AI-Based Store and Commercial Area Improvement Analysis System Using Anonymous Feedback

Abstract

The present invention relates to an AI-based store and commercial area improvement analysis system that analyzes customer opinions collected based on anonymous feedback to derive items for improvement of a store or commercial area and provides the results to a store operator or commercial area manager. The system according to the present invention induces a user terminal to connect through identification information of media placed in an offline store, and receives anonymous feedback by distinguishing between positive items and improvement items regarding the visited store. The collected anonymous feedback undergoes text analysis including sentiment analysis and core keyword extraction, and is structured into recurring problem types through keyword mapping according to an industry classification system and cluster analysis among similar feedback. The system calculates the priority of improvement items using values ​​corresponding to feedback frequency, sentiment analysis results, store operation items, or customer revisit probability, and generates improvement reports based on this. Furthermore, it evaluates feedback quality to generate improvement events designed to induce customer engagement or operational improvements, and forms a feedback structure where analysis results are reflected back into standard setting by dynamically updating priority calculation criteria through the analysis of changes before and after the event execution. Accordingly, the present invention provides the effect of improving the reliability of analysis data while maintaining customer participation, and systematically deriving improvement tasks at the store and commercial district levels.

Description

Technology Field

[0001] The present invention relates to store operation and commercial area analysis technology, and more specifically, to an AI-based store and commercial area improvement analysis system that analyzes anonymously collected customer feedback using artificial intelligence to identify items requiring improvement at the store or commercial area level, calculates their priorities, and provides them in the form of a report.

[0002] In particular, the present invention belongs to the field of technology that structures free-form opinions collected from customers visiting offline stores through text analysis, sentiment analysis, and pattern analysis, and supports store operators or commercial area managers in utilizing the results for practical operational improvements. Background Technology

[0003] Recently, in the operating environment of offline stores and commercial districts, the quality of customer experience has a direct impact on sales, return rates, and brand credibility. Accordingly, there are continuous attempts to collect and analyze customer opinions and reflect them in store operations.

[0004] Traditionally, methods such as customer satisfaction surveys, questionnaires, star ratings, and review boards have been primarily used to collect customer opinions. However, these methods have the following limitations.

[0005] First, in a structure where customers submit opinions based on their real names or accounts, problems arise where the submission of negative feedback is discouraged or responses remain merely perfunctory.

[0006] Second, opinions collected in a free-form format often amount to nothing more than a list of individual cases, making it difficult for store operators to systematically analyze them to identify structural problem types or priorities for improvement.

[0007] Third, existing simple statistical-based analysis or keyword frequency analysis methods fail to adequately reflect the intensity of emotions, diversity of expressions, and recurring problem patterns, thus having limitations in providing analysis results that can be utilized for actual operational improvement.

[0008] Fourth, even if customer feedback is analyzed, there is rarely a feedback loop structure provided to evaluate whether the analysis results led to actual improvements in store operations or to reflect those results back into the analysis criteria.

[0009] As a result, despite possessing a large amount of customer feedback, there is a persistent problem in which store or commercial district operators are unable to utilize it as an objective and repeatable criterion for determining improvement. The problem to be solved

[0010] The present invention has been devised to solve the problems of the prior art as described above, and aims to solve the following technical problems.

[0011] First, the objective is to enable the more reliable collection of diverse customer experience data, including negative opinions, by encouraging customers visiting offline stores to provide free, anonymous feedback.

[0012] Second, the task is to derive structured improvement items rather than a mere list of individual opinions by performing text preprocessing, sentiment analysis, and key keyword extraction on the collected anonymous feedback, and by identifying recurring problem types.

[0013] Third, the task is to enable the reasonable calculation of the relative importance and priority among improvement items by comprehensively considering the frequency of feedback, sentiment analysis results, and values ​​corresponding to store operations or the likelihood of customer revisit.

[0014] Fourth, the task is to ensure that the results of data analysis lead to actual operational decision-making by providing improvement reports that store operators or commercial area managers can understand and utilize based on the priority of the calculated improvement items.

[0015] Fifth, the task is to provide an improvement analysis system in which analysis criteria dynamically adapt to operational results by analyzing performance before and after carrying out event or reward-based operational improvement activities based on improvement items, and reflecting the results back into the priority calculation criteria. means of solving the problem

[0016] According to one embodiment of the present invention, an AI-based store and commercial area improvement analysis system is disclosed, which analyzes customer opinions collected based on anonymous feedback to derive improvement items for a store or commercial area and provides the results to a store operator or commercial area manager, comprising: a connection induction unit that induces a connection of a user terminal through identification information included in a medium placed in an offline store; an anonymous feedback input unit configured to input positive items and improvement items regarding a visited store separately from each other from the user terminal; a data storage unit that collects and stores the anonymous feedback; a text analysis unit that generates text preprocessing results, sentiment analysis results, and core keyword extraction results for the collected anonymous feedback; a pattern analysis unit that derives problem types that appear repeatedly based on the result of mapping the core keywords according to an industry classification system and the clustering result between similar feedback; and a priority calculation unit and a report generation unit that calculate the priority of improvement items using a value corresponding to the frequency of appearance of feedback, a value corresponding to the sentiment analysis result, and a value corresponding to a store operation item or the possibility of customer revisit as input variables, and generates and provides an improvement report based on the calculated priority.

[0017] According to one embodiment of the present invention, the anonymous feedback input unit is configured to restrict the submission of feedback when positive items and improvement items are not input, thereby preventing biased, unidirectional feedback input; the text analysis unit generates a plurality of evaluation information including information indicating a difference in sentiment distribution, information indicating diversity of input expressions, or information corresponding to the volume of described content, by comparing sentiment analysis results derived for each of the positive items and improvement items; and the priority calculation unit is configured to adopt only feedback whose feedback reliability value derived by comparing the plurality of evaluation values ​​is greater than or equal to a standard as input for priority calculation. An AI-based store and commercial area improvement analysis system is disclosed.

[0018] According to one embodiment of the present invention, the AI-based store and commercial area improvement analysis system further comprises a reward control unit that induces access of a user terminal through a QR code or deep link identifier and calculates a feedback quality value based on an evaluation value corresponding to whether anonymous feedback input is completed, the number of input items, keyword diversity, or the clarity of the sentiment analysis result; wherein the reward control unit generates an improvement event to induce customer participation or store operation improvement based on the feedback quality value or the priority calculation result, and the improvement event is configured such that at least one of the purpose of the event, the target customer group, the participation condition, or the period is set in correspondence with the problem type or improvement item.

[0019] According to one embodiment of the present invention, an AI-based store and commercial area improvement analysis system is disclosed, characterized in that the priority calculation unit selectively applies different ranges of application ratios to each of the values ​​corresponding to the frequency of occurrence of feedback, the value corresponding to the result of sentiment analysis, and the value corresponding to the possibility of store operation items or customer revisit, and the application ratios are selected in a normalized state according to relative importance, thereby configuring the priority of improvement items to be calculated according to the result of the combination of the values.

[0020] According to one embodiment of the present invention, the AI-based store and commercial area improvement analysis system comprises: a common problem identification unit that classifies keywords or problem types commonly mentioned by multiple stores among feedback collected from multiple stores within the same commercial area as common commercial area problems; and a comparative analysis unit that aggregates satisfaction values ​​and key keywords of multiple stores belonging to the same industry to generate industry-specific benchmark data, and compares the value of a specific store with the benchmark data to identify relative strengths and elements requiring improvement, wherein the results are configured to be included in a commercial area analysis report.

[0021] According to one embodiment of the present invention, the AI-based store and commercial area improvement analysis system further comprises a performance analysis unit that analyzes changes in additional anonymous feedback, store operation values, or customer revisit values ​​collected before and after the performance of an improvement event according to claim 3, wherein the performance analysis unit determines whether the change in values ​​before and after the performance of the event is within a preset threshold range or is determined to be a significant change compared to a reference period, and wherein the priority calculation unit updates the selection criteria for the application ratio applied to at least one of a value corresponding to the frequency of appearance of feedback, a value corresponding to the sentiment analysis result, or a value corresponding to a store operation item or the possibility of customer revisit based on the determination result, thereby configuring the priority calculation criteria of the improvement item to dynamically adapt according to the event result. Effects of the invention

[0022] The store and commercial area improvement analysis system using AI-based anonymous feedback analysis according to the present invention provides the following effects.

