Robotic instruction using statistical analysis method, apparatus, device, and medium

Abstract

This invention provides a method, apparatus, device, and medium for statistical analysis of robot instruction usage, comprising: using organizations, business lines, and outlets as graph nodes; obtaining the relationships between nodes based on actual business rules and historical behavior data as graph edges; dynamically assigning edge weights based on factors including the frequency of organizational calls to business lines, the frequency of task collaboration between outlets, and the strength of instruction flow dependency; constructing a multi-dimensional business relationship graph; the multi-dimensional business relationship graph is used to depict the dynamically changing business dependencies among organizations, business lines, and outlets; and obtaining predicted values ​​of robot instruction usage, corresponding confidence intervals, and quantified business impact factors within a preset future time period through a prediction model based on the multi-dimensional business relationship graph.

Description

Technical Field

[0001] This invention relates to the field of logistics management technology, and more specifically, to statistical analysis methods, apparatus, equipment, and media for robot instruction usage. Background Technology

[0002] With the widespread application of intelligent robots in logistics operations, intelligent robot platforms generate a large number of robot instructions initiated and executed collaboratively by organizations, business lines, and outlets. These instructions involve collaborative interactions among multiple entities, levels, and business lines, including branch offices, regional centers, outlet managers, courier pickup and delivery, intelligent dispatching, city outlets, and community stations. The invocation, flow, and execution of these instructions are characterized by complex relationships, uneven frequency, and reliance on dynamic changes.

[0003] Traditional methods of statistical analysis of robot commands often rely on simple data aggregation and static statistics, making it difficult to effectively characterize the dynamic business dependencies between organizations, business lines, and outlets. Furthermore, they fail to fully integrate the interrelationships of business structures with time-series variation patterns. When predicting robot command usage, predictions are typically based solely on single time-series data, neglecting crucial information such as inter-entity collaboration, business affiliation, and task flow. This results in insufficient accuracy of the predictions and an inability to quantitatively assess their reliability.

[0004] Furthermore, existing platforms suffer from issues such as insufficiently granular data access control, inadequate protection of sensitive information, and difficulty in tracing access behavior during instruction data analysis. Therefore, how to achieve multi-dimensional relationship modeling, accurate prediction, and reliable evaluation of robot instructions, and provide a secure and controllable data access mechanism, has become a pressing technical problem to be solved in this field. Summary of the Invention

[0005] The main objective of this invention is to address the problem that existing technologies rely solely on single time-series data for prediction, neglecting crucial information such as inter-entity collaboration, business affiliation, and task flow, resulting in insufficient accuracy of prediction results and an inability to quantitatively assess the reliability of the prediction results.

[0006] The first aspect of this invention provides a statistical analysis method for robot instruction usage, comprising: Using organizations, business lines, and outlets as graph nodes, and obtaining the relationships between nodes based on actual business rules and historical behavior data as graph edges, a multi-dimensional business relationship graph is constructed by dynamically assigning edge weights based on factors such as the frequency of calls from organizations to business lines, the frequency of task collaboration between outlets, and the strength of instruction flow dependencies. This multi-dimensional business relationship graph is used to depict the dynamically changing business dependencies among organizations, business lines, and outlets. The organization mentioned above is the main body responsible for the instruction management and initiation of the intelligent robot platform. It is an enterprise operating entity with hierarchical affiliation and business jurisdiction attributes, including branch offices and regional centers. The business line is the instruction execution carrier that carries standardized business functions on the intelligent robot platform. It is a functional module oriented towards logistics operation scenarios, including network manager, courier pickup and delivery, and intelligent scheduling. The network points are the offline execution entities of the instructions from the intelligent robot platform. They are logistics operation units with geographical location and business coverage attributes, including urban network points, community stations, and last-mile service points. Based on a multidimensional business relationship graph and historical instruction usage time series, a prediction model is used to obtain the predicted value of robot instruction usage within a future preset time period. The prediction model includes a graph coding layer for extracting business structure features and a temporal coding layer for extracting time features. The robot instruction usage is predicted based on the fusion features of the graph coding layer and the temporal coding layer.

[0007] Optionally, in a first implementation of the first aspect of the present invention, the organization, business lines, and outlets are used as graph nodes, and the associations between nodes are obtained based on actual business rules and historical behavior data as graph edges. Edge weights are dynamically assigned based on factors including the frequency of calls from the organization to the business lines, the frequency of task collaboration between outlets, and the strength of instruction flow dependencies, to construct a multi-dimensional business relationship graph. This multi-dimensional business relationship graph is used to depict the dynamically changing business dependencies among the organization, business lines, and outlets, including: Collect robot instruction usage logs, organizational structure information, business line association rules, and branch network topology relationships; The robot instruction usage log is a standardized data log in the intelligent robot platform that records the entire lifecycle behavior of the organization, business lines, and outlets in calling, executing, transferring, and coordinating robot instructions. It includes quantitative information on the subject of the behavior, time, path, and result. The organizational structure information is structured data that includes the static attributes and hierarchical associations of various organizational entities, such as internal subsidiaries and regional centers, as well as their hierarchical affiliations, jurisdictions, and business responsibilities. The business line association rules are the inherent business association specifications within the intelligent robot platform, including the functional positioning, triggering conditions, and upstream and downstream linkage logic of various business lines such as network managers and courier pickup and delivery. The network topology is topological data that includes the geographical location, hierarchical affiliation, business coverage, business radiation and connection relationships between urban and community sites. The collected robot instruction usage logs, organizational structure information, business line association rules, and network topology relationships are preprocessed to obtain preprocessed robot instruction usage logs, organizational structure information, business line association rules, and network topology relationships. Based on preprocessed robot instruction usage logs, organizational structure information, business line association rules, and network topology, a multi-dimensional business relationship graph is constructed, with organizations, business lines, and networks as graph nodes. The relationships between nodes are obtained based on actual business rules and historical behavior data as graph edges. Edge weights are dynamically assigned based on factors such as the frequency of organization calls to business lines, the frequency of task collaboration between networks, and the strength of instruction flow dependencies.

[0008] Optionally, in a second implementation of the first aspect of the present invention, the preprocessing of the collected robot instruction usage logs, organizational structure information, business line association rules, and network topology relationships to obtain preprocessed robot instruction usage logs, organizational structure information, business line association rules, and network topology relationships includes: The collected robot instruction usage logs, organizational structure information, business line association rules, and network topology relationships are sequentially cleaned, normalized, and structured mapped to obtain preprocessed robot instruction usage logs, organizational structure information, business line association rules, and network topology relationships. The data cleaning process is used to remove invalid logs from robot instruction usage logs, organizational structure information, erroneous structure information in network topology relationships, and invalid rules in business line association rules, and to supplement missing business attributes that meet preset requirements. The normalization process is used to standardize business data of different formats and units. The structured mapping process is used to transform unstructured and / or semi-structured data into graph-recognizable structured key-value pair data, associating each data type with a unique identifier.

