, the method includes:

[0016] Update the hyperparameters of the agent according to the Q value and the target Q value;

[0017] The step of obtaining scene status data is performed according to the updated agent.

[0018] In one embodiment, before the step of obtaining the decision action data output by the agent based on the input data, the step includes:

[0019] The agent determines a target strategy according to the received input data;

[0020] Through the target strategy, the corresponding decision action data is selected and output based on the input data, wherein the decision action data includes rating the store locations as golden locations, high-quality locations, ordinary locations, and worthless locations.

[0021] In one embodiment, the step of determining the reward value according to the scene state data and the reward function comprises:

[0022] Acquire various business indicator data in the scenario status data and corresponding weight values, wherein the business indicator data includes cost item indicators, income item indicators and risk item indicators;

[0023] According to the instant reward function, business indicator data and corresponding weight values, the instant reward for each business indicator data is calculated;

[0024] The reward value is calculated based on the reward function and the instant reward of each business indicator data.

[0025] In one embodiment, before the step of acquiring scene state data based on the training module, the method further includes:

[0026] Collect and integrate store locations, competitive product information, community information, administrative district information, and main road information, and determine them as the original data set;

[0027] Using geographic information coding technology to convert the address information in the original data set into longitude and latitude coordinates, and mark the store locations, competitive product locations and community locations on the map;

[0028] According to the preset business circle range, various business indicator data corresponding to the original data set within the radius are calculated and determined as scene status data.

[0029] In one embodiment, after the step of calculating various business indicator data within the radius according to the preset business circle range and determining them as the scene state data, the step further includes:

[0030] Sorting the scene state data according to the numerical value, and determining the first quartile and the third quartile, wherein the first quartile is the numerical value at the 25% position of the sorting, and the third quartile is the numerical value at the 75% position of the sorting;

[0031] Determine a fourth quartile range according to the first quartile and the third quartile, and determine an abnormal range according to the fourth quartile range;

[0032] The data in the abnormal range is cleared according to the abnormal range, and the processed scene state data is output.

[0033] In one embodiment, after the step of determining the target store location selected by the agent based on the decision action data and feeding back the target store location to the user terminal, the step further includes:

[0034] Predicting potential customer traffic at the target store locations;

[0035] Based on the prediction results and the corresponding cost-benefit analysis, business recommendations for the target store locations are generated.

[0036] In addition, to achieve the above-mentioned purpose, the present application also provides a store location selection device, which includes: a memory, a processor, and a computer program stored on the memory and executable on the processor, wherein the computer program is configured to implement the steps of the store location selection method as described above.

[0037] In addition, to achieve the above-mentioned purpose, the present application also provides a storage medium, which is a computer-readable storage medium, and a computer program is stored on the computer-readable storage medium. When the computer program is executed by a processor, the steps of the store location selection method as described above are implemented.

[0038] The present application provides a store site selection method, a store site selection device and a storage medium, which obtains at least one site selection data based on the execution module, and then transmits the site selection data as input data to the intelligent agent, and obtains decision action data output by the intelligent agent based on the input data, and finally determines the target store location selected by the intelligent agent based on the decision action data, and feeds back the target store location to the user end, so that the store site selection can be performed through the trained intelligent agent to improve the accuracy of the store site selection decision. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the present application and, together with the description, serve to explain the principles of the present application.

[0040] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, for ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative labor.

[0041] Figure 1 A schematic diagram of the flow chart of the first embodiment of the store location selection method of the present application;

[0042] Figure 2 A schematic diagram of the flow chart of the second embodiment of the store site selection method of the present application;

[0043] Figure 3 A quantitative modeling diagram of indicators of scene status data involved in the solution of this embodiment;

[0044] Figure 4 A specific parameter diagram of the scene state data involved in the embodiment of the present application;

[0045] Figure 5 This is a decision action data table involved in the embodiment of the present application;

[0046] Figure 6 This is a Q value matrix diagram involved in the embodiment of the present application;

[0047] Figure 7 This is a hyperparameter assignment table involved in the embodiment of this application;

[0048] Figure 8 This is a hyperparameter adjustment table involved in the embodiment of the present application;

[0049] Fig. 9 A graph showing the association between the reward function and the business indicator data involved in the solution of this embodiment;

[0050] Fig.10 This is a flow chart of the third embodiment of the store location selection method of the present application;

[0051] Fig.11 This is a point calibration diagram involved in the embodiment of the present application;

[0052] Fig.12 This is a statistical table of data involved in the embodiments of this application;

[0053] Fig.13A schematic diagram of the data preprocessing process involved in the embodiment of the present application;

[0054] Fig.14 This is a schematic diagram of the architecture of the hardware operating environment of the store site selection device involved in the embodiment of the present application.

[0055] The realization of the purpose, functional features and advantages of this application will be further explained in conjunction with embodiments and with reference to the accompanying drawings. DETAILED DESCRIPTION

[0056] It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0057] In order to better understand the above technical solution, exemplary embodiments of the present application will be described in more detail below with reference to the accompanying drawings. Although exemplary embodiments of the present application are shown in the accompanying drawings, it should be understood that the present application can be implemented in various forms and should not be limited by the embodiments described herein. On the contrary, these embodiments are provided in order to enable a more thorough understanding of the present application and to fully convey the scope of the present application to those skilled in the art.

