Express performance quality monitoring data processing method and device

Abstract

The present application relates to the field of express delivery performance quality monitoring, and discloses an express delivery performance quality monitoring data processing method and device. The method comprises: accessing multi-source heterogeneous service data; extracting multi-element features, including spatial dimension features, performance index dimension features, time sequence dimension features and service label dimension features; the performance index dimension features include a plurality of performance indexes, and the service label dimension features include a signature type, a work order type, and whether to cancel delivery; obtaining a warning threshold for each performance index, comparing real-time data of the performance index with the warning threshold, marking an abnormal index in combination with the time sequence dimension features and the spatial dimension features, and triggering a warning; and performing a reverse depth-first search, tracing a starting node including the marked abnormal index, sequentially connecting and querying service label dimension features under corresponding dimensions, and restoring a business link where the abnormality occurs, so as to solve the problem of lacking intelligent abnormality identification and tracing capabilities.

Description

Technical Field

[0001] This invention relates to the field of express delivery fulfillment quality monitoring, and in particular to a method and apparatus for processing express delivery fulfillment quality monitoring data. Background Technology

[0002] Currently, the express delivery industry's fulfillment reports generally adopt a traditional static statistical architecture, resulting in scattered data sources and weak data integration capabilities. The mainstream monitoring methods still rely on manual screening, lacking intelligent anomaly identification and traceability capabilities. Faced with the high-concurrency business of millions of orders per day for dedicated delivery services, the traditional technical architecture can no longer meet the business needs of real-time monitoring, anomaly warning, and rapid traceability, resulting in common pain points in the industry such as lagging management, low efficiency, and high misjudgment rate.

[0003] Therefore, there is an urgent need to provide a data processing method and device for monitoring express delivery fulfillment quality, which can effectively reduce service quality risks, improve customer service quality, achieve refined and differentiated operation and management, and ensure stable fulfillment services. Summary of the Invention

[0004] The main purpose of this invention is to solve the problems of lack of intelligent anomaly identification and tracing capabilities, which cannot meet the business needs of real-time monitoring, anomaly warning, and rapid tracing, resulting in lagging management, low efficiency, and high misjudgment rate.

[0005] The first aspect of this invention provides a method for processing express delivery fulfillment quality monitoring data, comprising: accessing multi-source heterogeneous business data; extracting multi-dimensional features, including spatial dimension features, fulfillment indicator dimension features, temporal dimension features, and business tag dimension features, wherein the spatial dimension features record the physical ownership of each data item, the fulfillment indicator dimension features include several fulfillment indicators, the temporal dimension features depict the trend of fulfillment indicators changing over time, and the business tag dimension features include the type of receipt, the type of work order, and whether door-to-door delivery has been cancelled; obtaining the warning threshold for each fulfillment indicator, comparing the real-time data of the fulfillment indicators with the warning threshold, marking abnormal indicators by combining the temporal dimension features and spatial dimension features, and triggering a warning; performing a reverse depth-first search, wherein the starting node for tracing the source includes the marked abnormal indicators, prioritizing the traversal of core related entities with higher weights, sequentially querying the business tag dimension features under the corresponding dimensions, and reconstructing the business link that caused the abnormality.

[0006] Optionally, in a first implementation of the first aspect of the present invention, the access to multi-source heterogeneous business data includes: performing filtering of invalid data and data format standardization processing, wherein invalid data includes returned data, and data format standardization processing includes unifying the time format, numerical format, and encoding rules of multi-source heterogeneous business data.

[0007] Optionally, in a second implementation of the first aspect of the present invention, the access to multi-source heterogeneous business data includes: associating multi-source heterogeneous business data scattered across different platforms, wherein the data association uses the waybill number, branch code, and salesperson code as globally unique association keys.

[0008] Optionally, in a third implementation of the first aspect of the present invention, the extraction of multi-dimensional features includes: introducing a dynamic weighted fusion algorithm, combining business weight configuration rules, assigning high weights to core risk indicators among the performance indicators, and assigning basic weights to other indicators.

[0009] Optionally, in the fourth implementation of the first aspect of the present invention, obtaining the early warning threshold of each performance indicator includes: retrieving historical characteristic data of each performance indicator within a preset time period, using an exponential moving average algorithm combined with a quantile statistical algorithm for joint modeling, calculating the long-term historical average of each performance indicator, defining the normal fluctuation range of the indicator, and generating an early warning threshold.

[0010] Optionally, in the fifth implementation of the first aspect of the present invention, the step of comparing the real-time data of the performance indicators with the early warning threshold and marking abnormal indicators by combining time-series and spatial features includes: taking the current day as the end of the window, continuously capturing the time-series data of the indicators for 7 consecutive days, calculating the fluctuation amplitude, fluctuation frequency, and offset duration of the data within the window; simultaneously introducing a Markov chain state transition model to divide the data state into several discrete states, and analyzing the evolution law of the data state within the continuous window through the state transition probability matrix: if only a single window shows a data offset, it is determined to be a normal short-term fluctuation or an occasional anomaly; if multiple consecutive sliding windows maintain an abnormal state and the state transition probability approaches 1, it is determined to be a continuous operational failure; if multiple outlets in the same province / city simultaneously show abnormal indicators, it is determined to be a regional batch problem.