[0009] Optionally, in a third implementation of the first aspect of the present invention, the prediction model includes: extracting business structure association features from a multidimensional business relationship graph using a graph attention network; extracting time series change features from a historical instruction usage time series using a Transformer encoder; fusing the extracted business structure association features and time series change features; and predicting the amount of robot instruction usage within a preset time period based on the fused features.

[0010] Optionally, in a fourth implementation of the first aspect of the present invention, obtaining the predicted value of robot instruction usage within a future preset time period includes: By using the adaptive attention mechanism of graph attention network, differentiated attention weights are assigned to three types of nodes—organization, business line, and outlet—as well as the edges between nodes, to capture explicit multidimensional dependencies and implicit associations between nodes. Based on the captured explicit multidimensional dependencies and implicit associations between nodes, structural feature vectors are obtained. The Transformer encoder captures the periodic patterns, trends, and sudden fluctuations in instruction usage using time series analysis of historical instructions. Based on the captured periodic patterns, trends, and sudden fluctuations in instruction usage, a time series feature vector is obtained. By dimensionally aligning and cross-dimensionally fusing structural feature vectors and temporal feature vectors, a comprehensive feature vector that integrates business correlation logic and temporal patterns is generated. The comprehensive feature vector is based on the fully connected regression layer of the prediction model. It combines the mapping relationship between the historical comprehensive feature vector and the historical command usage learned during the training phase of the prediction model to predict the robot command usage of each node in the future preset time period.

[0011] Optionally, in a fifth implementation of the first aspect of the present invention, the prediction model further includes: The fully connected regression layer introduces a Gaussian probability regression mechanism to synchronously output the initial variance corresponding to the predicted value of robot command usage at each node. The initial variance characterizes the original fluctuation of the predicted value. Based on the historical prediction error distribution pattern learned during the training phase of the prediction model, the initial variance is dynamically calibrated to obtain the calibrated variance; the historical prediction error distribution pattern is the statistical distribution characteristic of the error value between the historical prediction mean and the historical actual command usage. Based on the preset confidence level, and using the predicted value and the calibrated variance, the confidence interval corresponding to the predicted value of robot command usage at each node is calculated by the normal distribution statistical method. The width of the confidence interval is used to quantify the confidence level of the corresponding predicted value; wherein, the smaller the interval width, the higher the confidence level of the predicted value, and the larger the interval width, the lower the confidence level of the predicted value.

[0012] Optionally, in a sixth implementation of the first aspect of the present invention, the method further includes: obtaining the organization, management scope, and business responsibilities of the current user based on the identity information of the current user; matching the node position and association boundary of the current user in the multidimensional business relationship graph based on the organization, management scope, and business responsibilities of the current user; and dynamically controlling the data visibility range based on the matched node position and association boundary of the current user in the multidimensional business relationship graph. At the same time, sensitive information that meets the preset requirements is automatically de-identified; when users access data, the data access path, operation behavior and decision basis are recorded.

[0013] A second aspect of the present invention provides a robot instruction usage statistical analysis device, characterized in that it comprises: The multi-dimensional business relationship graph construction module uses organizations, business lines, and outlets as graph nodes, and obtains the relationships between nodes as graph edges based on actual business rules and historical behavior data. The edge weights are dynamically assigned by factors including the frequency of organization calls to business lines, the frequency of task collaboration between outlets, and the strength of instruction flow dependencies. The multi-dimensional business relationship graph is used to depict the dynamically changing business dependencies among organizations, business lines, and outlets. The organization mentioned above is the main body responsible for the instruction management and initiation of the intelligent robot platform. It is an enterprise operating entity with hierarchical affiliation and business jurisdiction attributes, including branch offices and regional centers. The business line is the instruction execution carrier that carries standardized business functions on the intelligent robot platform. It is a functional module oriented towards logistics operation scenarios, including network manager, courier pickup and delivery, and intelligent scheduling. The network points are the offline execution entities of the instructions from the intelligent robot platform. They are logistics operation units with geographical location and business coverage attributes, including urban network points, community stations, and last-mile service points. The instruction usage prediction module is used to obtain the predicted value of robot instruction usage within a preset time period based on a multi-dimensional business relationship graph and historical instruction usage time series through a prediction model.

[0014] Optionally, in a first implementation of the second aspect of the present invention, the multi-dimensional business relationship graph construction module includes: The data collection submodule is used to collect robot instruction usage logs, organizational structure information, business line association rules, and network topology relationships. The robot instruction usage log is a standardized data log in the intelligent robot platform that records the entire lifecycle behavior of the organization, business lines, and outlets in calling, executing, transferring, and coordinating robot instructions. It includes quantitative information on the subject of the behavior, time, path, and result. The organizational structure information is structured data that includes the static attributes and hierarchical associations of various organizational entities, such as internal subsidiaries and regional centers, as well as their hierarchical affiliations, jurisdictions, and business responsibilities. The business line association rules are the inherent business association specifications within the intelligent robot platform, including the functional positioning, triggering conditions, and upstream and downstream linkage logic of various business lines such as network managers and courier pickup and delivery. The network topology is topological data that includes the geographical location, hierarchical affiliation, business coverage, business radiation and connection relationships between urban and community sites. The preprocessing submodule is used to preprocess the collected robot instruction usage logs, organizational structure information, business line association rules, and network topology relationships to obtain the preprocessed robot instruction usage logs, organizational structure information, business line association rules, and network topology relationships. The relationship graph construction submodule is used to construct a multi-dimensional business relationship graph based on preprocessed robot instruction usage logs, organizational structure information, business line association rules, and network topology relationships. Organizations, business lines, and networks are used as graph nodes, and the associations between nodes are obtained based on actual business rules and historical behavior data as graph edges. Edge weights are dynamically assigned by factors including the frequency of organization calls to business lines, the frequency of task collaboration between networks, and the strength of instruction flow dependencies, thus constructing a multi-dimensional business relationship graph.

[0015] Optionally, in a second implementation of the second aspect of the present invention, the preprocessing submodule includes: The collected robot instruction usage logs, organizational structure information, business line association rules, and network topology relationships are sequentially cleaned, normalized, and structured mapped to obtain preprocessed robot instruction usage logs, organizational structure information, business line association rules, and network topology relationships. The data cleaning process is used to remove invalid logs from robot instruction usage logs, organizational structure information, erroneous structure information in network topology relationships, and invalid rules in business line association rules, and to supplement missing business attributes that meet preset requirements. The normalization process is used to standardize business data of different formats and units. The structured mapping process is used to transform unstructured and / or semi-structured data into graph-recognizable structured key-value pair data, associating each data type with a unique identifier.

[0016] Optionally, in a third implementation of the second aspect of the present invention, the prediction model includes: extracting business structure association features from a multidimensional business relationship graph using a graph attention network; extracting time series change features from a historical instruction usage time series using a Transformer encoder; fusing the extracted business structure association features and time series change features; and predicting the amount of robot instruction usage within a preset time period based on the fused features.