[0058] In order to better understand the above technical solution, the above technical solution will be described in detail below in conjunction with the accompanying drawings and specific implementation methods.

[0059] The main solution of the present application is: to obtain at least one site selection data based on the execution module; to pass the site selection data as input data to the intelligent agent, and to obtain decision action data output by the intelligent agent based on the input data; to determine the target store location selected by the intelligent agent based on the decision action data, and to feed back the target store location to the user end.

[0060] Currently, in the process of store location selection, business personnel usually analyze the available store locations based on simple statistical analysis. This analysis process mostly relies on personal experience and intuition to evaluate the pros and cons of potential locations, lacks the ability to learn and iterate independently, and cannot adapt to the complex and changing market environment, which leads to low accuracy in store location decision-making.

[0061] The present application obtains at least one site selection data based on the execution module, and then transmits the site selection data as input data to the intelligent agent, and obtains decision action data output by the intelligent agent based on the input data, and finally determines the target store location selected by the intelligent agent based on the decision action data, and feeds back the target store location to the user end, so that the store location can be selected through the trained intelligent agent to improve the accuracy of the store location selection decision.

[0062] It should be noted that the execution subject of this embodiment can be a store location selection system, or a computing service device with data processing, network communication and program running functions, such as a tablet computer, a personal computer, a mobile phone, etc., or a store location selection device capable of realizing the above functions, etc., and this embodiment does not specifically limit this. The following takes the store location selection system as an example to illustrate this embodiment and the following embodiments.

[0063] Embodiment 1

[0064] Please refer to Figure 1 In the first embodiment, the store location selection method includes the following steps:

[0065] Step S10: obtaining at least one site selection data based on the execution module.

[0066] In this embodiment, the processing action is performed by the store location selection system, and the store location selection system includes an execution module, an intelligent agent, and a training module. After the store system detects that the intelligent agent training is completed, the execution module obtains and deploys the intelligent agent in the store location selection system. A location selection interface is displayed in the front end corresponding to the store location selection system. When the user uploads a location selection request based on the location selection interface, the front end feeds back the location selection request to the store location selection system. The store location selection system determines the corresponding location selection data based on the location selection request, wherein the location selection data is at least one, and the location selection data is pre-processed data. The specific steps are the data pre-processing steps described in Example 3, which will not be repeated here. In addition, the location selection data includes location selection data of multiple store locations, and the execution module inputs the location selection data of at least one store location into the intelligent agent, and the intelligent agent outputs decision action data of at least one store location.

[0067] Step S20: passing the site selection data as input data to the agent, and obtaining decision action data output by the agent based on the input data.

[0068] In this embodiment, the preprocessed site selection data is used as input data and transmitted to the agent, and the agent selects a corresponding target strategy based on the input data and outputs corresponding decision action data. The decision action data output by the agent is obtained, wherein the decision action data is the rating of the target point in the site selection data by the agent.

[0069] Step S30: Determine the target store location selected by the agent based on the decision action data, and feed back the target store location to the user end.

[0070] In this embodiment, the site selection data corresponds to data of multiple target points. According to the decision action data output by the agent for each target point, the point with the highest rating among the target points is determined, and the point is determined as the target store point. The target store point is then fed back to the user end, thereby completing the store site selection.

[0071] Optionally, in this embodiment, after the step of determining the target store location selected by the agent based on the target strategy and feeding back the target store location to the user terminal, the step further includes:

[0072] Predict the potential customer traffic of the target store location; and generate business proposals for the target store location based on the prediction results and corresponding cost-benefit analysis.

[0073] Specifically, predicting the potential customer traffic of the target store location involves multiple factors, including but not limited to the geographical location, surrounding demographic characteristics, traffic conditions, competition, seasonal changes, etc. By adopting a data-driven approach and combining data analysis technology. The specific location of the store and its surrounding environmental characteristics, including demographic data, traffic data, peak traffic hours, the location, type, price range, and historical sales data of surrounding competitors are analyzed to determine the potential customer traffic. Business recommendations are then determined based on the above analysis results or model output results. For example, if the predicted customer traffic indicates that the location is not suitable for opening a store, other locations need to be reconsidered. Furthermore, a marketing plan is formulated based on the characteristics of the target customer base, such as attracting young customers through social media promotion, so as to accurately predict potential customer traffic and conduct a cost-benefit analysis.

[0074] In addition, potential customer traffic can also be predicted by building a model. The specific steps are to first select a suitable prediction model, such as linear regression, time series analysis, machine learning model (such as random forest, neural network, etc.), and then create useful feature variables based on the collected data, and use the selected model to predict customer traffic in different time periods.

[0075] In the technical solution provided in this embodiment, at least one site selection data is obtained based on the execution module, and then the site selection data is passed to the intelligent agent as input data, and decision action data output by the intelligent agent based on the input data is obtained, and finally the target store location selected by the intelligent agent is determined based on the decision action data, and the target store location is fed back to the user end, so that the store location can be selected through the trained intelligent agent to improve the accuracy of the store location selection decision.

