Warehouse logistics data processing method

Abstract

The invention discloses a warehouse logistics data processing method, and relates to the technical field of warehouse logistics, and the main points of the technical scheme comprise the following steps: collecting the return reverse circulation data, stock overstock early warning data and traceability feedback data of warehouse goods in real time, and obtaining the initial detection reverse logistics data; the quality inspection rechecking time consumption and the refund flow connection duration of the stored goods in the reverse circulation process are collected in real time to obtain preprocessed reverse processing data; according to the initial measurement reverse logistics data, obtaining initial measurement reverse processing efficiency; obtaining current reverse constraint feature data according to the return reason reasonability difference value and the reverse sorting adaptation degree difference value of the stored goods, and obtaining actual reverse processing efficiency for the initial measurement reverse processing efficiency according to the first prediction reverse error value and the second prediction reverse error value; and outputting a processing result according to the actual reverse processing efficiency. The method has the effect of comprehensively improving the management level and the operation efficiency of storage reverse logistics.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of warehouse logistics, more specifically, it relates to a warehouse logistics data processing method. BACKGROUND

[0002] The traditional reverse logistics data processing mode has obvious short boards. Most warehouses only collect scattered basic data such as return quantity and inventory quantity, lack real-time collection of return reverse flow path, inventory accumulation cycle, and goods traceability link dimension, resulting in serious data fragmentation and inability to fully reflect the whole process state of reverse logistics; the calculation of reverse processing efficiency often only depends on the simple ratio of single goods flow volume and time consumption, without considering the reasonableness of return reason, the adaptability of sorting process, and the interference of process time fluctuation such as quality inspection review and refund connection on efficiency, thereby deviating from the actual operation situation. SUMMARY

[0003] In view of the deficiencies in the prior art, the purpose of the present application is to provide a warehouse logistics data processing method.

[0004] To achieve the above-mentioned purpose, the present application provides the following technical scheme: The warehouse logistics data processing method comprises the following steps: Real-time collection of return reverse flow data, inventory accumulation early warning data and traceability feedback data of warehouse goods to obtain initial measured reverse logistics data; real-time collection of quality inspection review time consumption and refund process connection time of warehouse goods in the reverse flow process to obtain preprocessed reverse processing data; According to the initial measured reverse logistics data, the initial measured reverse processing efficiency is obtained; According to the difference value of the reasonableness of the return reason of the warehouse goods and the difference value of the reverse sorting adaptability, the current reverse constraint characteristic data is obtained, and according to the current reverse constraint characteristic data and the correlation influence variation characteristic between the historical return reason reasonableness difference value, the historical reverse sorting adaptability difference value and the historical upper chain time delay error, the first predicted reverse error value is obtained; According to the proportion of each type of time consumption in the preprocessed reverse processing data, the current error influence reverse degree is obtained, and according to the correlation influence variation characteristic between the current error influence reverse degree and the historical error influence reverse degree and the historical upper chain time delay error, the second predicted reverse error value is obtained; According to the first predicted reverse error value and the second predicted reverse error value, the actual reverse processing efficiency of the initial measured reverse processing efficiency is obtained; According to the actual reverse processing efficiency, the processing result is output.

[0005] Preferably, the initial measured reverse processing efficiency is obtained according to the initial measured reverse logistics data, and specifically comprises the following steps: Extract the goods reverse flow node information corresponding to the return reverse flow data in the initial reverse logistics data, extract the backlog goods category information corresponding to the inventory backlog warning data in the initial reverse logistics data, and extract the complete goods traceability link information corresponding to the traceability feedback data in the initial reverse logistics data; Based on the information of the reverse flow nodes, determine the cargo dwell information of each reverse flow node, and distinguish the differences in the dwell information of different categories of goods at each reverse flow node according to the cargo dwell information and the information of the backlog of goods categories; Based on the complete information of the cargo traceability link, determine the traceability matching status of each category of goods in reverse flow, and obtain the node adaptation information of each category of goods in reverse flow based on the traceability matching status and the difference in the dwell time of different categories of goods at each reverse flow node. Based on node adaptation information, the average turnover time for different categories of goods to complete the entire reverse flow process is obtained. The average turnover time is then corrected based on the backlog relief information corresponding to the inventory backlog warning data to obtain the corrected average turnover time. The initial reverse processing efficiency is obtained based on the corrected average turnover time and the volume of goods reversed flow corresponding to the returned goods reverse flow data.