[0017] Optionally, in a fourth implementation of the second aspect of the present invention, the instruction usage prediction module includes: The structural feature vector acquisition submodule is used to assign differentiated attention weights to three types of nodes (organization, business line, and network point) and the edges between nodes through the adaptive attention mechanism of graph attention network, capture the explicit multidimensional dependencies and implicit association features between nodes, and obtain the structural feature vector based on the captured explicit multidimensional dependencies and implicit association features between nodes. The time-series feature vector acquisition submodule is used to capture the periodic patterns, trend changes, and sudden fluctuations in instruction usage using the Transformer encoder and time series data of historical instructions. Based on the captured periodic patterns, trend changes, and sudden fluctuations in instruction usage, the time-series feature vector is obtained. The feature fusion submodule is used to perform dimensional alignment and cross-dimensional fusion of structural feature vectors and temporal feature vectors to generate a comprehensive feature vector that integrates business correlation logic and time patterns. The prediction submodule is used to predict the robot command usage at each node in a future preset time period by using the fully connected regression layer of the prediction model based on the comprehensive feature vector and combining the mapping relationship between the historical comprehensive feature vector and the historical command usage learned during the prediction model training phase.

[0018] Optionally, in a fifth implementation of the second aspect of the present invention, the prediction model further includes: a credibility assessment submodule, used to assess the credibility of the predicted values ​​of robot instruction usage at each node based on a confidence interval; The fully connected regression layer introduces a Gaussian probability regression mechanism to synchronously output the initial variance corresponding to the predicted value of robot command usage at each node. The initial variance characterizes the original fluctuation of the predicted value. Based on the historical prediction error distribution pattern learned during the training phase of the prediction model, the initial variance is dynamically calibrated to obtain the calibrated variance; the historical prediction error distribution pattern is the statistical distribution characteristic of the error value between the historical prediction mean and the historical actual command usage. Based on the preset confidence level, and using the predicted value and the calibrated variance, the confidence interval corresponding to the predicted value of robot command usage at each node is calculated by the normal distribution statistical method. The width of the confidence interval is used to quantify the confidence level of the corresponding predicted value; wherein, the smaller the interval width, the higher the confidence level of the predicted value, and the larger the interval width, the lower the confidence level of the predicted value.

[0019] Optionally, in a sixth implementation of the second aspect of the present invention, the apparatus further includes: a business-aware permission control module, used to obtain the organization, management scope, and business responsibilities of the current user based on the identity information of the current user; match the node position and association boundary of the current user in the multidimensional business relationship graph based on the organization, management scope, and business responsibilities of the current user; and dynamically control the data visibility range based on the matched node position and association boundary of the current user in the multidimensional business relationship graph. At the same time, sensitive information that meets the preset requirements is automatically de-identified; when users access data, the data access path, operation behavior and decision basis are recorded.

[0020] A third aspect of the present invention provides an electronic device, the electronic device comprising a memory and at least one processor, wherein the memory stores instructions; The at least one processor invokes the instructions in the memory to cause the electronic device to execute the steps of the robot instructions using statistical analysis methods as described above.

[0021] A fourth aspect of the present invention provides a computer-readable storage medium storing instructions, characterized in that, when executed by a processor, the instructions implement the steps of the statistical analysis method for robot instructions as described above.

[0022] Compared with the prior art, the present invention has the following beneficial effects: 1. This invention uses organizations, business lines, and outlets as core nodes, and combines actual business rules and historical behavioral data to construct a multi-dimensional business relationship graph. By assigning dynamic weights to edges based on call frequency, collaboration frequency, and flow intensity, it not only achieves a full-coverage depiction of the entire business chain from instruction initiation to function execution to offline implementation, but also captures the relationship changes brought about by business adjustments in real time through dynamic updates of weights. This effectively solves the problem that traditional static statistics cannot depict the dynamic business dependencies of multiple entities and multiple levels, making business relationships more concrete and more in line with actual operational scenarios. 2. The prediction model of this invention innovatively integrates graph encoding layer and time-series encoding layer. It extracts business structure association features through graph attention network to capture explicit dependencies and implicit associations between nodes; it extracts time-series features through Transformer encoder to capture the periodic patterns, trend changes and sudden fluctuations in instruction usage, and then generates a comprehensive feature vector for prediction through dimension alignment and cross-dimensional fusion. It breaks through the limitations of traditional prediction based on only a single time-series data, fully integrates business structure association features and time-series change patterns, making the prediction results more consistent with the business logic of logistics operations, and greatly improving the accuracy of robot instruction usage prediction. 3. This invention introduces a Gaussian probability regression mechanism into the fully connected regression layer of the prediction model, synchronously outputs the initial variance corresponding to the predicted value, and dynamically calibrates the initial variance by combining the historical prediction error distribution pattern. Finally, the confidence interval is calculated based on the preset confidence level, and the credibility of the predicted value is quantitatively characterized by the interval width. This realizes a quantitative credibility assessment of the prediction results, solves the problem that the existing technology cannot effectively assess the credibility of the prediction results, provides a more valuable prediction basis for logistics operation decisions, and avoids decision bias caused by a single prediction value. 4. This invention relies on a multi-dimensional business relationship graph to achieve dynamic access control based on business awareness. It matches the user's identity information with the node position and association boundary in the graph, dynamically controls the data visibility range, and realizes "authorization by position and control by graph". At the same time, it automatically desensitizes sensitive information and records the data access path, operation behavior and decision basis throughout the process. It solves the problems of rough data access control, insufficient protection of sensitive information and difficulty in tracing access behavior in existing platforms. It realizes full-process traceability and auditability from raw data to operational decisions, and ensures the security and compliance of business data use. 5. The prediction model of this invention can output predicted values ​​of robot instruction usage and corresponding confidence intervals for each node of the organization, business line, and network, realizing multi-dimensional and refined prediction results output, rather than the traditional overall summary prediction. Based on the prediction results of each node, logistics companies can carry out targeted robot resource scheduling, business line capability configuration, and network manpower arrangement, effectively improving the level of refined management of logistics operations, reducing operating costs, and improving overall operating efficiency. Attached Figure Description

[0023] Other features, objects, and advantages of the present invention will become more apparent from the following detailed description of non-limiting embodiments with reference to the accompanying drawings: Figure 1 The first flowchart of the use of statistical analysis methods for robot instructions provided in the embodiments of the present invention is shown.

[0024] Figure 2 A second flowchart illustrating the use of statistical analysis methods for robot instructions provided in an embodiment of the present invention.

[0025] Figure 3 This is a schematic diagram of a statistical analysis device for robot instructions provided in an embodiment of the present invention.