[0076] Embodiment 2

[0077] Please refer to Figure 2In the second embodiment, after step S20, the following steps are further included:

[0078] Step S40: Acquire scene status data based on the training module.

[0079] In this embodiment, the processing action is performed by the training module in the store location selection system, and the scene status data is pre-processed data, wherein the scene status data includes store rent, transfer fee, management fee, average wage in the city, number of communities within a radius of 250 meters, community occupancy rate, rent per square meter in the community, second-hand housing price per square meter in the community, number of competing stores within a radius of 250 meters, and number of two-way roads with four lanes or more within a radius of 250 meters.

[0080] Optionally, the scenario status data in this embodiment is divided into cost item indicators, income item indicators and risk item indicators, and all have fixed units, such as Figure 3 As shown, Figure 3 This is a quantitative modeling diagram of indicators of scene status data involved in the scheme of this embodiment. Among them, the cost item indicators include shop rent, transfer fee, management fee and average wage in the city, corresponding to the lower the overall monthly rent of the shop, the lower the cost, the lower the transfer fee, the lower the sunk cost, the lower the management fee per unit area, and the lower the city wage, the lower the labor cost. The income item indicators include the number of communities within a radius of 250 meters, the community occupancy rate, the rent per square meter of the community, and the second-hand housing price per square meter of the community, corresponding to the more communities, the more potential sales opportunities, the higher the occupancy rate, the more potential sales opportunities, the higher the rent, the higher the average transaction price, and the higher the housing price, the higher the average transaction price. The risk item indicators include the number of store floors, the number of competing stores within a radius of 250 meters, and the number of two-way roads with more than 4 lanes within a radius of 250 meters, corresponding to the more floors, the smaller the sales per unit area, the more competing products, the fewer sales opportunities, and the more lanes, the closer to the main road, the fewer potential sales opportunities and store entry time.

[0081] Specifically, the main cost items include labor cost expenditure, store rental expenditure, fixed ratio management fee expenditure, etc., which are the foundation for the establishment of a store. In today's business and real estate environment, controlling store cost expenditure has a more significant and far-reaching impact on store cash flow and profits than marketing growth. The main income items can be divided into the number of surrounding potential consumers, the level of surrounding potential consumption capacity, and the scope of business coverage. Such factors affect store revenue qualitatively rather than quantitatively. They are attached to the physical retail environment. The flow of people determines the exposure of the sales channel. The more people flow to the store, the higher the sales volume will be. The more prosperous the business district where the store is located, the higher the premium that can be attached to the corresponding goods, and the higher the single transaction amount. In addition to direct cost and income factors, there are also risk items that indirectly affect the profitability of the store. Such risks are known and unknown risks, that is, we know that the risk exists, but we don’t know the probability of the risk and the impact. If there are many main roads / expressways around the store, customers may not stop in to shop even if there is a large flow of people and many passers-by in the community. From the perspective of consumer psychology, the traffic flow of fast retail / fresh food retail stores should be as smooth and continuous as possible. If customers need to go up and down stairs frequently and there are many floors, it will greatly affect the customer shopping experience and cause sales losses. The number of similar competing stores that already exist near the site will also greatly affect the store sales, and may even form a customer flow monopoly, which is a direct threat to the store sales.

[0082] In addition, since the units (dimensions) and magnitudes of the evaluation indicators are different, a unified evaluation mechanism is introduced to perform dimensionless processing. Here, a 10-point scoring model is used. After unified quantitative modeling, all state s values ​​are converted into dimensionless data of the same magnitude. The specific scenario state data s parameters are as follows: Figure 4 . Determine the scenario status data based on the cost item indicators, income item indicators and risk item indicators, such as Figure 4 As shown, Figure 4 This is a specific parameter diagram of the scene state data involved in the embodiment of the present application. C1 to C4 represent cost item indicators, P1 to P4 represent income item indicators, R1 to R3 represent risk item indicators, and s represents the scene state data. The mathematical set of the scene state data s is:

[0083] s={(C1,C2,C3,...,R3)|C1,C2,C3,...,R2∈{0,1,2,...,10},R3∈{0,3,5}}

[0084] Step S50: passing the scene state data as input data to the agent, and obtaining decision action data output by the agent based on the input data.

[0085] In this embodiment, the agent refers to an entity that can perceive the environment and influence the environment by taking actions, and can be a software program, a module deployed in the store location selection system, or other modules that can make decisions autonomously. After the store location selection system obtains the scene state data, the scene state data is passed to the agent as input data, and when the action output by the agent is detected, the decision action data output by the agent based on the input data is obtained.

[0086] It can be understood that the agent maximizes the cumulative reward by interacting with the scene state data. The agent is provided with a perception module, a decision module and an action module. The agent obtains information about the current environment through the perception module. This information is usually provided in the form of scene state data, which can be a partial or complete description of the environment. Then the decision module takes corresponding actions based on the current state information, that is, decision action data. This decision process is based on the corresponding strategy or a complex machine learning model. Finally, the action module performs specific behaviors according to the action instructions output by the decision module. The workflow of the agent is that the agent obtains information about the current state from the environment. Based on the current state, the agent decides which action to take. This decision may be based on a predefined strategy or a strategy obtained through learning. The agent performs the selected action. The environment gives feedback based on the action of the agent, usually a formal reward signal. The agent updates its internal state or strategy based on the feedback so that it can make better decisions in the future.