[0011] Optionally, in the sixth implementation of the first aspect of the present invention, the execution of reverse depth-first search includes: after determining the abnormal indicators, calling a pre-built logistics fulfillment knowledge graph, whose entity nodes contain spatial dimension features and business tag dimension features, whose relational edges define the logical relationships between entity nodes, and configuring relational weights and relational directions for each relational edge; using the marked abnormal indicators and abnormal spatial dimensions as the source tracing starting nodes, starting from the starting nodes, and using the reverse depth-first search algorithm to perform a full-link traversal; during the traversal process, data cross-validation and logical reasoning are performed synchronously, and the abnormal trigger point is located by combining the indicator calculation formula and business rules.

[0012] The second aspect of this invention provides a data processing device for monitoring express delivery fulfillment quality, comprising: a data processing module for accessing multi-source heterogeneous business data; an extraction module for extracting multi-dimensional features, including spatial dimension features, fulfillment indicator dimension features, temporal dimension features, and business tag dimension features, wherein the spatial dimension features record the physical ownership of each data item, the fulfillment indicator dimension features include several fulfillment indicators, the temporal dimension features depict the trend of fulfillment indicators changing over time, and the business tag dimension features convert several discrete tags into feature values, wherein the discrete tags include the type of receipt, the type of work order, and whether door-to-door delivery has been cancelled; an early warning module for obtaining the early warning threshold for each fulfillment indicator, comparing the real-time data of the fulfillment indicator with the early warning threshold, marking abnormal indicators by combining the temporal dimension features and spatial dimension features, and triggering an early warning; and a search and restoration module for performing a reverse depth-first search, wherein the starting node for tracing the source includes the marked abnormal indicators, prioritizing the tracing of core related entities with higher weights, sequentially querying the business tag dimension features under the corresponding dimension, and restoring the business link that caused the abnormality.

[0013] Optionally, in the first implementation of the second aspect of the present invention, the data processing module is further configured to perform filtering of invalid data and data format standardization processing. Invalid data includes returned data, and data format standardization processing includes unifying the time format, numerical format, and encoding rules of multi-source heterogeneous business data.

[0014] Optionally, in a second implementation of the second aspect of the present invention, the data processing module is further configured to associate multi-source heterogeneous business data scattered across different platforms, wherein the data association uses the waybill number, branch code, and salesperson code as globally unique association keys.

[0015] Optionally, in a third implementation of the second aspect of the present invention, the extraction module further includes a dynamic weighted fusion unit, which is used to assign high weights to the core risk indicators in the performance indicators and basic weights to other indicators.

[0016] Optionally, in the fourth implementation of the second aspect of the present invention, the early warning module further includes a dynamic threshold generation unit. The dynamic threshold generation unit retrieves historical characteristic data of each performance indicator within a preset time period, uses an exponential moving average algorithm combined with a quantile statistical algorithm to jointly model, calculates the long-term historical average of each performance indicator, delineates the normal fluctuation range of the indicator, and generates an early warning threshold.

[0017] Optionally, in the fifth implementation of the second aspect of the present invention, the early warning module takes the current day as the end of the window, continuously captures the time series data of indicators for 7 consecutive days, and calculates the fluctuation amplitude, fluctuation frequency, and offset duration of the data within the window; at the same time, a Markov chain state transition model is introduced to divide the data state into several discrete states, and the evolution law of the data state within the continuous window is analyzed through the state transition probability matrix: if only a single window shows a data offset, it is judged as a normal short-term fluctuation or an occasional anomaly; if multiple consecutive sliding windows maintain an abnormal state and the state transition probability approaches 1, it is judged as a continuous operational failure; if multiple outlets in the same province / city simultaneously show abnormal indicators, it is judged as a regional batch problem.

[0018] Optionally, in the sixth implementation of the second aspect of the present invention, after determining the abnormal indicators, the search and restoration module calls a pre-built logistics fulfillment knowledge graph, whose entity nodes include spatial dimension features and business tag dimension features, whose relational edges define the logical relationships between entity nodes, and configures the relational weight and relational direction for each relational edge; taking the marked abnormal indicators and abnormal spatial dimensions as the source tracing starting nodes, starting from the starting nodes, the reverse depth-first search algorithm is used to perform full-link traversal; during the traversal process, data cross-validation and logical reasoning are performed synchronously, and the abnormal trigger point is located by combining the indicator calculation formula and business rules.

[0019] A third aspect of the present invention provides an electronic device, comprising: a memory and at least one processor, wherein the memory stores instructions, and the memory and the at least one processor are interconnected via a circuit; the at least one processor invokes the instructions in the memory to cause the electronic device to execute the express delivery quality monitoring data processing method described above.