[0026] Figure 4 This is a schematic diagram of the structure of an electronic device provided in an embodiment of the present invention. Detailed Implementation

[0027] This invention provides a method, apparatus, device, and medium for statistical analysis of robot instruction usage. The method includes: using organizations, business lines, and outlets as graph nodes; obtaining the relationships between nodes based on actual business rules and historical behavior data as graph edges; dynamically assigning edge weights based on factors such as the frequency of organizational calls to business lines, the frequency of task collaboration between outlets, and the strength of instruction flow dependencies; and constructing a multi-dimensional business relationship graph. This multi-dimensional business relationship graph is used to depict the dynamically changing business dependencies among organizations, business lines, and outlets. Based on the multi-dimensional business relationship graph, a prediction model is used to obtain predicted robot instruction usage values, corresponding confidence intervals, and quantified business impact factors within a preset future time period. This invention solves the problem that existing technologies rely solely on single time-series data for prediction, neglecting key information such as inter-subject collaboration relationships, business affiliation, and task flow, resulting in insufficient accuracy of prediction results and an inability to quantitatively assess the reliability of the prediction results.

[0028] The terms "first," "second," "third," "fourth," etc. (if present) in the specification, claims, and accompanying drawings of this invention are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" or "having" and any variations thereof are intended to cover a non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0029] For ease of understanding, the specific process of the embodiments of the present invention is described below. Please refer to [link / reference]. Figure 1 The first embodiment of the present invention, which uses statistical analysis methods for robot instructions, includes: 101. Using organizations, business lines, and outlets as graph nodes, and obtaining the relationships between nodes based on actual business rules and historical behavior data as graph edges, a multi-dimensional business relationship graph is constructed by dynamically assigning edge weights based on factors such as the frequency of calls from organizations to business lines, the frequency of task collaboration between outlets, and the strength of instruction flow dependencies. The multi-dimensional business relationship graph is used to depict the dynamically changing business dependencies among organizations, business lines, and outlets. The organization mentioned above is the main body responsible for the instruction management and initiation of the intelligent robot platform. It is an enterprise operating entity with hierarchical affiliation and business jurisdiction attributes, including branch offices and regional centers. The business line is the instruction execution carrier that carries standardized business functions on the intelligent robot platform. It is a functional module oriented towards logistics operation scenarios, including network manager, courier pickup and delivery, and intelligent scheduling. The network points are the offline execution entities of the instructions from the intelligent robot platform. They are logistics operation units with geographical location and business coverage attributes, including urban network points, community stations, and last-mile service points. In this embodiment, robot instruction usage logs, organizational structure information, business line association rules, and network topology relationships are collected. The robot instruction usage log is a standardized data log in the intelligent robot platform that records the entire lifecycle behavior of the organization, business lines, and outlets in calling, executing, transferring, and coordinating robot instructions. It includes quantitative information on the subject of the behavior, time, path, and result. Its core purpose is to obtain historical behavior data to provide a core basis for edge generation and weight assignment.

[0030] The organizational structure information is structured data that includes the static attributes and hierarchical associations of various organizational entities, such as internal subsidiaries and regional centers, as well as their jurisdictions and business responsibilities. Its core purpose is to obtain the attributes and hierarchical associations of organizational nodes to ensure the accuracy of organizational nodes and the rationality of their hierarchy.

[0031] The business line association rules are inherent business association specifications within the intelligent robot platform, including the functional positioning, triggering conditions, and upstream and downstream linkage logic of various business lines such as network point managers and courier pickup and delivery. The core purpose is to obtain static association rules for business line nodes and clarify the binding relationships between business lines and between business lines and organizations / network points. For example, the rules clearly state that the intelligent dispatch business line can only be called by organizations at the regional center level and above, and the last-mile service point can only connect to the courier pickup and delivery business line. These rules are important bases for filtering valid associations and eliminating unreasonable edges.

[0032] The network topology relationship includes topological data on the geographical location, hierarchical affiliation, business coverage, business radiation, and connection relationships between urban and community network points; its core purpose is to obtain the attributes and relationships of network point nodes to ensure the accuracy of the spatial attributes and hierarchical associations of network point nodes.

[0033] The collected robot instruction usage logs, organizational structure information, business line association rules, and network topology relationships are preprocessed to obtain preprocessed robot instruction usage logs, organizational structure information, business line association rules, and network topology relationships. In this embodiment, the collected robot instruction usage logs, organizational structure information, business line association rules, and network topology relationships are sequentially cleaned, normalized, and structured mapped to obtain preprocessed robot instruction usage logs, organizational structure information, business line association rules, and network topology relationships. The data cleaning process is used to remove invalid logs from robot instruction usage logs, organizational structure information, erroneous structure information in network topology relationships, and invalid rules in business line association rules, and to supplement missing business attributes that meet preset requirements. The normalization process is used to standardize business data of different formats and units. The structured mapping process is used to transform unstructured and / or semi-structured data into graph-recognizable structured key-value pair data, associating each data type with a unique identifier.

[0034] Based on preprocessed robot instructions, logs, organizational structure information, business line association rules, and network topology, a graph is constructed with organizations, business lines, and networks as nodes. The associations between nodes are obtained based on actual business rules and historical behavior data as graph edges. Edge weights are dynamically assigned based on factors such as the frequency of calls from organizations to business lines, the frequency of task collaboration between networks, and the strength of instruction flow dependencies. This constructs a multi-dimensional business relationship graph, thereby fully depicting the complex and dynamically changing business dependencies among organizations, business lines, and networks. In this embodiment, all three nodes revolve around the intelligent robot platform, each undertaking a role in the entire process of instruction initiation, instruction execution, and offline implementation. The attributes and positioning of the three nodes are matched with each other, forming a closed loop. The graph edges are the embodiment of the business relationships between nodes, and the edge weights are the quantitative representation of the strength of the relationships. Both are generated based on actual business rules and historical behavior data to ensure dynamism and authenticity.

[0035] 102. Based on a multi-dimensional business relationship graph and historical instruction usage time series, a prediction model is used to obtain the predicted value of robot instruction usage within a future preset time period. In this embodiment, the prediction model includes: extracting business structure association features from a multidimensional business relationship graph using a graph attention network; extracting time series change features from historical instruction usage time series using a Transformer encoder; fusing the extracted business structure association features and time series change features; and predicting the robot instruction usage within a preset time period based on the fused features; wherein the historical instruction usage time series is robot instruction usage data for each node ordered chronologically.

[0036] More specifically, obtaining the predicted value of robot instruction usage within a future preset time period includes: By using the adaptive attention mechanism of graph attention network, differentiated attention weights are assigned to three types of nodes—organization, business line, and outlet—as well as the edges between nodes, to capture explicit multidimensional dependencies and implicit associations between nodes. Based on the captured explicit multidimensional dependencies and implicit associations between nodes, structural feature vectors are obtained. The Transformer encoder captures the periodic patterns, trends, and sudden fluctuations in instruction usage using time series analysis of historical instructions. Based on these captured patterns, a time-series feature vector is obtained. In this embodiment, the granularity is first set according to the prediction requirements: hourly granularity for short-term predictions, daily granularity for medium-term predictions, and weekly granularity for long-term predictions, ensuring that the time-series data can capture the changing patterns of the corresponding periods. By dimensionally aligning and cross-dimensionally fusing structural feature vectors and temporal feature vectors, a comprehensive feature vector that integrates business correlation logic and temporal patterns is generated. The comprehensive feature vector is based on the fully connected regression layer of the prediction model. It combines the mapping relationship between the historical comprehensive feature vector and the historical command usage learned during the training phase of the prediction model to predict the robot command usage of each node in the future preset time period.