[0087] For example, the decision action data output by the intelligent agent includes golden locations, high-quality locations, ordinary locations, and worthless locations, corresponding to the three categories of A / B / C. The business significance of store A is the golden location that can be signed within 1 month, store B is defined as a high-quality location that can be signed within 1 to 3 months, and store C is defined as an ordinary location that can be signed for more than 3 months. In addition, there is a worthless location X that has been abandoned by the actual business and has no potential for signing and making profits. Figure 5 As shown, Figure 5 The decision action data table involved in the embodiment of this application. The decision actions a of the agent when facing different scene states s are A, B, C and X. Figure 5 It can be concluded that the mathematical set of decision action a is: a = {A, B, C, X}.

[0088] Optionally, in this embodiment, before the step of obtaining the decision action data output by the agent based on the input data, the step includes:

[0089] The intelligent agent determines a target strategy based on the received input data; through the target strategy, selects and outputs the corresponding decision action data based on the input data, wherein the decision action data includes rating store locations as golden locations, high-quality locations, ordinary locations, and worthless locations.

[0090] Specifically, after the execution module obtains the scene state data, or the training module obtains the scene state data, the execution module or the training module transmits the scene state data as input data to the agent. After the agent receives the input data, the decision module in the agent automatically selects the target strategy according to the scene state data, and performs a specific decision action according to the action instruction output by the target strategy, thereby outputting decision action data. The decision action data includes evaluating the store location as a golden location, a high-quality location, an ordinary location, and a worthless location.

[0091] Step S60: Calculate the Q value according to the scene state data and the decision action data based on a preset formula.

[0092] In this embodiment, after the agent outputs the decision action data, the Q value is calculated based on Q(st,at), where s t represents the scene state data at time step t, a t represents the decision action of the agent for the scene state data s at time step t, and the Q value is calculated based on the above data and based on Q(st,at).

[0093] Specifically, according to the set expressions of state s and action a, the Q value table can be formed by s×a, that is, the Cartesian product of the two sets. Figure 6 The Q value matrix shown, Figure 6 This is the Q value matrix diagram involved in the embodiment of this application. This matrix is ​​the core Q value table. Relying on the update of the values ​​of this matrix, the action decisions of the intelligent body in different states are recorded. By continuously updating this table through the iterative formula, the learning process of the intelligent body can be realized.

[0094] Step S70: determining a reward value according to the scene state data and the reward function, and determining a target Q value according to the reward value and the Q value.

[0095] In this embodiment, the scene state data corresponds to a reward function. In the above steps, the scene state data is divided according to cost item indicators, income item indicators and risk item indicators. Each type of indicator data corresponds to a state code for identification, and the identification is associated with a reward function. According to the value of the scene state data, the value is substituted into the reward function to obtain the reward value. Then, the target Q value is calculated according to the reward value and the Q value.

[0096] Specifically, in the scheme of this embodiment, the target Q value is calculated according to the reward value and the Q value, and the Q value is iteratively calculated. Further, in this embodiment, when the difference between the target Q value and the Q value is continuously iterated for a target number of times and is less than a preset threshold, before the agent is output, it includes:

[0097] The hyperparameters of the agent are updated according to the Q value and the target Q value; and the step of acquiring the scene state data is performed according to the updated agent.

[0098] Specifically, each type of scene state data corresponds to a time step. After the end of this time step, the hyperparameters of the agent need to be updated according to the Q value and the target Q value, wherein the hyperparameters include learning rate, immediate reward, discount factor, exploration rate, exploration rate decay, minimum exploration rate and exploration strategy. After the agent is updated, the store location selection system will randomly select another target scene state data, so that the agent executes the next time step. The agent selects the target decision action data based on the target state data and the corresponding strategy, calculates the Q value of this time step, and then calculates the difference, that is, iteratively calculates the Q value based on the Q-learning algorithm and the Bellman equation. The formula is:

[0099] Q(s t ,a t )←Q(s t ,a t )+α[r t+1 +γQ a (s t+1 ,a)-Q(s t ,a t )]

[0100] Q(s t ,a t ) is the state s at time step t t Next, take action a t The value of is (0,1]. Where "s" is the state of the external environment, that is, the scene state data, "a" is the action that the corresponding agent can take under a given state, that is, the decision action data; t is the time step. In Q learning, time slices are discrete concepts of time steps, that is, the time nodes when actions are to be taken in the face of the environment, and many discrete time slices are fitted into a continuous decision process. α is the learning rate, which determines the extent to which the agent uses new Q values ​​to update old Q values, that is, the proportion of using new knowledge to replace old knowledge, and takes values ​​(0,1]; r is the immediate reward obtained after executing the action, which is the main reward and punishment factor after executing the action, and is used as the main feedback to regulate the action selection of the agent. γ is the discount factor, which determines the current value of the discounted future reward, and takes values ​​(0,1); st and t+1 They are the states at time steps t and t+1 respectively. When iteratively calculating the Q value based on the Q-learning algorithm and the Bellman equation, the state function is initialized first:

[0101] s={(C1,C2,C3,...,R3)|C1,C2,C3,...,R2∈{0,1,2,...,10},R3∈{0,3,5}}

[0102] All states are at position 0, and the Q value table is also initialized to 0. The time step t is divided into two steps. t is the time node for making judgments based on historical experience, and t+1 is the time node for comparing the agent's decision based on the actual point rating. learning =r process +r final Reward function, substitute the process reward and result reward into the formula, use the historical Q value to make a decision, and the process reward and result reward are converted by coefficients. The iterative part is updated into the Q indicator again, so as to iterate repeatedly until the optimal solution is approached. In addition, when determining the agent's exploration strategy, there is another hyperparameter - the exploration rate ε, which is not reflected in the formula and is equally important for agent learning. ε controls the frequency of the agent's exploration of unknown actions when using a greedy strategy, which is similar to the aggressiveness of exploring new actions. It takes a value of (0,1]. For the learning rate α, discount factor γ, immediate reward r, and exploration rate ε, there are default values ​​at the beginning of the iteration, such as Figure 7 As shown, Figure 7 This is the hyperparameter assignment table involved in the embodiment of this application. After the agent converges, the production parameters are tuned to better retain the existing model features and avoid decision divergence. The specific adjustments are as follows: Figure 8 As shown, Figure 8 This is a hyperparameter adjustment table involved in the embodiment of the present application.

[0103] As an optional implementation method for obtaining scene state data, after the hyperparameters of the agent are updated, target scene state data different from the previous training step can be obtained, and the target scene state data can be input into the updated agent to continue the training iteration process of the agent. By using multiple different types of scene state data, the accuracy of agent training can be improved.

[0104] As another optional implementation method for obtaining scene state data, after the hyperparameters of the agent are updated, the scene state data of the previous training step can continue to be used until the decision action data output by the agent is calculated, and when the difference between the target Q value and the Q value is less than a preset threshold, the target scene decision action data is replaced and input into the agent training. By using the same scene state data to train the agent multiple times, the amount of training data prepared can be reduced to improve the efficiency of agent training.

[0105] Optionally, in this embodiment, the step of determining the reward value according to the scene state data and the reward function includes:

[0106] Obtain various business indicator data and corresponding weight values ​​in the scenario status data, wherein the business indicator data include cost item indicators, revenue item indicators and risk item indicators; calculate the instant reward of each business indicator data according to the instant reward function, the business indicator data and the corresponding weight values; calculate the reward value based on the reward function and the instant reward of each business indicator data.

[0107] Specifically, the business indicator data includes cost item indicators, income item indicators and risk item indicators. Each business indicator data is correspondingly set with a weight value. The reward function includes giving an instant reward to the action taken by the agent at each time step. The reward value is calculated by each instant reward, such as Fig. 9 As shown, Fig. 9 This is a graph showing the association between the reward function and the business indicator data involved in the solution of this embodiment. The reward function can be described as:

[0108]

[0109] The instant reward for each business indicator data is calculated by multiplying each instant reward function by the weight value. Finally, the judgment of the agent's decision result is mainly based on the decision action a given by the agent after judging the current single-dimensional state s. If the action a given by the agent is different from the actual result of the training set, it is considered that the decision is wrong. Because the model training of this method is result-oriented, the agent must be given an overwhelming order of magnitude penalty, which is enough to offset the reward in the process guidance and still have a margin of more than an order of magnitude, and vice versa. This reward function is only applicable in the agent learning stage. There is no judgment error after the model converges and is put into production. The specific judgment is as follows:

[0110]

[0111] In summary, the reward function of the Q learning algorithm during agent training is r learning =r process +r final .

[0112] Step S80: When the target Q value and the difference between the Q value are continuously iterated for a target number of times and are both less than a preset threshold, the agent is output.

[0113] In this embodiment, the target number of times can be preset, specifically 500 times, and the preset threshold is set to 0.01. For the data set, the iterative stability of the difference in the observed Q value is used to determine its convergence, so as to achieve a balance between data volume and efficiency. When the difference in the change of 500 consecutive Q value iterations is less than 0.01, it can be considered that the agent has converged, and then the agent is output and deployed in the store location selection system.

[0114] In the technical solution provided in this embodiment, scene state data is obtained based on the training module, and the scene state data is input into the intelligent agent, and decision action data output by the intelligent agent is determined, then the Q value is calculated based on the scene state data and the decision action data and based on a preset formula, and then the reward value is determined based on the scene state data and the reward function, and the target Q value is determined based on the reward value and the Q value. Finally, when the difference between the target Q value and the Q value is continuously iterated for a target number of times and is less than a preset threshold, the intelligent agent is output, so that store location selection can be performed through the trained intelligent agent to improve the accuracy of store location selection decisions.

[0115] Embodiment 3

[0116] Please refer to Fig.10 In the second embodiment, before the step of acquiring scene state data based on the training module, it also includes:

[0117] Step S60: Collect and integrate store locations, competitive product information, community information, administrative district information, and main road information, and determine them as the original data set.

[0118] Step S70: Use geographic information coding technology to convert the address information in the original data set into longitude and latitude coordinates, and mark the store locations, competitive product locations and community locations on the map.