[0023] First, by inducing user terminal access through media placed in offline stores and configuring feedback input anonymously, distinguishing between positive and improvement items, it is possible to reliably collect feedback based on actual experiences, including negative opinions, while minimizing the psychological burden on customers.

[0024] Second, by performing text preprocessing, sentiment analysis, and key keyword extraction on the collected anonymous feedback, and deriving recurring problem types based on clustering results among similar feedback, it is possible to systematically identify structured improvement items rather than a simple listing of individual opinions.

[0025] Third, by calculating the priority of improvement items through a comprehensive consideration of values ​​corresponding to the frequency of feedback, values ​​corresponding to sentiment analysis results, and values ​​corresponding to store operation items or the likelihood of customer revisit, store or commercial district operators can determine elements requiring improvement based on objective criteria.

[0026] Fourth, by generating and providing improvement reports based on the calculated priorities, the analysis results are not merely limited to providing data but are provided in a form that can be directly utilized for actual store operation decision-making.

[0027] Fifth, by generating events corresponding to improvement items and analyzing changes in additional feedback, store operation values, or customer revisit values ​​collected before and after the execution of the event, and reflecting the results in the priority calculation criteria, a feedback-based improvement analysis structure can be provided in which the analysis criteria dynamically adapt according to operational results.

[0028] Sixth, through such a structure, the present invention provides the effect of enabling customer experience analysis and operational improvement at the store or commercial district level to be implemented as an iterative and cumulative improvement process rather than a one-time analysis. Brief explanation of the drawing

[0029] FIG. 1 is a block diagram schematically illustrating the overall configuration of an AI-based store and commercial area improvement analysis system according to one embodiment of the present invention. FIG. 2 is a flowchart illustrating a process in which anonymous feedback is collected and feedback reliability is determined based on a plurality of evaluation values ​​according to an embodiment of the present invention. FIG. 3 is a diagram illustrating a concept in which the priority of improvement items is calculated using input variables corresponding to the frequency of feedback appearance, sentiment analysis results, and the possibility of store operation or customer revisit, according to one embodiment of the present invention. FIG. 4 is a flowchart illustrating a process in which the performance before and after the execution of an improvement event is analyzed according to an embodiment of the present invention, and the results are reflected in the update of the priority calculation criteria. Specific details for implementing the invention

[0030] The following detailed description of the invention refers to the accompanying drawings, which illustrate specific embodiments in which the invention may be practiced. These embodiments are described in sufficient detail to enable those skilled in the art to practice the invention. It should be understood that various embodiments of the invention are different but need not be mutually exclusive. For example, specific shapes, structures, and characteristics described herein may be implemented in other embodiments without departing from the spirit and scope of the invention in relation to one embodiment.

[0031] Furthermore, it should be understood that the location or arrangement of individual components within each disclosed embodiment may be changed without departing from the spirit and scope of the invention. Accordingly, the following detailed description is not intended to be taken in a limiting sense, and the scope of the invention is limited only by the appended claims, including all equivalents thereof, provided appropriately described. Similar reference numerals in the drawings refer to the same or similar functions across various aspects.

[0032] Meanwhile, throughout this specification, when a part is described as “comprising” a certain component, this means that, unless specifically stated otherwise, it does not exclude other components but may include additional components. Furthermore, the “part” for a component as used in this specification performs at least one function or operation. And the “part” may perform the function or operation by hardware, software, or a combination of hardware and software.

[0033] Terms including ordinal numbers, such as first, second, etc., may be used to describe various components, but these components are not limited by the aforementioned terms. The aforementioned terms are used solely for the purpose of distinguishing one component from another.

[0034] In this specification, terms such as “comprising” are intended to indicate the existence of the features, numbers, steps, actions, components, parts, or combinations thereof described in the specification, and should be understood as not precluding the existence or addition of one or more other features, numbers, steps, actions, components, parts, or combinations thereof. When a component is referred to as being “connected” to another component, it should be understood that it may be directly connected to or coupled with the other component, or that there may be other components in between.

[0035] In this document, network functions, artificial neural networks, and neural networks may be used interchangeably. The various embodiments described herein may be implemented, for example, in computer-readable recording media and storage media using software, hardware, or a combination thereof.

[0036] According to hardware implementation, the embodiments described herein may be implemented using at least one of ASICs (application specific integrated circuits), DSPs (digital signal processors), DSPDs (digital signal processing devices), PLDs (programmable logic devices), FPGAs (field programmable gate arrays), processors, controllers, microcontrollers, microprocessors, and other electrical units for performing functions. In some cases, the embodiments described herein may be implemented as the processor itself of an electronic device.

[0037] The processor may be composed of one or more cores and may include processors for data analysis and deep learning, such as a server's Central Processing Unit (CPU), General Purpose Graphics Processing Unit (GPGPU), and Tensor Processing Unit (TPU). The processor may read a computer program stored in memory and perform data processing for machine learning according to one embodiment of the present invention.

[0038] The processor can perform computations for training neural networks, such as processing input data for training in deep learning (DL), extracting features from input data, calculating errors, and updating neural network weights using backpropagation. At least one of the processor's CPU, GPGPU, and TPU can handle the training of network functions. For example, the CPU and GPGPU can work together to handle the training of network functions and data classification using network functions.

[0039] Furthermore, in describing the present invention, if it is determined that a detailed description of related known functions or configurations may unnecessarily obscure the essence of the invention, such detailed description is abbreviated or omitted.

[0040] Hereinafter, preferred embodiments of the present invention will be described in more detail with reference to the drawings.

[0042] FIG. 1 is a block diagram schematically illustrating the overall configuration of an AI-based store and commercial area improvement analysis system according to one embodiment of the present invention.

[0043] The system according to the present invention is configured to collect anonymous feedback from customers visiting an offline store, analyze the feedback based on artificial intelligence to derive items for improvement of the store or commercial area, and provide the results to the store operator or commercial area manager.

[0044] The system first induces a user terminal to connect through media placed in the offline store. The media contains identification information recognizable by the user terminal, and when the user recognizes or selects such identification information, the user terminal connects to the system. Through this, the system can initiate feedback input based on the customer's actual visit experience within the store.

[0045] When a user terminal connects to the system, the system provides an anonymous feedback input environment that includes input fields separating positive and improvement items regarding the visited store. The user inputs content corresponding to each item according to the input environment, and the system receives this input anonymously to alleviate the user's psychological burden and encourage the submission of more honest opinions.

[0046] The system stores the received anonymous feedback in the data storage unit.

[0047] The stored feedback data is managed for subsequent analysis procedures and can be collected cumulatively from multiple users.

[0048] Subsequently, the system performs text analysis on the stored feedback.

[0049] The system performs text preprocessing on the text included in the collected anonymous feedback to enable comparison and pattern derivation in subsequent analysis stages by organizing redundant expressions, meaningless symbols, or descriptions that do not contribute to analysis, and to reduce analysis deviations caused by differences in expression forms.

[0050] In addition, the system produces a sentiment analysis result for the preprocessed text that indicates whether the expression included in the feedback is closer to a positive, negative, or neutral tendency, and the sentiment analysis result is subsequently used as an input value for determining the reliability of the feedback or the priority of improvement items.

[0051] The core keywords and sentiment analysis results generated in this way are passed to the pattern analysis stage.

[0052] The system sorts the extracted core keywords in correspondence with the industry classification system, thereby processing items that may have different meanings depending on the characteristics of the industry, even if they are the same expression.