[0037] This embodiment is based on dual-feature fusion prediction. It captures the static structural dependence and dynamic correlation strength of "organization-business line-network" through a multi-dimensional business relationship graph, and captures the time pattern of instruction usage through historical instruction usage time series. The fusion of the two achieves more accurate and more business-realistic instruction usage prediction than single time series prediction.

[0038] Please see Figure 2 A second embodiment of the present invention, which uses statistical analysis methods for robot instructions, includes: 201. Using organizations, business lines, and outlets as graph nodes, and obtaining the relationships between nodes based on actual business rules and historical behavior data as graph edges, a multi-dimensional business relationship graph is constructed by dynamically assigning edge weights based on factors such as the frequency of calls from organizations to business lines, the frequency of task collaboration between outlets, and the strength of instruction flow dependencies. The multi-dimensional business relationship graph is used to depict the dynamically changing business dependencies among organizations, business lines, and outlets. In this embodiment, the core role of the organization is that of an instruction management and initiating entity, essentially serving as the decision-making and management level of the enterprise. Its hierarchical affiliation, such as head office → regional center → branch office, determines the initiation authority and jurisdiction of instructions, while the business jurisdiction attribute clarifies the business lines that the organization can dispatch and the range of outlets it can manage. For example, a regional center, as an organizational node, can initiate instructions for intelligent dispatching of business lines and manage all city outlets and community sites within its region, ensuring the accuracy and hierarchical rationality of instruction issuance.

[0039] The core positioning of each business line is as an instruction execution carrier, essentially a functional module layer of the intelligent robot platform, and a core bridge connecting the organization and its branches. Its standardized business functions correspond to specific logistics operation scenarios. The differences in triggering conditions and upstream / downstream linkage logic among different business lines determine the diversity of relationships between nodes. For example, the branch management business line is mainly used for daily branch operation management, such as order statistics and inventory inquiries, and relies on route planning data from the intelligent dispatch business line. Meanwhile, the courier pickup and delivery business line directly connects to the offline execution tasks of the branches, forming a linkage link of organization initiation → branch management / intelligent dispatch → courier pickup and delivery → branch execution.

[0040] The core positioning of a delivery network is as the offline execution entity, essentially the execution layer of logistics operations, and the final carrier for the delivery of instructions. Its geographical location and business coverage determine the spatial boundaries and service capabilities of instruction execution. The hierarchical affiliation between network points, such as city network point → community station → last-mile service point, corresponds to the hierarchical breakdown of instruction execution; for example, the overall instruction is broken down and executed by each community station. For instance, a city network point, as a primary execution node, receives pickup and delivery instructions from the organization, breaks them down to community stations within its jurisdiction, and then the last-mile delivery is completed by the last-mile service point, forming a closed-loop hierarchical system for offline execution.

[0041] Edge generation logic: The relationships between edges are not pre-set manually, but are determined by two types of data, including: First, business rules (static relationships), such as the binding relationship between an organization and a business line, such as a branch office only being able to call a specific business line, and the adaptation relationship between a business line and a branch, such as a last-mile service point only connecting to the courier pickup and delivery business line; Second, historical behavior data (dynamic relationships), such as the historical records of an organization calling a business line, the historical records of collaborative task completion between branches, and the historical paths of instructions flowing between different nodes, ensuring that the relationships between edges conform to the actual business operation logic.

[0042] Weight assignment rules: Weight is a dynamic quantitative indicator of edge association strength, calculated through three core indicators, and directly reflects the degree of business dependence between nodes. The specific calculation logic is as follows: Organizational call frequency to business line: The number of times an organization calls a certain business line within a certain period is statistically analyzed. The call duration and instruction completion rate are combined for weighted calculation. The more frequent the call and the higher the completion rate, the greater the weight, reflecting the organization's dependence on the business line. Inter-branch task collaboration frequency: This counts the number of times two branches collaborate to complete the same task within a certain period. The higher the collaboration frequency, the greater the weight, reflecting the closeness of business linkage between branches. Instruction flow dependency strength: The probability of an instruction flowing from one node to another within a certain period is statistically analyzed. The higher the flow probability and the lower the instruction delay rate, the greater the weight, reflecting the rigidity of instruction dependency between nodes.

[0043] Dynamic updates of weights: Edge weights are not fixed values, but are updated regularly as business data is updated, ensuring that the graph can reflect changes in business relationships in real time and avoiding the graph from becoming disconnected from actual business due to business adjustments.

[0044] This embodiment covers all entities from the organization to the business line to the outlets, achieving full coverage of the relationships among the three core entities. It clearly depicts the entire business chain from instruction initiation to function execution to offline implementation, avoiding fragmentation in business relationship analysis. It dynamically assigns edge weights based on historical behavioral data and supports regular updates, which can promptly capture relationship changes brought about by business adjustments, ensuring that the graph is consistent with actual business and avoiding the problem of static graphs being out of touch with business, ensuring that the graph can truly reflect business dependencies.

[0045] 202. Based on a multi-dimensional business relationship graph and historical instruction usage time series, a prediction model is used to obtain the predicted value of robot instruction usage within a future preset time period; In this embodiment, the prediction model employs a graph coding layer, a temporal coding layer, a feature fusion layer, and a fully connected regression layer. Each layer independently completes the feature extraction or fusion task, and the final prediction is achieved through a progressive process. More specifically, the graph encoding layer includes: inputting a multi-dimensional business relationship graph into a graph attention network; and assigning differentiated attention weights to each node and its associated edges through the multi-head attention mechanism of the graph attention network. The core logic is that the closer the relationship, the higher the weight, and the higher the attention ratio. For example, if the edge weight between a regional center and the intelligent dispatch business line is 9 (0-10), and the edge weight between the center and the branch manager business line is 5, then the intelligent dispatch business line will be given a higher attention weight to ensure that the model prioritizes capturing core relationships. It also captures explicit dependencies between nodes, such as the organization-business line binding relationship with clear business rules, and implicit relationships, such as the potential collaborative relationship between two branches reflected in historical data.