[0006] Preferably, the reverse flow data of returns includes return type, return path nodes, and return liability determination results; The inventory backlog warning data includes the amount of returned goods backlog and the inventory turnover cycle.

[0007] Preferably, the current reverse constraint feature data is obtained based on the difference in the reasonableness of the reasons for the return of stored goods and the difference in the adaptability of reverse sorting, specifically including the following steps: Analyze the return reason correlation information corresponding to the reasonableness difference of the return reason, distinguish the correlation type between the return reason and the attributes of the goods themselves based on the return reason correlation information, and extract the difference fluctuation range information corresponding to the reasonableness difference of the return reason; The correctness of return reasons is determined based on the association type, and the strength of reasonableness constraints corresponding to each association type is determined based on the correctness and difference fluctuation range information. Analyze the sorting process association information corresponding to the reverse sorting fit difference, identify the matching gap between the sorting link and the reverse flow of goods based on the sorting process association information, and extract the change trend information of the difference corresponding to the reverse sorting fit difference. Based on the matching gap, the impact level of the gap is divided, and the fitness constraint strength corresponding to each impact level of the gap is determined according to the difference change trend information. The current reverse constraint feature data is obtained by processing the strength of the correlation rationality constraint, the strength of the fit constraint, and the level of the gap influence.

[0008] Preferably, the current reverse constraint feature data is obtained by processing the strength of the correlation rationality constraint, the strength of the fit constraint, and the level of the gap influence, specifically including the following steps: After associating the rationality constraint strength and the fit constraint strength, the constraint items that affect the reverse flow process are screened out; The current reverse constraint feature data is formed by combining the association weights of the association types and gap impact levels corresponding to each constraint item.

[0009] Preferably, based on the current reverse constraint feature data and the correlation characteristics between the historical return reason reasonableness difference, the historical reverse sorting fit difference, and the historical chain delay error, the first predicted reverse error value is obtained, specifically including the following steps: After classifying the differences in the rationality of historical return reasons, the differences in historical reverse sorting adaptability, and the historical chain delay error based on their common origin, a combination of historical features is formed. For each set of historical feature combinations, the correlation between the fluctuation range of the reasonableness difference of historical return reasons, the change gradient of the historical reverse sorting adaptability difference, and the change direction and amplitude of the corresponding historical chain delay error is determined, thus obtaining the historical correlation influence feature set. The current reverse constraint feature data is decomposed into the fluctuation feature of the current return reason reasonableness difference and the change feature of the current reverse sorting suitability difference. The current return reason reasonableness difference fluctuation feature and the current reverse sorting suitability difference change feature are compared with the corresponding historical features in the historical correlation influence feature set to determine the target historical correlation influence feature set that matches the current feature trend. The deviation between the current return reason reasonableness difference fluctuation feature and the corresponding historical fluctuation feature in the target historical feature set, as well as the deviation between the current reverse sorting suitability difference change feature and the corresponding historical change feature in the target historical feature set, are calculated to form the feature difference quantification result. Based on the correlation pattern between historical features and historical chain delay error 1 in the target historical correlation influence feature set, and combined with the feature difference quantification results, the deviation correction of historical chain delay error 1 corresponding to the target historical features is performed; the initial prediction reverse error value is generated by comprehensively considering the impact of the corrected historical chain delay error 1 and the feature difference quantification results on the current reverse constraint feature data. The first prediction inverse error value is obtained by calibrating the initial prediction inverse error value using the fit of historical feature combinations.