[0053] In addition, the system groups similar feedbacks into the same problem category based on judgment results corresponding to the similarity of expression, commonality of mentioned subjects, or similarity of occurrence context among multiple feedbacks containing the aforementioned aligned keywords.

[0054] Here, the aforementioned problem categories are not limited to problem types observed individually in a single store; if they appear repeatedly in multiple stores classified within the same commercial area, they may be interpreted as items requiring improvement at the commercial area level. Accordingly, the system can structure problem types from the perspective of at least one of the store unit or commercial area unit analysis.

[0055] At this point, the system performs a judgment procedure to distinguish whether the problem type is due to individual characteristics at the store level or corresponds to a structural problem repeatedly observed in multiple stores within the same commercial area.

[0056] Specifically, the system refers to at least one of the following as a judgment criterion: whether the same problem type or similar core keywords are repeatedly derived from feedback from multiple stores during a certain analysis period; whether the pattern of occurrence of problem types appears similarly among stores belonging to the same industry; or whether the problem type appears in relation to the operating environment of the entire commercial district rather than being limited to a specific store.

[0057] Based on these judgments, the system classifies problem types into individual store improvement items or common improvement items at the commercial district level, thereby ensuring that the scope of improvement to which the analysis results should be applied is clearly distinguished.

[0058] Through this, the system does not simply list individual user opinions, but identifies problem types that appear repeatedly in multiple feedbacks and structures these problem types into items requiring improvement at the store or commercial district level.

[0059] Subsequently, the system performs a priority calculation step for the previously identified problem types to determine the relative importance of each problem type as a target for improvement.

[0060] In this process, the system uses at least some of the following as input values: a value corresponding to the frequency of occurrence, which indicates the extent to which an individual problem type is repeatedly mentioned in the entire feedback; a value corresponding to the result of sentiment analysis, which reflects the emotional tendency of expressions related to the problem type; and a value corresponding to store operation items or customer revisit potential, which indicates the impact of the problem type on store operation efficiency or customer revisit potential.

[0061] The system organizes the above input values ​​so that they can be compared under the same criteria, determines the relative importance between each problem type, and calculates the priority of improvement items based on the result of that determination.

[0062] In this case, the above values ​​are not limited to predetermined absolute figures, but are utilized in response to relative differences between multiple problem types, whether criteria are met, or trends of change, thereby allowing priorities to be calculated flexibly according to the operational characteristics of the store or commercial area.

[0063] In addition, the above priority calculation may be applied to at least one of the cases where it is to determine the degree of improvement required for individual stores or to determine the importance of problem types that commonly appear in multiple stores within the same commercial area, and the scope of analysis may be selectively set depending on the application target.

[0064] Once the priority is determined, the system generates an improvement report based on the results.

[0065] The improvement report includes high-priority improvement items, which can be used by store operators or commercial area managers to make decisions for improving store operations.

[0066] The system provides the generated improvement reports to store operators or commercial area managers, ensuring that the results of customer feedback analysis lead to actual operational improvements.

[0067] As such, the system according to the present invention systematically supports the improvement of store and commercial area operations through a series of procedures including the collection of anonymous feedback, text-based analysis, identification of problem types, calculation of priorities, and provision of reports.

[0069] FIG. 2 is a flowchart supporting the “feedback reliability judgment,” and the configuration of Paragraph 2 will be explained in detail below following the flow of FIG. 2.

[0070] In this embodiment, when a user terminal is connected, the system provides an input environment for anonymous feedback input, wherein the input environment is configured such that a positive item input area and an improvement item input area are separated from each other.

[0071] The user enters text corresponding to each input area and performs a submission action, and the system first checks whether both positive and improvement items have been entered at the time of submission.

[0072] At this point, the system prevents submission from being completed if any single item is left unentered, thereby blocking feedback skewed in a single direction—either positive or improvement—from flowing into the reliability judgment stage.

[0073] These submission restrictions are not merely a feature for user convenience, but serve as a prerequisite to prevent evaluation information generated in subsequent stages from being distorted by biased input.

[0074] If submission is permitted, the system performs preprocessing and sentiment analysis on the positive item text and improvement item text, respectively, through the text analysis unit, and generates multiple evaluation information based on the results.

[0075] Here, preprocessing can be performed by cleaning up elements that do not contribute to analysis, such as spaces, symbols, and repetitive characters, and arranging them into a form comparable at the sentence or phrase level, in order to reduce instability in comparison results caused by differences in expressions of the same meaning.

[0076] Sentiment analysis is configured to produce results indicating whether the expressions included in the text are closer to a positive, negative, or neutral tendency, and these results are generated in separate forms for positive items and improvement items, respectively.

[0077] The system compares the sentiment analysis results generated separately in this way to generate evaluation information to determine whether sentiment trends are appropriately distinguished between positive items and improvement items.

[0078] For example, the system considers a case where a positive tendency predominates in positive items and a negative or improvement request tendency predominates in improvement items as a normal response trend; however, if both items converge to the same emotional trend or appear abnormally reversed in opposite directions, it may classify it as a state where the authenticity of the response or input accuracy is likely low.

[0079] In this case, the system does not conclude the emotional tendency itself as an absolute value, but constructs evaluation information as information corresponding to whether the relative difference or degree of separation within the same user feedback satisfies a certain standard.

[0080] In addition, the system generates evaluation information corresponding to the diversity of expression.

[0081] Expression diversity may include characteristics corresponding to the breadth of information provided by the input text, such as whether expressions of the same meaning are repeated, the scope of meaningful keywords included, or whether descriptions of specific subjects, situations, or reasons are included.

[0082] The system can generate evaluation information to determine inputs in which identical or similar expressions are repeated without descriptions of meaningful objects or experiences as low variety, and inputs in which expressions are expanded by including specific objects, situations, or reasons as high variety.

[0083] In particular, in this invention, even if both positive and improvement items are submitted, if both items are entered as variations of the same phrase or contain little semantic information, their value as actual experience-based feedback may be low; therefore, diversity evaluation functions as important input information for reliability judgment.

[0084] The above evaluation information is not intended to assign absolute scores or numerical values ​​to individual feedback, but rather consists of information representing the characteristics of the input text to compare and classify multiple feedbacks within the same analysis frame.

[0085] The system also generates evaluation information corresponding to the description length.

[0086] Descriptive length can be configured to include information corresponding to whether the input contains a minimum unit of meaning, whether a sentence structure containing core keywords is maintained, or whether it is excessively short or excessively verbose, thereby impairing analysis stability, rather than simply the number of characters itself.

[0087] For example, since extremely short inputs may lack specificity of experience, and conversely, excessive inputs that repeat the same meaning may increase noise and be disadvantageous for subsequent analysis, the system may not fix length-related evaluation information to a single value but may instead configure it as comparative information, such as whether it exists within an acceptable range or whether the length balance between positive items and improvement items is significantly disrupted.

[0088] In this case, since length evaluations can vary depending on individual users or industry characteristics, it is desirable for the system to generate length-related evaluation information in a form corresponding to whether criteria are met or relative comparison results, rather than limiting it to absolute numerical values.

[0089] The system derives a feedback reliability value by combining at least some of the evaluation information corresponding to the difference in emotion distribution, the evaluation information corresponding to expression diversity, and the evaluation information corresponding to narrative length.

[0090] In this context, "synthesis" does not refer to an arbitrary subjective combination, but is structured in a way that allows for adoption or exclusion decisions based on pre-established criteria, after organizing each piece of evaluation information to enable comparison within the same judgment frame by considering the direction in which it contributes to the reliability judgment.

[0091] For example, the system may classify feedback as reliable feedback if the evaluation information regarding differences in emotional distribution is above the standard, the diversity evaluation information is above the standard, and the length evaluation information is within the acceptable range, and conversely, if any one falls below the standard, it may derive a low reliability value or determine that the reliability value is below the standard.

[0092] In addition, considering that conflicts may occur between evaluation information, the system may be configured to exceptionally accept cases where other evaluation information sufficiently strongly meets the criteria even if specific evaluation information does not meet the criteria, or conversely, to exclude cases where specific evaluation information strongly exhibits an abnormal pattern even if other evaluation information is satisfactory.