[0020] A fourth aspect of the present invention provides a computer-readable storage medium storing instructions that, when executed on a computer, cause the computer to perform the express delivery quality monitoring data processing method described above.

[0021] In the technical solution of this invention, by accessing multi-source heterogeneous business data, returning items are removed according to preset rules, and full data cleaning, format standardization and data association are completed. Four-element features of spatial dimension, performance indicator dimension, time sequence dimension and business tag dimension are extracted, and each performance indicator is identified. The entire chain traceability of waybill, personnel, work order and signing link is carried out, which solves the problem of lack of intelligent anomaly identification and traceability capabilities in the prior art. Attached Figure Description

[0022] Figure 1 A flowchart of a data processing method for monitoring express delivery fulfillment quality provided in an embodiment of the present invention; Figure 2 This is a schematic diagram of a data processing device for monitoring express delivery fulfillment quality, provided in an embodiment of the present invention. Figure 3 This is a schematic diagram of the structure of an electronic device provided in an embodiment of the present invention; Detailed Implementation This invention provides a method and apparatus for processing express delivery fulfillment quality monitoring data. By accessing multi-source heterogeneous business data, returning items are removed according to preset rules. The process involves full data cleaning, format standardization, and data association. Four-dimensional features—spatial, fulfillment indicator, temporal, and business tag dimensions—are extracted to identify various fulfillment indicators and enable full-chain traceability of waybills, personnel, work orders, and receipt stages. This invention solves the problems of lacking intelligent anomaly identification and traceability capabilities, resulting in lagging management, low efficiency, and high misjudgment rates.

[0023] 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.

[0024] 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 a method for processing express delivery fulfillment quality monitoring data in this invention includes: 110. Integrate multi-source heterogeneous business data; then perform invalid data filtering (full data cleaning, data format standardization, and data association; specifically, First, the system connects to multi-source heterogeneous business data in parallel via interfaces, pulling various types of raw business data, such as waybill scan tables, work order detail tables, AI intelligent outbound call record tables, branch file tables, and salesperson information tables. Then, it performs full data retrieval and temporary caching operations. Then, invalid data filtering is performed according to the established business rules: invalid data includes returned item data. In this embodiment, based on the scanning table rules, returned item data with scanning type 88 and abnormal receipt type 15 / 49 are accurately filtered and removed. After removing dirty data, perform operations such as filling empty values, deduplicating duplicate data, filtering illegal fields, and removing special characters in sequence. Subsequently, data format standardization processing was carried out to unify the time format, numerical format, and encoding rules of all multi-source heterogeneous business data, and to align the statistical time granularity of all data to the T-1 day caliber required by the business, so as to ensure the consistency of the time series statistical benchmark. During the data association phase, the waybill number, branch code, and salesperson code are used as globally unique association primary keys. Multi-table lookup and primary key mapping algorithms are adopted to connect the entire chain of data, including waybill, distribution, receipt, work order, and AI outbound calls, and to stitch together isolated data scattered across different platforms into complete single waybill lifecycle data.

[0025] 120. Extract quaternary features and construct a multidimensional feature matrix; The four-dimensional features include spatial dimension features, performance indicator dimension features, time series dimension features, and business tag dimension features. Feature extraction, encoding, and quantification are performed in layers according to the above four categories. Spatial Dimension Features: Traversing six spatial levels—region, province, city, primary outlet, signing outlet, and salesperson—a one-hot encoding algorithm is used to digitize different regions, outlets, and personnel, transforming textual geographic and subject information into discrete feature vectors that the algorithm can recognize, and recording the physical ownership of each data point. Performance indicator dimension characteristics: It includes several performance indicators. This embodiment focuses on 29 core performance indicators defined by the business, including signed receipts, door-to-door visits, warehousing, various percentages, work order statistics, month-on-month values, etc., and completes indicator value extraction and dimensionless normalization processing. Time-series features: Based on continuous time-series data from the past year, a timestamp sequence is generated with the natural day as the smallest time unit. Derived time-series features such as daily indicator values, weekly averages, monthly averages, historical values ​​for the same period, and time-series fluctuation differences are extracted to construct a continuous time-series feature chain, depicting the trend of performance indicators changing over time. Business tag dimension features: Extract discrete classification tags such as receipt type, work order type, whether delivery to the door was canceled, AI outbound call connection status, and outbound call intent. In some embodiments, a tag quantization algorithm is used to convert the above classification tags into feature values.