[0119] Step S80: Calculate various business indicator data corresponding to the original data set within the radius according to the preset business circle range, and determine them as the scene status data.

[0120] In this embodiment, various types of original data sets are obtained before the data source, including store locations, competitive product information, community information, administrative district information and main road information. The data types of the original data sets are integers, floating point numbers and strings. The collected original data may not be directly usable, and it is necessary to convert indicators into definable and quantifiable data. Specifically, it involves data conversion and information collection with the help of front-end visual geographic information tools, and batch processing of indicator data in combination with automated scripts. Then, through the geographic information coding (Geocoding) function provided by the geographic information service provider, the address string is converted into a floating point number of longitude and latitude, and the site selection point, competitive product point and community point are rendered on the map.

[0121] Specifically, Fig.11 As shown, Fig.11 This is the point calibration diagram involved in the embodiment of this application. Among them, the cross is the competitor, the five-pointed star is the target point, the solid circle is the center point of the community, the gray dotted line is the lane dividing line, the black solid line is the community range and the road boundary, and the arrow indicates the flow direction of the main road traffic. With the target point as the center of the circle and a radius of 250 meters, a blue semi-transparent layer of the business coverage area is drawn, which can convert some counting indicators into plane geometric relationship problems.

[0122] According to the calibrated map above, determine the number of communities within a radius of 250 meters. The blue circle range intersects with the community polygon line segment. The intersection means the count is increased by 1. A single community is not accumulated repeatedly, and the value is 5. The number of competing stores within a radius of 250 meters. The geometric inclusion relationship between the circle and the point is determined by the blue circle range and the competing points. Inclusion means the count is increased by 1, and the value is 2. The number of two-way roads with more than 4 lanes within a radius of 250 meters is determined by judging the number of lanes. Roads with more than 4 lanes are selected, and any road edge segment is determined to intersect with the circle, that is, the count is increased by 1, which is determined to be 1 according to the map. After data completion and indicator conversion, the data is associated through the foreign keys of each table, and unnecessary auxiliary related data is discarded. Finally, a unified data format with the site selection point as the dimension can be output, and each field is scored and filled in according to the scoring scheme to obtain Fig.12 , Fig.12 This is a statistical table of data involved in the implementation scheme of this application.

[0123] For example, Fig.13 As shown, Fig.13This is a flow chart of data preprocessing involved in the embodiment of the present application. The user selects or inputs the selected target point in the front-end interface, and the front-end obtains the original data set of the target point through the interface and determines the geographic information therein. The address information in the original data set is converted into longitude and latitude coordinates using geographic information coding technology, and the store locations, competitor locations, and community locations are marked on the map, and fed back to the front-end for display. After the calculation is completed, the plane geometric relationship and the cumulative indicator count are determined, the indicator data of the site selection point is updated, and it is determined as scene state data and saved in the database.

[0124] Optionally, in this embodiment, after the step of calculating various business indicator data within the radius according to the preset business circle range and determining them as the scene state data, the step further includes:

[0125] The scene status data are sorted according to the numerical value, and the first quartile and the third quartile are determined, wherein the first quartile is the value located at the 25% position of the sort, and the third quartile is the value located at the 75% position of the sort; the fourth quartile range is determined according to the first quartile and the third quartile, and the abnormal range is determined according to the fourth quartile range; the data in the abnormal range is cleared according to the abnormal range, and the processed scene status data is output.

[0126] Specifically, after the data is processed, it is necessary to remove outliers in the data and then refine the data. The specific cleaning method is the quartile method. First, calculate and determine the interquartile range (IQR), calculate the first quartile (Q1, 25% quantile) and the third quartile (Q3, 75% quantile), and then calculate the IQR: IQR = Q3-Q1, and then determine the range of outliers. The specific abnormal range is that if the data point is lower than Q1-1.5×IQR or higher than Q3+1.5×IQR, it is considered an outlier. Delete the corresponding abnormal data points in the data set according to the abnormal range. In addition, after removing the outliers, the remaining data set is re-analyzed by the quartile method. It is determined that there is no value outside the normal range, and the cleaning is considered complete.

[0127] In the technical solution provided in this embodiment, by collecting and integrating store locations, competitive product information, community information, administrative district information and main road information, and determining them as original data sets, the address information in the original data sets is converted into longitude and latitude coordinates using geographic information coding technology, and the store locations, competitive product locations and community locations are marked on the map, and finally, according to the preset business district range, the various business indicator data corresponding to the original data sets within the radius are calculated and determined as the scene state data. This embodiment scheme ensures the reliability of the training data of the intelligent agent by preprocessing the data, thereby improving the training efficiency of the intelligent agent.

[0128] Since the system introduced in the embodiment of the present application is a system used to implement the method of the embodiment of the present application, based on the method introduced in the embodiment of the present application, the person skilled in the art can understand the specific structure and deformation of the system, so it is not repeated here. All systems used in the method of the embodiment of the present application belong to the scope of protection of this application.

[0129] The present application provides a store site selection device, which includes: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can execute the store site selection method in the above-mentioned embodiment one.