[0046] The temporal encoding layer includes: inputting preprocessed historical instructions into the Transformer encoder using time series, and adding temporal position information to each time step through position encoding to ensure that the model can recognize the chronological order; calculating the correlation weights between different time steps through the self-attention mechanism of the Transformer encoder to capture the long-term and short-term dependencies of the time series—for example, capturing the peak of instruction volume on Mondays and Fridays (cyclical pattern), the steady increase of instruction volume over the past 3 months (trend change), and the surge in instruction volume on a certain day due to promotional activities (sudden fluctuation); and performing dimensionality reduction processing on the captured temporal features through the feedforward neural network (FFN) of the Transformer encoder to finally generate a temporal feature vector with the same dimension as the structural feature vector, ensuring that subsequent dimension alignment and fusion can be achieved. The feature fusion layer includes: fusing the structural feature vector output by the graph coding layer with the temporal feature vector output by the temporal coding layer to generate a comprehensive feature vector that combines business association logic and temporal regularity; specifically, by using a fully connected layer to adjust the two types of feature vectors, the structural feature vector and the temporal feature vector, to the same dimension to ensure that there is no fusion deviation caused by dimensional differences; and by using a dynamic weighted fusion method to fuse the structural feature vector and the temporal feature vector to generate a comprehensive feature vector, which contains both the association strength information of organization-business line-network and the temporal variation information of instruction usage.

[0047] The fully connected regression layer includes: inputting the comprehensive feature vector output by the feature fusion layer into the fully connected regression layer, matching the input comprehensive feature vector based on the mapping relationship between the historical comprehensive feature vector and the historical instruction usage learned during the model training phase, and initially outputting the predicted value of the instruction usage of each node within a preset time period in the future; Meanwhile, the fully connected regression layer introduces a Gaussian probability regression mechanism to synchronously output the initial variance corresponding to the predicted value of robot command usage at each node. The initial variance characterizes the original fluctuation of the predicted value. Based on the historical prediction error distribution pattern learned during the training phase of the prediction model, the initial variance is dynamically calibrated to obtain the calibrated variance; the historical prediction error distribution pattern is the statistical distribution characteristic of the error value between the historical prediction mean and the historical actual command usage. Based on the preset confidence level, and using the predicted value and the calibrated variance, the confidence interval corresponding to the predicted value of robot command usage at each node is calculated by the normal distribution statistical method. The width of the confidence interval is used to quantify the confidence level of the corresponding predicted value; wherein, the smaller the interval width, the higher the confidence level of the predicted value, and the larger the interval width, the lower the confidence level of the predicted value.

[0048] 203. Obtain the current user's organization, management scope, and business responsibilities based on the current user's identity information; match the current user's node position and association boundaries in the multidimensional business relationship graph based on the current user's organization, management scope, and business responsibilities; dynamically control the data visibility range based on the matched node position and association boundaries in the multidimensional business relationship graph; simultaneously, automatically anonymize sensitive information that meets preset requirements; and record the data access path, operation behavior, and decision basis when the user accesses data.

[0049] In this embodiment, dynamic business-aware access control relies on a multi-dimensional business relationship graph to achieve fine-grained, dynamically adjustable data access management and full-process auditing. Specifically, based on the current user's identity, organization, management scope, and business responsibilities, the system automatically matches the user's node position and association boundaries in the business relationship graph, dynamically controlling the data visibility range and display granularity. This ensures that branch administrators can only view data related to their own branch and directly associated branches and their organizations, while regional administrators can view summary information within their corresponding regions, achieving refined access management based on "authorization by position and control by graph." Simultaneously, sensitive information is automatically anonymized to prevent leakage of key business indicators, organizational structure information, and branch data. The system also records the entire data access path, operational behavior, and decision-making basis, achieving full-process traceability and auditability from raw data, business graph, model prediction to operational decisions, ensuring secure and compliant data usage.

[0050] The statistical analysis method used for robot instructions in embodiments of the present invention has been described above. The statistical analysis device used for robot instructions in embodiments of the present invention will be described below. Please refer to [link to relevant documentation]. Figure 3 One embodiment of the present invention, which uses a statistical analysis device for robot instructions, includes: The multi-dimensional business relationship graph construction module 301 is used to construct a multi-dimensional business relationship graph by using organizations, business lines, and outlets as graph nodes, and obtaining the relationships between nodes based on actual business rules and historical behavior data as graph edges. Edge weights are dynamically assigned based on factors including the frequency of organization calls to business lines, the frequency of task collaboration between outlets, and the strength of instruction flow dependencies. This multi-dimensional business relationship graph is used to depict the dynamically changing business dependencies among organizations, business lines, and outlets. Specifically, the organization is the instruction management and initiation entity of the intelligent robot platform, and is an enterprise operating entity with hierarchical affiliation and business jurisdiction attributes, including branches and regional centers. The business line is the instruction execution carrier on the intelligent robot platform that carries standardized business functions; it is a functional module oriented towards logistics operation scenarios, including outlet managers, courier pickup and delivery, and intelligent scheduling. The outlet is the offline execution entity of the intelligent robot platform's instructions, and is a logistics operation unit with geographical location and business coverage attributes, including city outlets, community stations, and last-mile service points. The multidimensional business relationship graph construction module 301 includes: The data collection submodule 3011 is used to collect robot instruction usage logs, organizational structure information, business line association rules, and network topology relationships. The robot instruction usage log is a standardized data log in the intelligent robot platform that records the entire lifecycle behavior of the organization, business lines, and outlets in calling, executing, transferring, and coordinating robot instructions. It includes quantitative information on the subject of the behavior, time, path, and result. The organizational structure information is structured data that includes the static attributes and hierarchical associations of various organizational entities, such as internal subsidiaries and regional centers, as well as their hierarchical affiliations, jurisdictions, and business responsibilities. The business line association rules are the inherent business association specifications within the intelligent robot platform, including the functional positioning, triggering conditions, and upstream and downstream linkage logic of various business lines such as network managers and courier pickup and delivery. The network topology is topological data that includes the geographical location, hierarchical affiliation, business coverage, business radiation and connection relationships between urban and community sites. The preprocessing submodule 3012 is used to preprocess the collected robot instruction usage logs, organizational structure information, business line association rules and network topology relationships respectively to obtain the preprocessed robot instruction usage logs, organizational structure information, business line association rules and network topology relationships. In this embodiment, the preprocessing submodule 3012 includes: The collected robot instruction usage logs, organizational structure information, business line association rules, and network topology relationships are sequentially cleaned, normalized, and structured mapped to obtain preprocessed robot instruction usage logs, organizational structure information, business line association rules, and network topology relationships. The data cleaning process is used to remove invalid logs from robot instruction usage logs, organizational structure information, erroneous structure information in network topology relationships, and invalid rules in business line association rules, and to supplement missing business attributes that meet preset requirements. The normalization process is used to standardize business data of different formats and units. The structured mapping process is used to transform unstructured and / or semi-structured data into graph-recognizable structured key-value pair data, associating each data type with a unique identifier.