[0093] The existence of such exception handling supports the fact that reliability judgment does not stop at simple format checking but functions as a practical filter for judging the quality of input text.

[0094] When a confidence value is derived, the system determines whether the confidence value is above a preset threshold, and adopts only feedback that is above the threshold as input for the priority calculation step.

[0095] Feedback deemed below the standard may be stored in the data storage unit, but may be managed separately so as not to be directly reflected in priority calculation, and may be utilized only as statistical information for quality improvement as necessary.

[0096] As a result, the system of the present embodiment is configured such that, as illustrated in FIG. 2, through a series of procedures including prevention of bias in the input stage, generation of multiple evaluation information based on text analysis, comparison between evaluation information and derivation of reliability values, determination of whether it is above a standard, and adoption of priority calculation input, the quality of anonymous feedback data is controlled while feedback that can contribute to actual operational improvement is prioritized and reflected in the analysis.

[0098] In one embodiment of the present invention, the AI-based store and commercial area improvement analysis system is configured to link anonymous feedback to a control step for inducing customer participation after anonymous feedback is collected and reliability is determined.

[0099] In other words, the system does not stop at simply storing the input and analysis results of anonymous feedback or outputting them as reports, but extends the feedback processing flow to enable additional input and verification corresponding to the analysis results.

[0100] To this end, the system identifies the connection path of the user terminal and checks whether the connection has been successfully completed as a feedback input. Here, the connection path can be distinguished based on identification information provided to induce the connection of the user terminal, such as a QR code or a deep link identifier, and the system links the connection path information with the completion status of the feedback input and uses it as input information for the subsequent participation control stage.

[0101] When feedback input is completed, the system calculates a feedback quality value corresponding to the input quality of the anonymous feedback. In this case, the feedback quality value is not used to directly determine the veracity of the feedback content, but rather as information to assess whether the input act was carried out faithfully to a level usable for analysis and encouraging subsequent participation.

[0102] For example, the system may construct a feedback quality value by referring to information corresponding to at least one of the following: whether actual input exists for each positive item and improvement item, the number of input items, the range containing core keywords, or whether the sentiment analysis results appear separated into different trends. It is desirable that such feedback quality values ​​are not limited to absolute scores but are set in a form corresponding to whether criteria are met or the results of a relative comparison with other feedback within the same period.

[0103] In this embodiment, the system is configured not to terminate the processing after feedback input with simple data collection, but to generate improvement events to induce customer participation or store operation improvement by referencing the feedback quality value or the priority calculation result of the previously calculated improvement item as input information.

[0104] In this case, the system does not generate improvement events by inducing random participation or providing uniform rewards, but is designed with a participation control structure that logically corresponds to the problem types or improvement items derived from the analysis results.

[0105] Specifically, prior to automatically generating improvement events, the system sets classification result information calculated by problem type, priority level information of the corresponding problem type, and distribution information of feedback quality values ​​as input variables, and calculates multiple decision parameters required for event design in stages based on the above input variables.

[0106] First, based on the classification results of the problem type, the system determines whether the problem type is a single-store issue or a common issue across a commercial district, and accordingly sets the scope of application of the event to either a store unit or a commercial district unit.

[0107] Next, the system determines the purpose of the event by referring to pre-set priority level information corresponding to the aforementioned problem type. If the priority level is above a pre-set upper standard, it is classified as a verification-type event aimed at collecting additional information or verifying improvement effects; if it is at an intermediate level, it is classified as a confirmation-type event aimed at supplementing problem recognition; and if it is at a lower level, it is configured to be classified as an observation-type event for monitoring.

[0108] Next, the system analyzes the distribution information of feedback quality values ​​to determine the target customer group for the event. At this time, the system sets a set of users whose feedback quality values ​​are judged to be above a pre-set standard as a primary candidate group, and selectively includes only users with access paths or input history that are highly correlated with the problem type, thereby ensuring the reliability of the analysis of the event participation results.

[0109] In addition, the system automatically sets participation conditions by referring to the problem categories classified by the above problem type and the attributes of the feedback content associated with that problem type.

[0110] Here, the nature of a problem type refers to whether the problem type is directly associated with any of the following: contextual information, temporal elements, service touchpoint elements, or operational process elements; and the system is configured to select the type of participation condition according to predefined rules based on the problem category to which the problem type belongs.

[0111] Specifically, in generating improvement events, the system is configured to set the problem type classification result, the priority level of the corresponding problem type, and the distribution of feedback quality values ​​as input variables, and to determine the design elements of the event step by step by sequentially referencing said input variables.

[0112] First, the system determines the type of participation condition based on the problem type classification results.

[0113] If the problem type is classified into a category associated with temporal factors such as time of day, waiting time, or congestion, the system sets at least one of the following as a participation condition: whether a visit was made during a specific time period, the level of perceived waiting time, or satisfaction by time period input.

[0114] On the other hand, if the problem type is classified into a category associated with service touchpoint elements, such as employee response, service attitude, or clarity of explanation, the system sets at least one of a confirmation input, a multiple-choice evaluation response, or an additional descriptive input for a specific response situation as a participation condition.

[0115] In addition, if the problem type is classified into a category associated with operational process elements such as order flow, payment process, or store flow, the system sets the experience with the relevant process or the selection of inconvenience factors as participation conditions.

[0116] Next, the system calculates the event period by referring to the frequency of occurrence of problem types and the time of their most recent appearance.

[0117] It is configured so that the event period corresponding to the occurrence pattern of the problem type is automatically determined by setting short-term events with a limited duration for problem types observed intensively over a short period, and setting repetitive events spanning multiple periods for problem types observed repeatedly for a period longer than a certain duration.

[0118] Next, the system determines the nature of the event by referring to the priority level of the problem type.

[0119] It is configured to generate verification events to verify improvement effects through additional input for problem types assessed as high priority, and to generate confirmation events to supplement the problem recognition level for problem types with medium priority.

[0120] In addition, the system analyzes the distribution of feedback quality values ​​to determine the target customer group for the event.

[0121] In this case, the system selects only users whose feedback quality value is judged to be above a standard as the target customer group for the event, or limits the event targets to only users who entered through access paths associated with the problem type, thereby ensuring that the results of event participation can be utilized as reliable input in the subsequent analysis stage.

[0122] Accordingly, the system according to the present invention designs improvement events based on the analysis results of anonymous feedback and utilizes the execution results of the improvement events as input to the performance analysis stage, thereby forming an iterative analysis structure in which priority calculation criteria and analysis criteria are updated according to actual operation results.

[0123] With this structure, the system can selectively apply different judgment criteria even for the same type of problem depending on the time of analysis, event execution results, or changes in the operating environment, and is configured so that analysis criteria are not maintained in a fixed state but are gradually adjusted in response to operational results.

[0124] In other words, the present invention provides a self-adaptive analysis structure in which feedback analysis, customer engagement induction, improvement event execution, performance verification, and standard updates are cyclically connected, thereby having a technical effect that is clearly distinguished from conventional statistical analysis systems or one-time report provision methods in which analysis is performed based on predetermined fixed weights or uniform judgment criteria.

[0125] Accordingly, the system reduces the generation of unnecessary events and the identification of meaningless improvement items, while providing the effect of continuously improving analysis accuracy and operational suitability by prioritizing the reflection of problem types whose improvement effects have been verified in actual operating environments.

[0127] FIG. 3 is a schematic diagram illustrating the judgment path selection and analysis value combination structure of a priority calculation unit according to an embodiment of the present invention. Hereinafter, with reference to FIG. 3, a specific judgment procedure for calculating priority for each problem type, the definition of input values, the order of operations, and constraints will be explained step by step.

[0128] In this embodiment, for each problem type derived from the pattern analysis unit, the system defines the input information to be used for priority calculation as a set of independent analysis values.