[0026] 130. Obtain the early warning thresholds for each performance indicator, compare the real-time data of the performance indicators with the early warning thresholds, and mark abnormal indicators by combining time-series and spatial features. Identify the anomaly types and trigger early warnings in a tiered manner, including: Using the current day as the end of the window, the time series data of the indicator is continuously extracted for 7 consecutive days. The fluctuation amplitude, fluctuation frequency, and offset duration of the data within the window are calculated. At the same time, a Markov chain state transition model is introduced to divide the data state into several discrete states, and the evolution pattern of the data state within the continuous window is analyzed through the state transition probability matrix. If only a single window shows data offset, it is judged as normal short-term fluctuation or occasional anomaly; if multiple sliding windows remain in an abnormal state and the state transition probability approaches 1, it is judged as a continuous operational failure (abnormal indicators exist); if multiple outlets in the same province / city simultaneously show abnormal indicators, it is judged as a regional batch problem (abnormal indicators and abnormal spatial dimensions exist).

[0027] 140. For abnormal indicators, perform a reverse depth-first search based on the entity-relationship graph, prioritize traversing the core related entities with higher weights, and sequentially query the business tag dimension features under the corresponding dimension to restore the business link that caused the anomaly and complete the full-link traceability of waybill, personnel, work order and receipt. Specifically, after identifying abnormal indicators and triggering an alert, the pre-built logistics fulfillment knowledge graph is invoked. Its entity nodes include spatial dimension features and business tag dimension features (region, province, city, outlet, salesperson, waybill, work order, receipt record, AI outbound call record, etc., covering all business entities). Its relational edges define the logical relationships between entity nodes (distribution, receipt, order dispatch, customer complaint, outbound call, etc., and configure the association weight and association direction for each relational edge). Using the marked abnormal indicators and abnormal spatial dimensions as the starting nodes for tracing the source, the reverse depth-first search algorithm is used to perform a full-link traversal: starting from the starting node, the algorithm recursively searches downwards along the reverse association edges of the graph, prioritizing the traversal of core related entities with higher weights, and sequentially querying all data in the corresponding dimensions, including full waybill details, waybill distribution records, salesperson operation records, platform work order records, actual receipt details, and AI intelligent outbound call interaction records, to completely reconstruct the entire business chain that caused the anomaly.

[0028] In this embodiment, by accessing multi-source heterogeneous business data, returning items are removed according to preset rules, and full data cleaning, format standardization and data association are completed. Four-element features of spatial dimension, performance indicator dimension, time sequence dimension and business tag dimension are extracted to identify each performance indicator and perform full-link traceability of waybill, personnel, work order and signing process. This method greatly improves the intelligent anomaly identification and traceability capabilities.

[0029] A second embodiment of a method for processing express delivery fulfillment quality monitoring data in this invention includes: 110. Access multi-source heterogeneous business data; then perform invalid data filtering (full data cleaning, data format standardization processing, and data association); First, the system connects to multi-source heterogeneous business data in parallel via interfaces, pulling various types of raw business data, such as waybill scan tables, work order detail tables, AI intelligent outbound call record tables, branch file tables, and salesperson information tables. Then, it performs full data retrieval and temporary caching operations. Next, invalid data filtering is performed according to established business rules: invalid data includes returned item data; then, data format standardization processing is carried out to unify the time format, numerical format, and encoding rules of all multi-source heterogeneous business data; in the data association stage, the waybill number, branch code, and salesperson code are used as globally unique association primary keys, and multi-table lookup and primary key mapping algorithms are adopted to connect the entire chain of data of waybills, distribution, receipt, work orders, and AI outbound calls, and stitch together the isolated data scattered on different platforms into complete single waybill full lifecycle data.

[0030] 120. Extract quaternary features and construct a multidimensional feature matrix; The four-dimensional features include spatial dimension features, performance indicator dimension features, time series dimension features, and business tag dimension features. Feature extraction, encoding, and quantification are performed in layers according to the above four categories. Spatial Dimension Features: Traversing six spatial levels—region, province, city, primary outlet, signing outlet, and salesperson—a one-hot encoding algorithm is used to digitize different regions, outlets, and personnel, transforming textual geographic and subject information into discrete feature vectors that the algorithm can recognize, and recording the physical ownership of each data point. Performance indicator dimension characteristics: It includes several performance indicators. This embodiment focuses on 29 core performance indicators defined by the business, including signed receipts, door-to-door visits, warehousing, various percentages, work order statistics, month-on-month values, etc., and completes indicator value extraction and dimensionless normalization processing. Time-series features: Based on continuous time-series data from the past year, a timestamp sequence is generated with the natural day as the smallest time unit. Derived time-series features such as daily indicator values, weekly averages, monthly averages, historical values ​​for the same period, and time-series fluctuation differences are extracted to construct a continuous time-series feature chain, depicting the trend of performance indicators changing over time. Business tag dimension features: Extract discrete classification tags such as receipt type, work order type, whether delivery to the door is cancelled, AI outbound call connection status, and outbound call intent. In some embodiments, a tag quantization algorithm is used to convert the above classification tags into feature values. In addition, a dynamic weighted fusion algorithm is introduced. Combined with the business weight configuration rules, high weights are assigned to the core risk indicators in the performance indicators, and basic weights (low weights) are assigned to other indicators. According to the rule that rows correspond to single waybills / statistical dimensions and columns correspond to feature items, all fused feature vectors are matrix-concatenated to generate a structured four-element multidimensional feature matrix.