[0051] The relationship graph construction submodule 3013 is used to construct a multi-dimensional business relationship graph based on preprocessed robot instruction usage logs, organizational structure information, business line association rules, and network topology relationships. Organizations, business lines, and networks are used as graph nodes, and the associations between nodes are obtained based on actual business rules and historical behavior data as graph edges. Edge weights are dynamically assigned by factors including the frequency of organization calls to business lines, the frequency of task collaboration between networks, and the strength of instruction flow dependencies.

[0052] The instruction usage prediction module 302 is used to obtain the predicted value of robot instruction usage within a preset time period based on a multi-dimensional business relationship graph and historical instruction usage time series through a prediction model. In this embodiment, the instruction usage prediction module 302 includes: The structural feature vector acquisition submodule 3021 is used to assign differentiated attention weights to three types of nodes (organization, business line, and network point) and the edges between nodes through the adaptive attention mechanism of graph attention network, capture the explicit multidimensional dependency and implicit association features between nodes, and obtain the structural feature vector based on the captured explicit multidimensional dependency and implicit association features between nodes. The time-series feature vector acquisition submodule 3022 is used to capture the periodic patterns, trend changes and sudden fluctuations of instruction usage using time series data from historical instructions through the Transformer encoder, and obtain the time-series feature vector based on the captured periodic patterns, trend changes and sudden fluctuations of instruction usage. The feature fusion submodule 3023 is used to perform dimensional alignment and cross-dimensional fusion of structural feature vectors and temporal feature vectors to generate a comprehensive feature vector that integrates business correlation logic and time patterns. The prediction submodule 3024 is used to predict the robot command usage of each node in a future preset time period by using the fully connected regression layer of the prediction model based on the comprehensive feature vector and combining the mapping relationship between the historical comprehensive feature vector and the historical command usage learned during the training phase of the prediction model. The credibility assessment submodule 3025 is used to assess the credibility of the predicted values ​​of robot instruction usage at each node based on the confidence interval. In this embodiment, the credibility assessment submodule 3025 includes: The fully connected regression layer introduces a Gaussian probability regression mechanism to synchronously output the initial variance corresponding to the predicted value of robot command usage at each node. The initial variance characterizes the original fluctuation of the predicted value. Based on the historical prediction error distribution pattern learned during the training phase of the prediction model, the initial variance is dynamically calibrated to obtain the calibrated variance; the historical prediction error distribution pattern is the statistical distribution characteristic of the error value between the historical prediction mean and the historical actual command usage. Based on the preset confidence level, and using the predicted value and the calibrated variance, the confidence interval corresponding to the predicted value of robot command usage at each node is calculated by the normal distribution statistical method. The width of the confidence interval is used to quantify the confidence level of the corresponding predicted value; wherein, the smaller the interval width, the higher the confidence level of the predicted value, and the larger the interval width, the lower the confidence level of the predicted value.

[0053] The business-aware access control module 303 is used to obtain the organization, management scope, and business responsibilities of the current user based on the identity information of the current user; match the node position and association boundary of the current user in the multi-dimensional business relationship graph based on the organization, management scope, and business responsibilities of the current user; and dynamically control the data visibility range based on the matched node position and association boundary of the current user in the multi-dimensional business relationship graph. At the same time, sensitive information that meets the preset requirements is automatically de-identified; when users access data, the data access path, operation behavior and decision basis are recorded.

[0054] above Figure 3 The robot instructions in this embodiment of the invention are described in detail from the perspective of modular functional entities using statistical analysis devices. The electronic devices in this embodiment of the invention are described in detail from the perspective of hardware processing.

[0055] Figure 4 This is a schematic diagram of the structure of an electronic device 700 provided in an embodiment of the present invention. The electronic device 700 can vary significantly due to different configurations or performance characteristics. It may include one or more central processing units (CPUs) 710 (e.g., one or more processors) and a memory 720, and one or more storage media 730 (e.g., one or more mass storage devices) for storing application programs 733 or data 732. The memory 720 and storage media 730 can be temporary or persistent storage. The program stored in the storage media 730 may include one or more modules (not shown in the diagram), each module including a series of instruction operations on the electronic device 700. Furthermore, the processor 710 may be configured to communicate with the storage media 730 and execute the series of instruction operations in the storage media 730 on the electronic device 700.

[0056] Electronic device 700 may also include one or more power supplies 740, one or more wired or wireless network interfaces 750, one or more input / output interfaces 750, and / or one or more operating systems 731, such as Windows Server, Mac OS X, Unix, Linux, FreeBSD, etc. Those skilled in the art will understand that... Figure 4 The illustrated electronic device structure does not constitute a limitation on electronic devices and may include more or fewer components than illustrated, or combine certain components, or have different component arrangements.

[0057] The present invention also provides a computer-readable storage medium, which can be a non-volatile computer-readable storage medium or a volatile computer-readable storage medium, wherein the computer-readable storage medium stores instructions that, when executed on a computer, cause the computer to perform the steps of the robot instructions using statistical analysis methods.

[0058] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working process of the system, device, or unit described above can be referred to the corresponding process in the foregoing method embodiments, and will not be repeated here.

[0059] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0060] The above-described embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A statistical analysis method for robot instructions, characterized in that, include: Using organizations, business lines, and outlets as graph nodes, and obtaining the relationships between nodes based on actual business rules and historical behavior data as graph edges, a multi-dimensional business relationship graph is constructed by dynamically assigning edge weights based on factors such as the frequency of calls from organizations to business lines, the frequency of task collaboration between outlets, and the strength of instruction flow dependencies. This multi-dimensional business relationship graph is used to depict the dynamically changing business dependencies among organizations, business lines, and outlets. The organization mentioned above is the main body responsible for the instruction management and initiation of the intelligent robot platform. It is an enterprise operating entity with hierarchical affiliation and business jurisdiction attributes, including branch offices and regional centers. The business line is the instruction execution carrier that carries standardized business functions on the intelligent robot platform. It is a functional module oriented towards logistics operation scenarios, including network manager, courier pickup and delivery, and intelligent scheduling. The network points are the offline execution entities of the instructions from the intelligent robot platform. They are logistics operation units with geographical location and business coverage attributes, including urban network points, community stations, and last-mile service points. Based on a multidimensional business relationship graph and historical instruction usage time series, a prediction model is used to obtain the predicted value of robot instruction usage within a future preset time period. The prediction model includes a graph coding layer for extracting business structure features and a temporal coding layer for extracting time features. The robot instruction usage is predicted based on the fusion features of the graph coding layer and the temporal coding layer.