[0129] The set of analysis values ​​includes frequency values ​​corresponding to the frequency of feedback occurrence by problem type, sentiment values ​​corresponding to the average intensity and recurrence persistence of the sentiment analysis results of feedback related to the problem type, and operational impact values ​​indicating the extent to which the problem type corresponds to store operation items or predefined associated items with customer revisit potential.

[0130] The frequency value is calculated based on the total number of times the problem type appeared within the analysis period, the rate of concentrated appearance within the most recent reference period, and whether it appeared consecutively.

[0131] The sentiment value is calculated by managing the sentiment analysis results of feedbacks belonging to the same problem type as a set, and reflecting the average value of negative sentiment intensity, the number of times intensity exceeding a threshold was repeated, and the duration.

[0132] The operational impact value is defined as a value assigned based on whether the problem type maps to predefined operational impact items, such as order delays, customer service complaints, payment errors, and factors hindering revisiting.

[0133] Since the above analysis values ​​differ in units and distribution, the system manages them in a normalized state to enable relative comparison for the same problem type.

[0134] The system does not immediately combine the above analysis values, but instead performs a decision path selection step in advance to determine which analysis value to apply first.

[0135] To this end, the system generates evaluation information on the temporal distribution of occurrence frequency, evaluation information on the intensity distribution and repetition persistence of sentiment analysis results, and evaluation information on operational impact for each problem type.

[0136] Information on the temporal distribution of occurrence frequency is defined as information for determining whether a problem type occurs evenly over the entire analysis period or occurs intensively during a specific period, and the system distinguishes between short-term intensive type and continuous type by determining whether the proportion of occurrence within the most recent reference period is above the concentration threshold or whether it occurs continuously, based on the timing of feedback occurrence for each problem type.

[0137] Information regarding the intensity distribution and repetition persistence evaluation of sentiment analysis results is defined as information used to determine whether negative or improvement-requesting emotional responses are repeated above a certain level; the system manages the sentiment analysis results of feedbacks belonging to the same problem type as a set and determines whether the dissatisfaction is temporary or structural by comparing the number of times the emotional intensity remains above a standard range and the duration thereof.

[0138] Operational impact assessment information is defined as information used to determine whether a problem type is directly related to store operation items or the likelihood of customer revisit, and this determination is made based on a predefined association table or set of rules.

[0139] The system compares whether any of the aforementioned evaluation information preferentially satisfies a pre-set judgment criterion, and selects one of the occurrence frequency-centered judgment path, the sentiment impact-centered judgment path, or the operational impact-centered judgment path based on the comparison result. This selection of a judgment path is a meta-decision step performed prior to the combination of analysis values, and determines the type of priority application criterion to be applied in the subsequent priority calculation.

[0140] When a judgment path by problem type is selected, the system sets the analysis value corresponding to the selected judgment path as the primary application criterion for priority calculation, and sets the remaining analysis values ​​as secondary criteria.

[0141] Subsequently, the system combines auxiliary criteria stepwise based on the intermediate priority values ​​calculated by the primary application criteria. At this stage, the combination of auxiliary criteria is not a simple weighted sum, but is permitted only within a limited range to ensure that the auxiliary criteria do not reverse the relative order between problem types determined by the primary application criteria.

[0142] Specifically, the system defines the range within which correction based on auxiliary criteria is permitted as a pre-set correction limit range, and applies upper and lower limit conditions to ensure that the correction result does not exceed the boundaries of the upper and lower groups formed by the priority application criteria. Accordingly, the auxiliary criteria serve to refine alignment within the same group, but are constrained so as not to invalidate the results of the decision path selection.

[0143] The final priority result calculated according to the above procedure is transmitted to the improvement report generation unit and the improvement event design unit. In addition, the event execution result is analyzed again in the performance analysis step shown in Fig. 4 and reflected in the judgment path selection criteria or correction limit range, thereby forming an adaptive priority calculation structure in which the priority calculation criteria are not fixed but are gradually adjusted according to the actual operation results.

[0144] In order to specifically explain the process by which the above-mentioned priority calculation structure is actually implemented and operated, an operational example targeting multiple cafe stores within the same commercial area is presented below.

[0145] In this embodiment, the system places media at multiple cafe stores classified within the same commercial district and induces connection from customer terminals through connection identification information included in each medium. Only feedback for which anonymous feedback input is completed at the customer terminal and reliability judgment is completed above a certain threshold is transmitted to the priority calculation unit, and the system stores the time of feedback occurrence, key keywords extracted from the feedback content, sentiment analysis results, and store identification information together for each problem type.

[0146] During the operation period, the system clusters feedback related to order waiting during peak hours into the same problem type and aggregates the proportion of occurrence and consecutive occurrence within the recent reference period based on the time of occurrence of each feedback included in that problem type.

[0147] In this case, if the system determines that the same problem type appears consecutively during weekday lunch hours within a specific week and that the proportion of appearances within the recent reference period exceeds the concentration criterion, it determines that the occurrence pattern of the problem type corresponds to a short-term concentrated type. Based on this determination, the system selects a frequency-centered judgment path for the problem type and sets the frequency value as the priority application criterion.

[0148] Next, the system manages the sentiment analysis results of feedback belonging to the same problem type as a set, and calculates the average value of negative sentiment intensity, the number of times intensity exceeding a threshold is maintained, and the duration. Additionally, the system calculates an operational impact value by querying the association table to determine whether the problem type maps to operational impact items related to order processing delays. Here, the sentiment value and the operational impact value are set as secondary criteria, and the system performs fine-tuning by combining the secondary criteria with the intermediate priority value calculated by the primary application criteria.

[0149] In this case, to prevent the auxiliary criteria from overturning the results of the primary application criteria, the system divides multiple problem types derived during the same period by the primary application criteria into upper and lower groups. The upper group is the set of problem types classified as higher by the primary application criteria, and the lower group is the set of other problem types.

[0150] The system restricts the combined result of auxiliary criteria from exceeding the boundaries between upper and lower groups. Specifically, if the correction amount calculated by the auxiliary criteria exceeds the correction limit range, the system truncates the correction amount to the correction limit range and applies it; furthermore, if the result of the applied correction involves shifting problem types from the lower group to the upper group or vice versa, the system invalidates the correction or readjusts it within the allowable range. Accordingly, while auxiliary criteria contribute to order refinement within the upper or lower group, they cannot alter the group boundaries themselves established by the primary applied criteria.

[0151] As a result of applying this constraint structure, the system sorts order types that appear repeatedly and intensively during peak hours by prioritizing them based on frequency; however, if there are problem types with higher sentiment values ​​within the same top group, those types are given higher priority within that group. Conversely, even if a problem type has a high frequency, it is placed at a relatively lower rank within the top group if it has low operational impact and sentiment values. This prevents items requiring immediate improvement but with weak linkages to actual operational effectiveness from being excessively fixed at the very top.

[0152] As another operational example, this is a case where a small number of feedback complaints regarding employee service appeared at a specific store during the same period, but the sentiment analysis results showed that the intensity of negative sentiment was repeated above a standard and the duration was calculated to be long.

[0153] The system selects the emotion-centered judgment path if it determines that the evaluation information on emotion intensity and repetition persistence meets the criteria, even if the evaluation information on the temporal distribution of occurrence frequency falls short of the concentration criteria.

[0154] In this case, the sentiment value is set as the primary application criterion, while the frequency and operational impact values ​​are set as secondary criteria. As a result, the system incorporates response complaint types that might be relegated to lower levels in simple frequency-based analysis into a higher group; however, the secondary criteria are applied only to refine the alignment within the same higher group, ensuring that the selection of a judgment path centered on sentiment impact is not neutralized by the secondary criteria.

[0155] As such, in this embodiment, the system independently evaluates the temporal distribution of occurrence frequency, emotional intensity and persistence, and operational impact for each problem type, selects a judgment path based on the judgment criteria that are satisfied first among them, sets the analysis value of the selected path as the priority application criterion, and then calculates the priority by combining auxiliary criteria only within a limited range.