[0031] 130. Obtain the early warning thresholds for each performance indicator, compare the real-time data of the performance indicators with the early warning thresholds, and mark abnormal indicators by combining time-series and spatial features. Identify the anomaly types and trigger early warnings in a tiered manner, including: Using the current day as the end of the window, the time series data of the indicator is continuously extracted for 7 consecutive days. The fluctuation amplitude, fluctuation frequency, and offset duration of the data within the window are calculated. At the same time, a Markov chain state transition model is introduced to divide the data state into several discrete states, and the evolution pattern of the data state within the continuous window is analyzed through the state transition probability matrix. If only a single window shows data offset, it is judged as normal short-term fluctuation or occasional anomaly; if multiple sliding windows remain in an abnormal state and the state transition probability approaches 1, it is judged as a continuous operational failure; if multiple outlets in the same province / city show abnormal indicators simultaneously, it is judged as a regional batch problem.

[0032] 140. For abnormal indicators, perform a reverse depth-first search based on the entity-relationship graph, prioritize traversing the core related entities with higher weights, and sequentially query the business tag dimension features under the corresponding dimension to restore the business link that caused the anomaly and complete the full-link traceability of waybill, personnel, work order and receipt. Specifically, after identifying abnormal indicators and triggering an alert, the pre-built logistics fulfillment knowledge graph is invoked. Its entity nodes include spatial dimension features and business tag dimension features (region, province, city, outlet, salesperson, waybill, work order, receipt record, AI outbound call record, etc., covering all business entities). Its relational edges define the logical relationships between entity nodes (distribution, receipt, order dispatch, customer complaint, outbound call, etc., and configure the association weight and association direction for each relational edge). Using the marked abnormal indicators and abnormal spatial dimensions as the starting nodes for tracing the source, the reverse depth-first search algorithm is used to perform a full-link traversal: starting from the starting node, the algorithm recursively searches downwards along the reverse association edges of the graph, prioritizing the traversal of core association entities with higher weights, and sequentially querying the full range of data such as waybill details, waybill distribution records, salesperson operation records, platform work order records, actual receipt details, and AI intelligent outbound call interaction records under the corresponding dimensions, so as to completely restore the entire business link that caused the anomaly; During the traversal, data cross-validation and logical reasoning are performed synchronously. Combined with indicator calculation formulas and business rules, abnormal trigger points are located. For example, when the on-site delivery rate is abnormal, the signature field and signature type label are automatically screened to determine whether the salesperson failed to deliver the goods as required or the customer actively chose to receive them on behalf of the customer. When the false work order rate is abnormal, the work order source, work order type, and waybill signing status are matched in reverse to distinguish whether it is a real customer complaint or a system misjudgment.

[0033] In this embodiment, various performance indicators can be identified, enabling full-chain traceability of waybills, personnel, work orders, and receipt processes. Furthermore, dynamic weighted fusion algorithms are introduced, which, combined with business weight configuration rules, allocate weights to improve the accuracy of anomaly identification and traceability.

[0034] A third embodiment of a method for processing express delivery fulfillment quality monitoring data in this invention includes: 110. Access multi-source heterogeneous business data; then perform invalid data filtering (full data cleaning, data format standardization processing, and data association); First, the system connects to multi-source heterogeneous business data in parallel via interfaces, pulling various types of raw business data, such as waybill scan tables, work order detail tables, AI intelligent outbound call record tables, branch file tables, and salesperson information tables. Then, it performs full data retrieval and temporary caching operations. Next, invalid data filtering is performed according to established business rules: invalid data includes returned item data; then, data format standardization processing is carried out to unify the time format, numerical format, and encoding rules of all multi-source heterogeneous business data; in the data association stage, the waybill number, branch code, and salesperson code are used as globally unique association primary keys, and multi-table lookup and primary key mapping algorithms are adopted to connect the entire chain of data of waybills, distribution, receipt, work orders, and AI outbound calls, and stitch together the isolated data scattered on different platforms into complete single waybill full lifecycle data.

[0035] 120. Extract quaternary features and construct a multidimensional feature matrix; The four-dimensional features include spatial dimension features, performance indicator dimension features, time series dimension features, and business tag dimension features. Feature extraction, encoding, and quantification are performed in layers according to the above four categories. Spatial Dimension Features: Traversing six spatial levels—region, province, city, primary outlet, signing outlet, and salesperson—a one-hot encoding algorithm is used to digitize different regions, outlets, and personnel, transforming textual geographic and subject information into discrete feature vectors that the algorithm can recognize, and recording the physical ownership of each data point. Performance indicator dimension characteristics: It includes several performance indicators. This embodiment focuses on 29 core performance indicators defined by the business, including signed receipts, door-to-door visits, warehousing, various percentages, work order statistics, month-on-month values, etc., and completes indicator value extraction and dimensionless normalization processing. Time-series features: Based on continuous time-series data from the past year, a timestamp sequence is generated with the natural day as the smallest time unit. Derived time-series features such as daily indicator values, weekly averages, monthly averages, historical values ​​for the same period, and time-series fluctuation differences are extracted to construct a continuous time-series feature chain, depicting the trend of performance indicators changing over time. Business tag dimension features: Extract discrete classification tags such as receipt type, work order type, whether delivery to the door was canceled, AI outbound call connection status, and outbound call intent. In some embodiments, a tag quantization algorithm is used to convert the above classification tags into feature values.