2. The robot instruction using statistical analysis method according to claim 1, characterized in that, The graph uses organizations, business lines, and outlets as nodes, and obtains the relationships between nodes based on actual business rules and historical behavior data as edges. The edge weights are dynamically assigned based on factors such as the frequency of calls from organizations to business lines, the frequency of task collaboration between outlets, and the strength of instruction flow dependencies, to construct a multi-dimensional business relationship graph. The multidimensional business relationship graph is used to depict the dynamically changing business dependencies among the organization, business lines, and outlets, including: Collect robot instruction usage logs, organizational structure information, business line association rules, and branch network topology relationships; The robot instruction usage log is a standardized data log in the intelligent robot platform that records the entire lifecycle behavior of the organization, business lines, and outlets in calling, executing, transferring, and coordinating robot instructions. It includes quantitative information on the subject of the behavior, time, path, and result. The organizational structure information is structured data that includes the static attributes and hierarchical associations of various organizational entities, such as internal subsidiaries and regional centers, as well as their hierarchical affiliations, jurisdictions, and business responsibilities. The business line association rules are the inherent business association specifications within the intelligent robot platform, including the functional positioning, triggering conditions, and upstream and downstream linkage logic of various business lines such as network managers and courier pickup and delivery. The network topology is topological data that includes the geographical location, hierarchical affiliation, business coverage, business radiation and connection relationships between urban and community sites. The collected robot instruction usage logs, organizational structure information, business line association rules, and network topology relationships are preprocessed to obtain preprocessed robot instruction usage logs, organizational structure information, business line association rules, and network topology relationships. Based on preprocessed robot instruction usage logs, organizational structure information, business line association rules, and network topology, a multi-dimensional business relationship graph is constructed, with organizations, business lines, and networks as graph nodes. The relationships between nodes are obtained based on actual business rules and historical behavior data as graph edges. Edge weights are dynamically assigned based on factors such as the frequency of organization calls to business lines, the frequency of task collaboration between networks, and the strength of instruction flow dependencies.

3. The robot instructions according to claim 2 use a statistical analysis method, characterized in that, The collected robot instruction usage logs, organizational structure information, business line association rules, and network topology relationships are preprocessed to obtain preprocessed robot instruction usage logs, organizational structure information, business line association rules, and network topology relationships, including: The collected robot instruction usage logs, organizational structure information, business line association rules, and network topology relationships are sequentially cleaned, normalized, and structured mapped to obtain preprocessed robot instruction usage logs, organizational structure information, business line association rules, and network topology relationships. The data cleaning process is used to remove invalid logs from robot instruction usage logs, organizational structure information, erroneous structure information in network topology relationships, and invalid rules in business line association rules, and to supplement missing business attributes that meet preset requirements. The normalization process is used to standardize business data of different formats and units. The structured mapping process is used to transform unstructured and / or semi-structured data into graph-recognizable structured key-value pair data, associating each data type with a unique identifier.

4. The robot instruction using statistical analysis method according to claim 1, characterized in that, The prediction model includes: extracting business structure association features from a multidimensional business relationship graph using a graph attention network; extracting time series change features from historical instructions using a Transformer encoder; fusing the extracted business structure association features and time series change features; and predicting the amount of robot instructions used in a future preset time period based on the fused features.

5. The robot instructions according to claim 4 use a statistical analysis method, characterized in that, The step of obtaining the predicted value of robot command usage within a future preset time period includes: By using the adaptive attention mechanism of graph attention network, differentiated attention weights are assigned to three types of nodes—organization, business line, and outlet—as well as the edges between nodes, to capture explicit multidimensional dependencies and implicit associations between nodes. Based on the captured explicit multidimensional dependencies and implicit associations between nodes, structural feature vectors are obtained. The Transformer encoder captures the periodic patterns, trends, and sudden fluctuations in instruction usage using time series analysis of historical instructions. Based on the captured periodic patterns, trends, and sudden fluctuations in instruction usage, a time series feature vector is obtained. By dimensionally aligning and cross-dimensionally fusing structural feature vectors and temporal feature vectors, a comprehensive feature vector that integrates business correlation logic and temporal patterns is generated. The comprehensive feature vector is based on the fully connected regression layer of the prediction model. It combines the mapping relationship between the historical comprehensive feature vector and the historical command usage learned during the training phase of the prediction model to predict the robot command usage of each node in the future preset time period.

6. The robot instructions according to claim 5 use a statistical analysis method, characterized in that, The prediction model also includes: The fully connected regression layer introduces a Gaussian probability regression mechanism to synchronously output the initial variance corresponding to the predicted value of robot command usage at each node. The initial variance characterizes the original fluctuation of the predicted value. Based on the historical prediction error distribution pattern learned during the training phase of the prediction model, the initial variance is dynamically calibrated to obtain the calibrated variance; the historical prediction error distribution pattern is the statistical distribution characteristic of the error value between the historical prediction mean and the historical actual command usage. Based on the preset confidence level, and using the predicted value and the calibrated variance, the confidence interval corresponding to the predicted value of robot command usage at each node is calculated by the normal distribution statistical method. The width of the confidence interval is used to quantify the confidence level of the corresponding predicted value; wherein, the smaller the interval width, the higher the confidence level of the predicted value, and the larger the interval width, the lower the confidence level of the predicted value.

7. The robot instruction according to claim 1 uses a statistical analysis method, characterized in that, The method further includes: obtaining the organization, management scope, and business responsibilities of the current user based on the identity information of the current user; matching the node position and association boundary of the current user in the multidimensional business relationship graph based on the organization, management scope, and business responsibilities of the current user; and dynamically controlling the data visibility range based on the matched node position and association boundary of the current user in the multidimensional business relationship graph. At the same time, sensitive information that meets preset requirements is automatically de-identified; when users access data, the data access path, operation behavior, and decision basis are recorded.

8. A robot instruction using a statistical analysis device, characterized in that, include: The multi-dimensional business relationship graph construction module uses organizations, business lines, and outlets as graph nodes, and obtains the relationships between nodes as graph edges based on actual business rules and historical behavior data. The edge weights are dynamically assigned by factors including the frequency of organization calls to business lines, the frequency of task collaboration between outlets, and the strength of instruction flow dependencies. The multi-dimensional business relationship graph is used to depict the dynamically changing business dependencies among organizations, business lines, and outlets. The organization mentioned above is the main body responsible for the instruction management and initiation of the intelligent robot platform. It is an enterprise operating entity with hierarchical affiliation and business jurisdiction attributes, including branch offices and regional centers. The business line is the instruction execution carrier that carries standardized business functions on the intelligent robot platform. It is a functional module oriented towards logistics operation scenarios, including network manager, courier pickup and delivery, and intelligent scheduling. The network points are the offline execution entities of the instructions from the intelligent robot platform. They are logistics operation units with geographical location and business coverage attributes, including urban network points, community stations, and last-mile service points. The instruction usage prediction module is used to obtain the predicted value of robot instruction usage within a preset time period based on a multi-dimensional business relationship graph and historical instruction usage time series through a prediction model.

9. An electronic device comprising a memory and at least one processor, wherein the memory stores instructions; The at least one processor invokes the instructions in the memory to cause the electronic device to perform the steps of the robot instructions using the statistical analysis method as claimed in any one of claims 1-7.

10. A computer-readable storage medium storing instructions thereon, characterized in that, When the instruction is executed by the processor, it implements each step of the statistical analysis method used in the robot instruction as described in any one of claims 1-7.

Images