[0156] Therefore, the system is distinguished from methods that uniformly apply the same combination of analysis values ​​to all problem types, and forms a differential priority calculation structure in which different judgment criteria are actually applied depending on the occurrence pattern of the problem type.

[0158] As illustrated in FIG. 3, in this embodiment, the priority calculation unit outputs the priority results calculated by problem type to the report generation stage and the improvement event design stage, and by reflecting the results confirmed in the performance analysis stage illustrated in FIG. 4 back into the decision path selection and criteria update, it forms an adaptive priority calculation structure in which the priority calculation criteria are not fixed but are gradually adjusted according to actual operation results.

[0159] In this embodiment, the AI-based store and commercial district improvement analysis system is configured not only to provide analysis results derived from individual stores in a simple parallel manner, but also to integrate and analyze anonymous feedback collected from multiple stores classified within the same commercial district to derive common problems at the commercial district level and comparative results by industry type.

[0160] To this end, whenever the system collects feedback by store, it stores store identification information, commercial area identification information, business type identification information, collection time information, and problem type identification information together with the feedback, and when analyzing by commercial area, it calls and processes only the feedbacks that have the same commercial area identification information as a set.

[0161] The common problem identification unit compares problem types or core keywords derived for each store within the same commercial area set to distinguish between items observed in a single store and items repeatedly observed in multiple stores. In this case, whether an item is repeatedly observed is not determined solely by the number of occurrences, but is judged based on whether at least one of the following is satisfied: the ratio of the number of stores where the problem type is observed to the total number of stores in the commercial area, the duration of repeated observation within the analysis period, and whether it is observed simultaneously in stores of the same industry.

[0162] The system assigns classification information as commercial area-common issues to problem types that meet the above criteria, and manages this classification information to include values ​​corresponding to whether the issue is a commercial area-common problem, a list of related stores, the observation period, and the intensity of observation. Accordingly, the system can clearly separate and manage local operational problems of specific stores from structural problems affecting the entire commercial area, and items classified as commercial area-common issues can be configured to be automatically registered as candidate items for subsequent commercial area-unit improvement events.

[0163] The comparative analysis department generates industry-specific benchmark data by aggregating satisfaction values, key keywords, and the distribution of problem types from multiple stores belonging to the same industry. Here, the benchmark data is not limited to simple average values ​​but can be configured to include a list of top problem types repeatedly observed within the industry over a certain period, the distribution of sentiment trends by problem type, and trends in priority fluctuations.

[0164] The system compares the analysis results of a specific store with the aforementioned benchmark data to identify the store's relative strengths and areas requiring improvement. In this process, the identification of relative strengths and areas requiring improvement is not determined solely by the difference in absolute values, but is calculated based on at least one of the following: the position within the same industry quantile, the direction of change within the same period, and whether there is overlap with common problems in the commercial area.

[0165] For example, if a specific store has a high priority for congestion-related problem types compared to benchmarks in the same industry, and is simultaneously classified as a common problem for the commercial district, the system can be configured to display the store's improvement needs as a short-term action item, while also displaying them as an item linked to a common improvement task at the commercial district level.

[0166] The report generation unit organizes the results of the common problem identification unit and the comparative analysis unit into a commercial area analysis report.

[0167] During the report generation process, the system separates and arranges common commercial area issues and individual store improvement items, specifying the scope of application for each item either at the commercial area or store level. Furthermore, the system sorts items by importance based on priority calculation results; specifically, it is configured to sort items according to commercial area-level priorities in the common commercial area issues section and according to store-specific priorities in the store-level section. Moreover, the system adjusts the level of summary in the report according to the recipient's role.

[0168] Reports provided to store operators may include actionable operational items for the store, candidates for immediately applicable measures, and conditions required for participation in improvement events, while reports provided to commercial district managers or local governments may include the distribution of common problems in the commercial district, infrastructure or operational environment factors requiring joint improvement, and comparative results of vulnerable items by industry.

[0169] In this case, the report generation unit does not terminate with the provision of output but is configured to retain identification information linked to the items included in the report, ensuring that the same items can be accurately recalled during subsequent improvement event design or performance analysis phases. In other words, the report is designed to function as an input structure for subsequent stages rather than as a result display.

[0170] Below, the process by which the above-mentioned commercial area analysis report is generated during the actual operation and linked to subsequent steps is explained as an example.

[0171] Anonymous feedback is collected over a certain period from multiple cafes classified as belonging to the same commercial district, and types of order waiting problems during peak hours are derived from the results of pattern analysis.

[0172] The system classifies a problem as a common commercial area issue if the number of stores within a commercial area where the problem type is observed exceeds a reference ratio, it is observed repeatedly during the same period, and a tendency to appear concentrated particularly on weekends and during lunchtime is confirmed.

[0173] Once classification is complete, the system assigns classification information, including a list of related stores and an observation period, to items marked as common problems in the commercial area, and places the corresponding items in the common problem area of ​​the commercial area analysis report.

[0174] Simultaneously, the system compares each store's satisfaction values ​​regarding waiting experiences and the distribution of problem types with industry-specific benchmark data. For instance, if congestion-related problem types appear at a moderate level in the industry benchmark, but a specific store consistently maintains a high priority for these types and exhibits a strong negative sentiment in feedback, the system identifies that store as vulnerable and places elements requiring improvement within the store-level action task area.

[0175] In this case, since the same problem type is also classified as a common problem for the commercial district, the system can be configured to display the linkage between store-unit implementation tasks and commercial district-unit joint improvement tasks together.

[0176] In addition, the system can be configured to automatically generate commercial area analysis reports on a monthly or quarterly basis, enabling time-series comparisons.

[0177] In this case, the system retrieves identification information for the same problem type stored in previous period reports to reflect changes in priority, increases or decreases in the number of observed stores, changes in feedback sentiment intensity, and the cumulative status of performance analysis results. Therefore, rather than relying on listed results from a single point in time, the recipient can verify through inter-period comparison whether the same problem type is actually moving in the direction of deterioration or improvement.

[0178] Furthermore, the system can grade and display the items of the commercial area analysis report.

[0179] Grading is not a simple arbitrary designation, but is performed by reflecting at least one of the following: the presence of common problems in the commercial area, the results of priority calculation, and the presence of improvement effects confirmed during the performance analysis stage.

[0180] For example, items with high priority but no confirmed improvement in performance analysis results may be maintained at a high grade or strengthened, while items with medium priority but consistently confirmed improvement in performance analysis results may be converted to a management grade. This grading system can be configured to be reused as input information for automatically setting the purpose and duration of subsequent improvement events during the design phase.

[0181] As another embodiment, an item classified as a common commercial area problem is selected as a target for a commercial area-unit improvement event, and after the improvement event is performed, if the performance analysis unit determines whether the decrease in the frequency of occurrence of the problem type and the alleviation of emotional intensity persist for a certain period or longer compared to the reference period, the system may be configured to transmit the result of the determination to the priority calculation unit to adjust the criteria for selecting the judgment path or the correction limit range.

[0182] Accordingly, judgment paths where actual improvement effects have been confirmed in a specific commercial area can be adjusted to be selected more frequently as initial priority criteria in the same or similar commercial areas in the future; conversely, if no improvement effects are confirmed, other judgment paths can be adjusted to be prioritized. Thus, the commercial area analysis report functions not as an endpoint of results, but as a key medium in an analysis feedback structure that leads to improvement events and standard updates.

[0184] FIG. 4 is a diagram illustrating the flow in which the result of an improvement event execution in a system according to the present embodiment is analyzed by the performance analysis unit, and the analysis result is reflected and updated in the judgment path selection and analysis value combination criteria of the priority calculation unit. Hereinafter, with reference to FIG. 4, the specific operation sequence and constraints regarding how input values ​​before and after the improvement event execution are defined, how they are compared and judged by what procedure, and how the result is reflected in the priority calculation criteria will be explained.