[0010] Preferably, the degree of reverse error impact is obtained based on the proportion of various time consumptions in the preprocessed reverse processing data, specifically including the following steps: If the proportion of quality inspection review time in the preprocessed reverse processing data is greater than the time threshold and the proportion of refund process connection time is greater than the time proportion threshold, then the preprocessed reverse processing data is determined to be negatively correlated feature data affecting reverse processing efficiency, and the first error affects the value-added reverse degree is output. If the proportion of quality inspection review time in the preprocessed reverse processing data is less than or equal to the time threshold and / or the proportion of refund process connection time is less than or equal to the time proportion threshold, then the preprocessed reverse processing data is determined to be positively correlated feature data of reverse processing efficiency, and the second error impact reduction reverse degree is output. The first error affects the inverse degree of added value, and the second error affects the inverse degree of subtraction. These two factors combine to form the inverse degree of the current error.

[0011] Preferably, the second predicted reverse error value is obtained based on the correlation characteristics between the current error affecting the reverse degree and the historical error affecting the reverse degree and the historical chain delay error, specifically including the following steps: The alignment result is obtained by performing time-series alignment processing on the historical error impact inverse degree and the historical chain delay error using the same time dimension division standard; Based on the alignment results, the feature changes corresponding to the two time-series features of historical error impact on reverse degree and historical chain delay error are filtered to obtain the association mapping relationship set; Hierarchical filtering retains the set of association mapping relationships that meet the requirements to form a valid set of association mapping relationships; Extract the current temporal features that affect the inverse degree of the current error and calculate the feature change of each temporal node. Match the feature change of each temporal node with the historical error increase / decrease impact feature change of the effective association mapping relationship set to determine the target historical association mapping relationship. Based on the historical correlation mapping relationship of the target, the change of the two features of the corresponding historical chain delay error is used as the reference value of the delay error change. The delay error change reference value is fused in time series to obtain the delay error change trend. The second predicted reverse error value is obtained based on the delay error change trend and the overall change of the current error affecting the reverse degree.

[0012] Preferably, the actual reverse processing efficiency is obtained by adjusting the initial reverse processing efficiency based on the first predicted reverse error value and the second predicted reverse error value, specifically including the following steps: Extract the error impact dimension features corresponding to the first and second predicted inverse error values, and clarify the constraint class impact dimension corresponding to the first predicted inverse error value and the process time class impact dimension corresponding to the second predicted inverse error value based on the error impact dimension features; Based on the weight differences between the constraint-type influence dimension and the process time-type influence dimension, the first and second predicted reverse error values ​​are assigned hierarchically to obtain the weighted first and second predicted reverse error values. The weighted first and second predicted reverse error values ​​are correlated and verified with the initial reverse processing efficiency, and error correction directions that are compatible with the initial reverse processing efficiency are selected. Based on the error correction direction, the first predicted reverse error value and the second predicted reverse error value are applied sequentially to the initial reverse processing efficiency for correction, and then the actual reverse processing efficiency is obtained.

[0013] Preferably, the processing result is output based on the actual reverse processing efficiency, specifically including the following steps: Determine the output direction by matching the corresponding optimization requirements based on the actual reverse processing efficiency level; Based on the output direction, the constraint-type impact information and process time-type impact information related to the actual reverse processing efficiency are integrated to generate basic output information including efficiency level, influencing factors and optimization direction. The basic output information is hierarchically organized, and key contents are sorted according to optimization priority to form a structured processing result; The structured processing results are adapted to the corresponding output format based on different application scenarios of warehouse logistics management to obtain the final processing result.

[0014] Compared with the prior art, the present invention has the following beneficial effects: This invention obtains the initial reverse processing efficiency based on preliminary reverse logistics data, and corrects it by incorporating constraint-related and process-time-related errors. It considers the impact of return reasons and sorting adaptation constraints, as well as the interference of time-consuming quality inspection and refund processes, resulting in an actual reverse processing efficiency that more closely reflects real-world operational conditions. The first predicted reverse error value corresponds to constraint-related issues, and the second predicted reverse error value corresponds to process-time-related issues. Combining these two factors clarifies whether the efficiency loss is due to unreasonable return reasons or a lengthy quality inspection process, providing a clear direction for optimization. For example, if error analysis reveals that excessively high quality inspection review time is the main influencing factor, the focus can be on optimizing the quality inspection process to improve processing efficiency. This method can output structured processing results adapted to different scenarios, enhancing the practicality and implementability of the solution. Optimization needs are matched according to the actual efficiency level, and influencing factors and optimization directions are prioritized to adapt to different output formats for warehousing operations, management decisions, and other scenarios. This promotes the refinement and efficiency of reverse logistics management. This method provides clear guidance for optimization decisions, comprehensively improving the management level and operational efficiency of warehousing reverse logistics. Attached Figure Description