[0130] Reference below Fig.14 , which shows a schematic diagram of the structure of a store location selection device suitable for implementing the embodiment of the present application. The store location selection device in the embodiment of the present application may include but is not limited to mobile terminals such as mobile phones, laptop computers, digital broadcast receivers, PDAs (Personal Digital Assistants), PADs (Portable Application Descriptions), PMPs (Portable Media Players), vehicle-mounted terminals (such as vehicle-mounted navigation terminals), etc., and fixed terminals such as digital TVs, desktop computers, etc. Fig.14 The store location selection device shown is merely an example and should not bring any limitation to the functions and scope of use of the embodiments of the present application.

[0131] like Fig.14As shown, the store location selection device may include a processing device 1001 (e.g., a core processor, a graphics processor, etc.), which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM: Read Only Memory) 1002 or a program loaded from a storage device 1003 to a random access memory (RAM: Random Access Memory) 1004. In RAM1004, various programs and data required for the operation of the store location selection device are also stored. The processing device 1001, ROM1002, and RAM1004 are connected to each other via a bus 1005. An input / output (I / O) interface 1006 is also connected to the bus. Generally, the following systems can be connected to the I / O interface 1006: an input device 1007 including, for example, a touch screen, a touch pad, a keyboard, a mouse, an image sensor, a microphone, an accelerometer, a gyroscope, etc.; an output device 1008 including, for example, a liquid crystal display (LCD: Liquid Crystal Display), a speaker, a vibrator, etc.; a storage device 1003 including, for example, a magnetic tape, a hard disk, etc.; and a communication device 1009. The communication device 1009 can allow the store location selection device to communicate with other devices wirelessly or by wire to exchange data. Although the store location selection device with various systems is shown in the figure, it should be understood that it is not required to implement or have all the systems shown. More or fewer systems can be implemented or have alternatively.

[0132] In particular, according to the embodiments disclosed in the present application, the process described above with reference to the flowchart can be implemented as a computer software program. For example, the embodiments disclosed in the present application include a computer program product, which includes a computer program carried on a computer-readable medium, and the computer program includes a program code for executing the method shown in the flowchart. In such an embodiment, the computer program can be downloaded and installed from a network through a communication device, or installed from a storage device 1003, or installed from a ROM 1002. When the computer program is executed by the processing device 1001, the above-mentioned functions defined in the method of the embodiment disclosed in the present application are executed.

[0133] The store location selection device provided by the present application adopts the store location selection method in the above embodiment, which can solve the technical problem that the accuracy of store location selection decisions in related solutions is low. Compared with the prior art, the beneficial effects of the store location selection device provided by the present application are the same as the beneficial effects of the store location selection method provided by the above embodiment, and the other technical features in the store location selection device are the same as the features disclosed in the method of the previous embodiment, which will not be repeated here.

[0134] It should be understood that the various parts disclosed in this application can be implemented by hardware, software, firmware or a combination thereof. In the description of the above embodiments, specific features, structures, materials or characteristics can be combined in any one or more embodiments or examples in a suitable manner.

[0135] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art who is familiar with the present technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

[0136] The present application provides a computer-readable storage medium having computer-readable program instructions (ie, computer programs) stored thereon, and the computer-readable program instructions are used to execute the store location selection method in the above-mentioned embodiment.

[0137] The computer-readable storage medium provided in the present application may be, for example, a USB flash drive, but is not limited to electrical, magnetic, optical, electromagnetic, infrared, or semiconductor systems, systems or devices, or any combination of the above. More specific examples of computer-readable storage media may include, but are not limited to: an electrical connection with one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above. In this embodiment, the computer-readable storage medium may be any tangible medium containing or storing a program that can be used by or in combination with an instruction execution system, system or device. The program code contained on the computer-readable storage medium may be transmitted using any appropriate medium, including but not limited to: wires, optical cables, RF (Radio Frequency), etc., or any suitable combination of the above.

[0138] The above-mentioned computer-readable storage medium may be included in the store location selection device; or it may exist independently without being assembled into the store location selection device.

[0139] The above-mentioned computer-readable storage medium carries one or more programs. When the above-mentioned one or more programs are executed by the store location selection device, the store location selection device: obtains at least one location selection data based on the execution module; passes the location selection data as input data to the intelligent agent, and obtains the decision action data output by the intelligent agent based on the input data; determines the target store location selected by the intelligent agent based on the decision action data, and feeds back the target store location to the user end.

[0140] Computer program code for performing the operations of the present application may be written in one or more programming languages ​​or a combination thereof, including object-oriented programming languages ​​such as Java, Smalltalk, C++, and conventional procedural programming languages ​​such as "C" or similar programming languages. The program code may be executed entirely on the user's computer, partially on the user's computer, as a separate software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In the case of a remote computer, the remote computer may be connected to the user's computer through any type of network, including a local area network (LAN) or a wide area network (WAN), or may be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0141] The flow chart and block diagram in the accompanying drawings illustrate the possible architecture, function and operation of the system, method and computer program product according to various embodiments of the present application. In this regard, each square box in the flow chart or block diagram can represent a module, a program segment or a part of a code, and the module, the program segment or a part of the code contains one or more executable instructions for realizing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the square box can also occur in a sequence different from that marked in the accompanying drawings. For example, two square boxes represented in succession can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each square box in the block diagram and / or flow chart, and the combination of the square boxes in the block diagram and / or flow chart can be implemented with a dedicated hardware-based system that performs a specified function or operation, or can be implemented with a combination of dedicated hardware and computer instructions.