[0185] In this embodiment, the AI-based store and commercial area improvement analysis system includes a performance analysis step for evaluating the impact of an improvement event on the problem type or improvement item under paragraph 3 after the improvement event is performed. At this time, the performance analysis step does not stop at simple post-hoc statistical verification but functions as a step for generating a basis for judgment so that the judgment path and the limit of the combination of auxiliary criteria selected in the subsequent priority calculation step can be adjusted according to actual operational results.

[0186] In this embodiment, the performance analysis unit refers to analysis values ​​generated based on the same problem type and the same analysis range before and after the execution of the improvement event as input information. Here, the analysis range can be set at the store level or the commercial district level, and the performance analysis is configured to be performed in correspondence with the range to which the event is applied.

[0187] The input analysis values ​​referenced by the performance analysis department may consist of frequency values, sentiment values, and operational performance values ​​calculated for the same problem type.

[0188] The frequency value is a value corresponding to the frequency of appearance of anonymous feedback containing the problem type during the same observation period before and after the event execution, the sentiment value is a value reflecting the negative intensity and persistence based on the sentiment analysis results of anonymous feedback related to the same problem type, and the operational performance value consists of at least one of the revisit probability value or the operational indicator value, and is defined in correspondence with the operational item related to the purpose of the event.

[0189] The Performance Analysis Department manages a set of reference values ​​stored prior to event execution and a set of comparison values ​​accumulated over a certain period after event execution by matching them based on the same problem type and analysis scope.

[0190] The set of reference values ​​is defined as values ​​calculated during the reference observation period immediately prior to the application of the event, and the set of comparison values ​​is defined as values ​​calculated during the comparison observation period after the application of the event.

[0191] The reference observation period and the comparison observation period can be set to have the same length and can be configured to mitigate the influence of external factors by aligning at least the day composition and time zone composition to be identical.

[0192] The performance analysis department generates derived information corresponding to the amount of change, the rate of change, and the persistence of change for each analysis value in order to calculate the difference between the above set of reference values ​​and the set of comparison values.

[0193] The amount of change is defined as the difference between the comparison value and the reference value, the rate of change is defined as the ratio of the amount of change to the reference value, and the persistence of change is defined as the period or number of times that a change in the direction of improvement is continuously maintained within the comparison observation period.

[0194] For example, in the case of frequency values, a time series of frequency values ​​calculated for each unit period during the comparison observation period can be constructed, and the length of the interval in which the direction of decrease is continuously maintained in that time series can be calculated as the persistence of change.

[0195] In the case of sentiment values ​​as well, a time series can be constructed by calculating the average negative intensity or the ratio of exceeding a standard of feedbacks belonging to the same problem type for each unit period, and the length of the interval in which the mitigation direction is maintained within that time series can be calculated as the persistence of change.

[0196] In the case of operational performance values, since the revisit probability value may be influenced by external factors, it can be configured to define the amount of change and the persistence of change by calculating the change in relative position with respect to the industry average or commercial area average of the same period, rather than judging it based solely on a single absolute value.

[0197] The performance analysis department performs an assessment of the improvement effect using the above-mentioned derived information.

[0198] The determination of the improvement effect is not merely based on confirming whether a change in the direction of improvement exists in at least one analysis value, but is performed based on whether the direction of improvement, the scope of improvement, and the sustainability of improvement simultaneously satisfy predefined judgment conditions.

[0199] For example, the improvement effect in frequency values ​​may be configured to be recognized only when the amount of change in frequency values ​​is in the direction of decrease, the rate of change is above the standard, and the persistence of change in the direction of decrease is above the standard.

[0200] The improvement effect in the appraisal value may be configured to be recognized only when the appraisal value changes in a mitigating direction, the rate of exceeding the standard decreases, and the persistence of the change in the mitigating direction is above the standard.

[0201] The improvement effect in operational performance values ​​may be configured to be recognized only when the relative position is maintained or improved without the revisit probability value falling below a standard.

[0202] In this case, the performance analysis unit may be configured to distinguish and determine primary and secondary indicators corresponding to the purpose of the event in preparation for cases where conflicting changes occur between the analysis values.

[0203] Primary metrics are defined as values ​​that directly correspond to the problem type or operational item set as the target for resolution during the event design phase, while secondary metrics are defined as values ​​that complementarily verify the impact of changes in primary metrics on actual customer experience or operational performance.

[0204] The results of the improvement effect judgment by the performance analysis unit are transmitted to the priority calculation unit and used to update the judgment path selection criteria and auxiliary criteria combination limits described in Fig. 3.

[0205] In this embodiment, the criterion update is performed not by uniformly raising or lowering the weights of specific analysis values, but by adjusting the selection rule for which judgment path is prioritized for the same problem type and the allowable range in which the auxiliary criterion can correct intermediate priority values.

[0206] Specifically, the performance analysis department identifies the types of analysis values ​​that served as primary indicators in determining improvement effects, and based on the identification results, generates update information to strengthen the probability or priority selection conditions for the judgment path corresponding to the analysis value in subsequent priority calculations for the same problem type.

[0207] For example, if a change in sentiment value serves as the key basis for judging the improvement effect, the priority selection condition of the sentiment-influence-centered judgment path may be relaxed to allow for more frequent selection; similarly, if a change in frequency value serves as the key basis, the priority selection condition of the occurrence frequency-centered judgment path may be relaxed to allow for more frequent selection. Conversely, if no improvement effect is confirmed, the existing priority selection condition may be maintained, or the priority selection condition of other judgment paths may be strengthened.

[0208] In addition, the standard update is not determined solely by the result of a single event, but is configured to cumulatively reflect the performance analysis results of multiple events performed on the same problem type.

[0209] To this end, the system stores problem type identification information, event type identification information, coverage scope information, a set of reference values, a set of comparison values, improvement effect judgment results, and update history together. Furthermore, the performance analysis unit determines whether improvement effects are repeatedly confirmed from the judgment results of multiple accumulated events for the same problem type, and can be configured to gradually expand the adjustment range of reference updates only if repeated confirmation is found. Accordingly, this prevents one-time fluctuations or accidental improvements caused by external factors from excessively inducing reference updates.

[0210] In this embodiment, constraints are applied to the update of the standard.

[0211] Constraints function as limitations to prevent specific analysis values ​​or specific decision paths from excessively dominating the priority calculation results.

[0212] For example, the adjustment range of the primary selection condition may be limited so as not to exceed a predefined upper limit, and the expansion or contraction of the secondary criterion combination limit may also be restricted to be performed in stages only within a predefined range. Additionally, if no further improvement effect is confirmed for a certain period, or if conflicting event results occur repeatedly in the same problem type, the system may be configured to maintain the previous update or revert to the previous stage.

[0213] Due to these constraints, the system is configured to prevent learning or updates from fluctuating drastically in the short term and to operate in a direction where reference updates are gradually refined.

[0214] The following describes the process by which the above performance analysis and standard update actually operate as an example.

[0215] Problem types related to waiting experiences during peak hours at specific stores are identified as having high priority, and improvement events aimed at improving waiting time guidance methods are carried out.

[0216] The system sets a reference observation period prior to event execution, calculates frequency and sentiment values ​​corresponding to the same problem type within the reference observation period, and also calculates operational performance values ​​corresponding to the likelihood of revisiting, storing them as a set of reference values.

[0217] At this time, the reference observation period is set to have the same day composition and time zone composition, and is sorted so that comparison can be made later.

[0218] After an event is performed, the system sets a comparison observation period and calculates frequency values, sentiment values, and operational performance values ​​for the same problem type in the same manner, storing them as a set of comparison values.

[0219] The Performance Analysis Department calculates the amount and rate of change between the set of reference values ​​and the set of comparison values, and calculates the continuity of change by verifying whether the direction of improvement is continuously maintained within the comparison observation period.

[0220] As a result, if the frequency value changes in a decreasing direction and this change is maintained for a significant portion of the comparative observation period, the sentiment value changes in a mitigating direction while the rate of exceeding the negative intensity standard decreases, and the operational performance value does not fall below the standard while maintaining the relative position compared to the industry average, the Performance Analysis Department determines that the event has a significant improvement effect.