[0036] 130. Dynamically generate early warning thresholds for each performance indicator based on the regional order volume scale, compare the real-time data of the performance indicator with the early warning thresholds, mark abnormal indicators by combining time-series and spatial features, and trigger early warnings. In this embodiment, historical characteristic data (e.g., one year's full historical characteristic data) of each performance indicator within a preset time period is first retrieved. An exponential moving average algorithm combined with a quantile statistical algorithm is used for joint modeling. The long-term historical mean of each performance indicator is calculated using the exponential moving average algorithm, thus mitigating the impact of occasional extreme values. Then, the normal fluctuation range of the indicators is defined based on the 95th percentile. In some embodiments, a regional coefficient parameter is also introduced, and the coefficient value is dynamically adjusted according to the average daily order volume and business volume of different provinces, cities and outlets. Large outlets with high average daily order volume are set with stricter threshold coefficients, while small outlets with low average daily order volume are appropriately relaxed in terms of threshold standards. In the end, exclusive adaptive early warning thresholds are generated for each performance indicator and each spatial level, completely replacing the traditional fixed threshold mode. Then, the real-time data of the performance indicators are compared with the early warning thresholds, and abnormal indicators are marked by combining time-series and spatial features, including: The measured values ​​of the indicators in the current period are compared with the warning thresholds of the corresponding levels and indicators line by line and indicator by indicator. The difference is used to determine whether the indicators exceed the normal range and to initially mark suspected abnormal data. Using the current day as the end of the window, the time series data of the indicator is continuously extracted for 7 consecutive days. The fluctuation amplitude, fluctuation frequency, and offset duration of the data within the window are calculated. At the same time, a Markov chain state transition model is introduced to divide the data state into three discrete states: normal state, slightly abnormal state, and severely abnormal state. The evolution pattern of the data state within the continuous window is analyzed through the state transition probability matrix. If only a single window shows data offset, it is judged as normal short-term fluctuation or occasional anomaly; if multiple sliding windows remain in an abnormal state and the state transition probability approaches 1, it is judged as a continuous operational failure; if multiple outlets in the same province / city show abnormal indicators simultaneously, it is judged as a regional batch problem.

[0037] 140. For abnormal indicators, perform a reverse depth-first search based on the entity-relationship graph, prioritize traversing the core related entities with higher weights, and sequentially query the business tag dimension features under the corresponding dimension to restore the business link that caused the anomaly and complete the full-link traceability of waybill, personnel, work order and receipt. Specifically, after identifying abnormal indicators and triggering an alert, the pre-built logistics fulfillment knowledge graph is invoked. Its entity nodes include spatial dimension features and business tag dimension features (region, province, city, outlet, salesperson, waybill, work order, receipt record, AI outbound call record, etc., covering all business entities). Its relational edges define the logical relationships between entity nodes (distribution, receipt, order dispatch, customer complaint, outbound call, etc., and configure the association weight and association direction for each relational edge). Using the marked abnormal indicators and abnormal spatial dimensions as the starting nodes for tracing the source, the reverse depth-first search algorithm is used to perform a full-link traversal: starting from the starting node, the algorithm recursively searches downwards along the reverse association edges of the graph, prioritizing the traversal of core association entities with higher weights, and sequentially querying the full range of data such as waybill details, waybill distribution records, salesperson operation records, platform work order records, actual receipt details, and AI intelligent outbound call interaction records under the corresponding dimensions, so as to completely restore the entire business link that caused the anomaly; During the traversal, data cross-validation and logical reasoning are performed simultaneously, and the abnormal trigger points are located by combining the indicator calculation formula and business rules.

[0038] In this embodiment, various performance indicators can be identified, enabling full-chain traceability of waybills, personnel, work orders, and receipt processes. Furthermore, early warning thresholds for each performance indicator can be dynamically generated based on the regional order volume, replacing the traditional fixed threshold mode. This approach greatly improves the flexibility and adaptability of anomaly identification.

[0039] The above describes a method for processing express delivery quality monitoring data in an embodiment of the present invention. The following describes a device for processing express delivery quality monitoring data in an embodiment of the present invention. Please refer to [link / reference]. Figure 2 One embodiment of the express delivery fulfillment quality monitoring data processing device of the present invention includes: Data processing module 210 is used to access multi-source heterogeneous business data, perform invalid data filtering, data format standardization processing, and associate multi-source heterogeneous business data scattered across different platforms. Invalid data includes returned item data. Data format standardization processing includes unifying the time format, numerical format, and encoding rules of multi-source heterogeneous business data. Data association uses waybill number, branch code, and salesperson code as globally unique association keys.