[0142] The modules involved in the embodiments described in this application may be implemented by software or hardware, wherein the name of the module does not constitute a limitation on the unit itself in some cases.

[0143] The readable storage medium provided in this application is a computer-readable storage medium, which stores computer-readable program instructions (i.e., computer programs) for executing the above-mentioned store location selection method, and can solve the technical problem that the accuracy of store location selection decisions in related solutions is low. Compared with the prior art, the beneficial effects of the computer-readable storage medium provided in this application are the same as the beneficial effects of the store location selection method provided in the above-mentioned embodiment, and will not be repeated here.

[0144] An embodiment of the present application provides a computer program product, including a computer program, which implements the steps of the above-mentioned store location selection method when executed by a processor.

[0145] The computer program product provided by the present application can solve the technical problem that the accuracy of store location selection decisions in related solutions is low. Compared with the prior art, the beneficial effects of the computer program product provided by the embodiment of the present application are the same as the beneficial effects of the store location selection method provided by the above embodiment, which will not be repeated here.

[0146] The above are only preferred embodiments of the present application, and are not intended to limit the patent scope of the present application. Any equivalent structure or equivalent process transformation made using the contents of the present application specification and drawings, or directly or indirectly applied in other related technical fields, are also included in the patent processing scope of the present application.

Claims

1. A store location selection method, characterized in that: Applied to a store location selection system, the store location selection system includes an execution module, an agent and a training module, and the method includes the following steps: Acquire at least one site selection data based on the execution module; Passing the site selection data as input data to the agent, and obtaining decision action data output by the agent based on the input data; The target store location selected by the agent is determined based on the decision action data, and the target store location is fed back to the user end.

2. The method according to claim 1, characterized in that Before the step of transmitting the site selection data as input data to the intelligent agent and obtaining the decision action data output by the intelligent agent based on the input data, the method includes: Acquiring scene state data based on the training module; Passing the scene state data as input data to the agent, and obtaining decision action data output by the agent based on the input data; Calculate the Q value based on the scene state data and the decision action data and based on a preset formula; Determine a reward value according to the scene state data and the reward function, and determine a target Q value according to the reward value and the Q value; When the target Q value and the difference between the Q value are continuously iterated for a target number of times and are both less than a preset threshold, the agent is output.

3. The method according to claim 2, characterized in that When the target Q value and the difference between the Q value are continuously iterated for a target number of times and are both less than a preset threshold, before the agent is output, the method includes: Update the hyperparameters of the agent according to the Q value and the target Q value; The step of obtaining scene status data is performed according to the updated agent.

4. The method according to claim 1 or 2, characterized in that: Before the step of obtaining the decision action data output by the agent based on the input data, the method includes: The agent determines a target strategy according to the received input data; Through the target strategy, the corresponding decision action data is selected and output based on the input data, wherein the decision action data includes rating the store locations as golden locations, high-quality locations, ordinary locations, and worthless locations.

5. The method according to claim 2, characterized in that The step of determining the reward value according to the scene state data and the reward function comprises: Acquire various business indicator data in the scenario status data and corresponding weight values, wherein the business indicator data includes cost item indicators, income item indicators and risk item indicators; According to the instant reward function, business indicator data and corresponding weight values, the instant reward for each business indicator data is calculated; The reward value is calculated based on the reward function and the instant reward of each business indicator data.

6. The method according to claim 2, characterized in that Before the step of acquiring scene state data based on the training module, the method further includes: Collect and integrate store locations, competitive product information, community information, administrative district information, and main road information, and determine them as the original data set; Using geographic information coding technology to convert the address information in the original data set into longitude and latitude coordinates, and mark the store locations, competitive product locations and community locations on the map; According to the preset business circle range, various business indicator data corresponding to the original data set within the radius are calculated and determined as the scene state data.

7. The method according to claim 6, characterized in that After the step of calculating various business indicator data within the radius according to the preset business circle range and determining them as the scene state data, the method further includes: Sorting the scene state data according to the numerical value, and determining the first quartile and the third quartile, wherein the first quartile is the numerical value at the 25% position of the sorting, and the third quartile is the numerical value at the 75% position of the sorting; Determine a fourth quartile range according to the first quartile and the third quartile, and determine an abnormal range according to the fourth quartile range; The data in the abnormal range is cleared according to the abnormal range, and the processed scene state data is output.

8. The method according to claim 1, characterized in that After the step of determining the target store location selected by the agent based on the decision action data and feeding back the target store location to the user terminal, the method further includes: Predicting potential customer traffic at the target store locations; Based on the prediction results and the corresponding cost-benefit analysis, business recommendations for the target store locations are generated.

9. A store location selection device, characterized in that: The store site selection device comprises: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the computer program is configured to implement the steps of the store site selection method according to any one of claims 1 to 8.

10. A storage medium, characterized in that: The storage medium is a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the steps of the store location selection method according to any one of claims 1 to 8 are implemented.

Images