[0221] At this time, the Performance Analysis Department classifies cases where the consistency of the change in emotional value is strongest and it is determined that the change is directly linked to the improvement of customer experience as emotional value-centered improvements, and transmits the classification result to the Priority Calculation Department.

[0222] The priority calculation unit updates the criteria to reflect the classification results, progressively strengthen the priority selection conditions of the emotion-influence-centered judgment path in subsequent priority calculations for the same problem type, and fine-tune the auxiliary criterion combination limits only within a predefined range.

[0223] Accordingly, if the same problem type is observed again thereafter, even if the frequency value is low, if the intensity and persistence of the emotional value meet the criteria, the emotion-centered judgment path is more likely to be selected first and placed at a higher priority.

[0224] As another embodiment, the performance analysis unit may be configured to integrate and analyze the results of improvement events commonly performed at multiple stores within the same commercial area.

[0225] The system calculates a set of standard values ​​and a set of comparison values ​​at the store level, respectively, and then generates a set of representative values ​​aggregated at the commercial district level.

[0226] The set of representative values ​​can be defined as median or quantile-based values ​​for the values ​​of stores within a commercial area, so that extreme values ​​of a specific store do not distort the judgment of the commercial area.

[0227] The performance analysis department can be configured to evaluate the amount of change and the persistence of change in a set of representative values ​​at the commercial district level, and to progressively update the criteria for selecting judgment paths applied to common commercial district issues only when improvement effects are repeatedly confirmed at the commercial district level. This ensures that priority criteria from the perspective of commercial district managers or local governments are not swayed by accidental fluctuations of a single store, but are updated according to the overall trends of the commercial district.

[0228] As another embodiment, the system may be configured to accumulate the results of performance analysis of specific improvement events as historical data and use them as initial settings for designing events and determining priorities for similar problem types in the future.

[0229] The system stores the problem type, event type, coverage, configuration of primary and secondary indicators, summary statistics of the reference value set and comparison value set, improvement effect assessment results, and reference update results together.

[0230] Subsequently, when the same or similar problem type is derived, the priority calculation unit sets the judgment path in which the improvement effect has been repeatedly confirmed in the past history as the initial priority path, but is configured to perform a limited combination so that the auxiliary criteria cannot reverse the primary judgment result according to the constraints of FIG. 3.

[0231] Accordingly, as operational experience accumulates, the system becomes capable of more stable and predictable prioritization and event design for the same type of problem.

[0232] Furthermore, the logic for generating judgment criteria and updating criteria within the performance analysis unit can be implemented as a separate, independent configuration. It can also be configured separately as an event performance-based criteria update unit that updates priority calculation criteria based on data before and after the execution of improvement events. Since this configuration includes a procedural structure and constraints that feed back the performance of improvement events into priority calculation, it functions as a unit that performs the technical operation of updating the criteria selection structure based on operational results, rather than simple statistical analysis.

[0234] Although preferred embodiments of the present invention have been illustrated and described above, the present invention is not limited to the specific embodiments described above. It is understood that various modifications can be made by those skilled in the art without departing from the essence of the invention as claimed in the claims, and such modifications should not be understood individually from the technical spirit or perspective of the present invention. Explanation of the symbols

[0235] 10: AI-based Store and Commercial District Improvement Analysis System

Claims

Claim 1 An AI-based store and commercial area improvement analysis system that analyzes customer opinions collected based on anonymous feedback to derive improvement items for a store or commercial area and provides the results to a store operator or commercial area manager, comprising: a connection induction unit that induces a user terminal to connect through identification information including a QR code or a deep link identifier included in media placed in an offline store; an anonymous feedback input unit configured to input positive items and improvement items regarding a visited store separately from the user terminal; a data storage unit that collects and stores the anonymous feedback; a text analysis unit that generates text preprocessing results, sentiment analysis results, and core keyword extraction results for the collected anonymous feedback; a pattern analysis unit that derives recurring problem types based on the results of mapping the core keywords according to an industry classification system and clustering results among similar feedback; a priority calculation unit that calculates the priority of improvement items using frequency values ​​corresponding to the frequency of appearance of feedback, sentiment values ​​corresponding to sentiment analysis results, and operation / revisit impact values ​​corresponding to store operation items or the possibility of customer revisit as input variables; a report generation unit that generates and provides an improvement report based on the priority calculated by the priority calculation unit; and a user through the identification information A reward control unit that calculates a feedback quality value based on evaluation information corresponding to the connection status of a terminal, the completion status of anonymous feedback input, the number of input items, keyword diversity, or the clarity of sentiment analysis results, and generates improvement events to induce customer participation or store operation improvement; a common problem identification unit that classifies keywords or problem types commonly mentioned by multiple stores among feedback collected from multiple stores within the same commercial area as commercial area common problems; and aggregates satisfaction values ​​and major keywords of multiple stores belonging to the same industry to generate industry-specific benchmark data,A comparative analysis unit that identifies relative strengths and areas needing improvement by comparing the value of a specific store with the above benchmark data; and a performance analysis unit that analyzes changes in additional anonymous feedback, store operation values, or customer revisit values ​​collected before and after the execution of the improvement event; wherein the anonymous feedback input unit is configured to prevent biased, unidirectional feedback input by controlling feedback submission to be restricted when positive items and improvement items are not entered, respectively; wherein the text analysis unit generates multiple evaluation information including information indicating differences in sentiment distribution, information indicating diversity of input expressions, or information corresponding to the volume of described content by comparing sentiment analysis results derived for each of the positive items and improvement items; wherein the priority calculation unit is configured to adopt only feedback with a feedback reliability value above a standard derived by comparing the multiple evaluation information as input for priority calculation; wherein the reward control unit generates the improvement event based on the feedback quality value or the priority calculation result, and wherein the improvement event is configured such that at least one of the event's purpose, target customer group, participation conditions, or period is set in correspondence with the problem type or improvement item; wherein the priority calculation unit selectively applies different ranges of application ratios to each of the frequency value, the sentiment value, and the operation / revisit influence value, and the application ratio is relative It is configured such that the priority of improvement items is calculated based on the combination result of the frequency value, the appraisal value, and the operation / revisit impact value by being selected in a normalized state according to importance, and the results of the common problem identification unit and the comparative analysis unit are configured to be included in the commercial area analysis report, and the priority calculation unit does not immediately combine the frequency value, the appraisal value, and the operation / revisit impact value, but after generating the temporal distribution evaluation information of the frequency of occurrence, the intensity distribution and repetition persistence evaluation information of the appraisal analysis results, and the operation impact evaluation information for each problem type,Evaluation information that primarily satisfies pre-set judgment criteria is selected, and one of a judgment path centered on appearance frequency, a judgment path centered on emotional influence, or a judgment path centered on operational influence is selected in response to the selected evaluation information; the priority calculation unit sets one of the frequency value, the emotional value, or the operational / revisit influence value corresponding to the selected judgment path as a priority application criterion, and sets the remaining value as a secondary criterion; the priority calculation unit divides multiple problem types into upper and lower groups based on the intermediate priority value calculated by the priority application criterion, and is configured so that the correction based on the secondary criterion is applied only within a correction limit range that does not invert the boundary between the upper and lower groups; the performance analysis unit determines whether the value change before and after the event execution is within a pre-set threshold range or is judged to be a significant change compared to the reference period; and the priority calculation unit updates at least one of the selection criteria for the application ratio applied to at least one of the frequency value, the emotional value, or the operational / revisit influence value, the selection criteria for the judgment path, or the correction limit range based on the judgment result of the performance analysis unit, thereby allowing the priority calculation criteria for the improvement item to dynamically adapt according to the result of the improvement event. AI-based store and commercial district improvement analysis system characterized by being configured. Claim 2 delete Claim 3 delete Claim 4 delete Claim 5 delete Claim 6 delete

Images