[0040] The extraction module 220 is used to extract multi-dimensional features, which include spatial dimension features, performance indicator dimension features, time series dimension features, and business tag dimension features. The spatial dimension features record the physical ownership of each data item. The performance indicator dimension features include several performance indicators. The time series dimension features depict the trend of performance indicators changing over time. The business tag dimension features convert several discrete tags into feature values. Discrete tags include receipt type, work order type, and whether delivery to the door has been cancelled. The extraction module 220 also includes a dynamic weighted fusion unit 221, which is used to assign high weights to core risk indicators among performance indicators and / or other characteristics, and assign basic weights to other indicators.

[0041] The early warning module 230 is used to obtain the early warning threshold of each performance indicator, compare the real-time data of the performance indicator with the early warning threshold, mark abnormal indicators by combining time-series and spatial features, and trigger an early warning. For example, the early warning module 230 uses the current day as the end of the window and continuously captures the time series data of indicators for 7 consecutive days, calculating the fluctuation amplitude, fluctuation frequency, and offset duration of the data within the window; at the same time, it introduces a Markov chain state transition model to divide the data state into several discrete states, and analyzes the evolution law of the data state within the continuous window through the state transition probability matrix: if only a single window shows a data offset, it is judged as a normal short-term fluctuation or an occasional anomaly; if multiple consecutive sliding windows maintain an abnormal state and the state transition probability approaches 1, it is judged as a continuous operational failure; if multiple outlets in the same province / city simultaneously show abnormal indicators, it is judged as a regional batch problem. In some embodiments, the early warning module 230 includes a dynamic threshold generation unit 231. The dynamic threshold generation unit 231 retrieves historical characteristic data of each performance indicator within a preset time period, uses an exponential moving average algorithm combined with a quantile statistical algorithm to jointly model, calculates the long-term historical average of each performance indicator, delineates the normal fluctuation range of the indicator, and generates an early warning threshold.

[0042] The search and restoration module 240 is used to perform a reverse depth-first search. The starting node for tracing the source includes the marked abnormal indicators. It prioritizes traversing the core related entities with higher weights, and sequentially queries the business tag dimension features under the corresponding dimension to restore the business chain that caused the abnormality. After determining the abnormal indicators, the search and restoration module 240 calls the pre-built logistics fulfillment knowledge graph. Its entity nodes contain spatial dimension features and business tag dimension features. Its relational edges define the logical relationships between entity nodes, and configure the relational weight and relational direction for each relational edge. Taking the marked abnormal indicators and abnormal spatial dimensions as the source tracing starting nodes, the reverse depth-first search algorithm is used to perform a full-link traversal. During the traversal, data cross-validation and logical reasoning can be performed simultaneously, and the abnormal trigger point can be located by combining the indicator calculation formula and business rules.

[0043] In this embodiment, the data processing module 210 accesses multi-source heterogeneous business data, the extraction module 220 extracts four-element features of spatial dimension, performance indicator dimension, time sequence dimension, and business tag dimension, the early warning module 230 identifies each performance indicator and compares and issues early warnings, and the search and restoration module 240 performs full-link traceability of waybills, personnel, work orders, and signing links, thereby improving the intelligent anomaly identification and traceability capabilities.

[0044] above Figure 2 This invention provides a detailed description of a data processing device for monitoring express delivery fulfillment quality from the perspective of modular functional entities. The following describes the electronic equipment in this invention from the perspective of hardware processing.

[0045] Figure 3 This is a schematic diagram of the structure of an electronic device 300 provided in an embodiment of the present invention. The electronic device 300 can vary significantly due to different configurations or performance characteristics. It may include one or more central processing units (CPUs) 310 (e.g., one or more processors) and a memory 320, and one or more storage media 330 (e.g., one or more mass storage devices) for storing application programs 333 or data 332. The memory 320 and storage media 330 can be temporary or persistent storage. The program stored in the storage media 330 may include one or more modules (not shown in the diagram), each module including a series of instruction operations on the electronic device 300. Furthermore, the processor 310 may be configured to communicate with the storage media 330 and execute the series of instruction operations in the storage media 330 on the electronic device 300.

[0046] Electronic device 300 may also include one or more power supplies 340, one or more wired or wireless network interfaces 350, one or more input / output interfaces 350, and / or one or more operating systems 331, such as Windows Server, Mac OS X, Unix, Linux, FreeBSD, etc. Those skilled in the art will understand that... Figure 3 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.

[0047] 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. The computer-readable storage medium stores instructions that, when executed on a computer, cause the computer to perform the steps of a method for processing express delivery fulfillment quality monitoring data.

[0048] 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.

[0049] 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.

[0050] 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 method for processing express delivery fulfillment quality monitoring data, characterized in that, include: Access to multi-source heterogeneous business data; Extract multi-dimensional features, including spatial dimension features, performance indicator dimension features, time series dimension features, and business tag dimension features. Spatial dimension features record the physical ownership of each data item. Performance indicator dimension features include several performance indicators. Time series dimension features depict the trend of performance indicators changing over time. Business tag dimension features include receipt type, work order type, and whether door-to-door delivery has been cancelled. The early warning thresholds for each performance indicator are obtained. The real-time data of the performance indicator is compared with the early warning thresholds. Abnormal indicators are marked by combining time-series and spatial features, and early warnings are triggered. Perform a reverse depth-first search, starting with the marked abnormal indicators, prioritize traversing the core related entities with higher weights, and sequentially query the business tag dimension features under the corresponding dimensions to reconstruct the business chain that caused the anomaly.

2. The method for processing express delivery fulfillment quality monitoring data according to claim 1, characterized in that, The access to multi-source heterogeneous business data includes: performing filtering of invalid data and data format standardization processing. Invalid data includes returned data. Data format standardization processing includes unifying the time format, numerical format, and encoding rules of multi-source heterogeneous business data.

3. A method for processing express delivery fulfillment quality monitoring data according to claim 1 or 2, characterized in that, The access to multi-source heterogeneous business data includes: associating multi-source heterogeneous business data scattered across different platforms, with the waybill number, branch code, and salesperson code serving as globally unique association keys.

4. The method for processing express delivery fulfillment quality monitoring data according to claim 1, characterized in that, The extraction of multi-dimensional features includes: introducing a dynamic weighted fusion algorithm, combining business weight configuration rules, assigning high weights to core risk indicators in the performance indicators, and assigning basic weights to other indicators.

5. The method for processing express delivery fulfillment quality monitoring data according to claim 1, characterized in that, The obtained early warning thresholds for each performance indicator include: Historical characteristic data of each performance indicator within a preset time period are retrieved, and the long-term historical average of each performance indicator is calculated by using a combination of exponential moving average algorithm and quantile statistical algorithm to define the normal fluctuation range of the indicator and generate early warning thresholds.

6. A method for processing express delivery fulfillment quality monitoring data according to claim 1 or 5, characterized in that, The process of comparing real-time data of performance indicators with early warning thresholds and marking abnormal indicators by combining time-series and spatial features includes: Using the current day as the end of the window, the time series data of the indicator is continuously extracted for 7 consecutive days. The fluctuation amplitude, fluctuation frequency, and offset duration of the data within the window are calculated. At the same time, a Markov chain state transition model is introduced to divide the data state into several discrete states, and the evolution pattern of the data state within the continuous window is analyzed through the state transition probability matrix. If only a single window shows data offset, it is judged as normal short-term fluctuation or occasional anomaly; if multiple sliding windows remain in an abnormal state and the state transition probability approaches 1, it is judged as a continuous operational failure; if multiple outlets in the same province / city show abnormal indicators simultaneously, it is judged as a regional batch problem.

7. The method for processing express delivery fulfillment quality monitoring data according to claim 1, characterized in that, The reverse depth-first search includes: after determining the abnormal indicators, calling the pre-built logistics fulfillment knowledge graph, whose entity nodes contain spatial dimension features and business tag dimension features, whose relationship edges define the logical relationship between entity nodes, and configuring the relationship weight and relationship direction for each relationship edge; Using the marked abnormal indicators and abnormal spatial dimensions as the starting nodes for tracing, a reverse depth-first search algorithm is used to perform a full-link traversal. During the traversal, data cross-validation and logical reasoning are performed simultaneously, and the abnormal trigger point is located by combining the indicator calculation formula and business rules.

8. A data processing device for monitoring express delivery fulfillment quality, characterized in that, include: The data processing module is used to access multi-source heterogeneous business data; The extraction module is used to extract multi-dimensional features, including spatial dimension features, performance indicator dimension features, time series dimension features, and business tag dimension features. Spatial dimension features record the physical ownership of each data item. Performance indicator dimension features include several performance indicators. Time series dimension features depict the trend of performance indicators changing over time. Business tag dimension features convert several discrete tags into feature values. Discrete tags include receipt type, work order type, and whether door-to-door delivery has been cancelled. The early warning module is used to obtain the early warning thresholds for each performance indicator, compare the real-time data of the performance indicator with the early warning thresholds, mark abnormal indicators by combining time-series and spatial features, and trigger early warnings. The search and restoration module is used to perform a reverse depth-first search. The starting node for tracing the source includes the marked abnormal indicators. It prioritizes traversing the core related entities with higher weights, and sequentially queries the business tag dimension features under the corresponding dimensions to restore the business chain that caused the anomaly.

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 express delivery fulfillment quality monitoring data processing method as described in any one of claims 1-7.

10. A computer-readable storage medium storing instructions thereon, characterized in that, When the instructions are executed by the processor, they implement the various steps of the express delivery performance quality monitoring data processing method as described in any one of claims 1-7.

Images