[0015] Fig. 1 A schematic diagram illustrating the steps of a warehousing and logistics data processing method according to an embodiment of the present invention; Fig. 2 This is a schematic diagram illustrating the reverse degree of the current error in the warehousing and logistics data processing method provided in this embodiment of the invention. Detailed Implementation

[0016] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0017] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of the invention. Therefore, the invention is not limited to the specific embodiments disclosed below.

[0018] Secondly, the term "an embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in different places throughout this specification does not necessarily refer to the same embodiment, nor is it a single embodiment or an embodiment selectively excluded from other embodiments.

[0019] Reference Figs. 1-2 As shown.

[0020] The embodiments further illustrate the warehousing and logistics data processing method proposed in this invention.

[0021] A method for processing warehousing and logistics data, comprising the following steps: Real-time collection of data on returned goods, reverse flow, inventory backlog warnings, and traceability feedback in the warehouse yields preliminary reverse logistics data; real-time collection of data on the time spent on quality inspection and verification of goods during the reverse flow process, as well as the time required for the refund process, yields pre-processed reverse processing data. The data on the reverse flow of returns includes the type of return, the nodes in the return path, and the results of the determination of responsibility for the return. Inventory backlog warning data includes the amount of returned goods backlog and the inventory turnover cycle.

[0022] Real-time collection of return reverse flow data, inventory backlog warning data, and traceability feedback data is used to integrate these data into preliminary reverse logistics data. Return reverse flow data records key information throughout the entire return process, such as distinguishing between return types (quality issues, incorrect specifications, and no-reason returns), return path nodes (recording each stage of the return process from the consumer to the warehouse, from the consumer's drop-off point to the regional transit warehouse, and then to the target warehousing center), and return liability determination results (clearly defining responsibility for returns, such as whether the return was due to supplier product quality issues or warehouse sorting errors). Inventory backlog warning data reflects the accumulation of returns in the warehouse; the backlog amount refers to the number of unprocessed returned goods in the current warehouse, and the backlog turnover cycle refers to the average time required for these backlogged returns to be processed from entering the warehouse.

[0023] The system collects real-time data on the time spent on quality inspection and verification during the reverse flow process, as well as the time required for the refund process to connect. This data is then integrated to form pre-processed reverse processing data. Quality inspection and verification time refers to the time spent on quality testing and information verification of returned goods, such as the time spent on functional testing and packaging integrity verification of returned electronic products. Refund process connection time refers to the interval between completing the return quality inspection and initiating the refund process, such as the time it takes to transmit information to the financial system after quality inspection and for the financial system to process the refund application.

[0024] The initial reverse processing efficiency was obtained based on the initial reverse logistics data. Based on the difference in the reasonableness of the reasons for the return of stored goods and the difference in the adaptability of reverse sorting, the current reverse constraint feature data is obtained. Based on the current reverse constraint feature data and the correlation and variation characteristics between the historical difference in the reasonableness of the reasons for the return, the historical difference in the adaptability of reverse sorting and the historical chain delay error, the first predicted reverse error value is obtained. The current error impact degree is obtained based on the proportion of various time consumptions in the preprocessed reverse processing data. The second predicted reverse error value is obtained based on the correlation and variation characteristics between the current error impact degree, the historical error impact degree, and the historical chain delay error. The actual reverse processing efficiency is obtained by analyzing the initial reverse processing efficiency based on the first predicted reverse error value and the second predicted reverse error value. Output the processing results based on the actual reverse processing efficiency.

[0025] The initial reverse logistics efficiency is determined based on the initial reverse processing data, specifically including the following steps: Extract the goods reverse flow node information corresponding to the return reverse flow data in the initial reverse logistics data, extract the backlog goods category information corresponding to the inventory backlog warning data in the initial reverse logistics data, and extract the complete goods traceability link information corresponding to the traceability feedback data in the initial reverse logistics data; Based on the information of the reverse flow nodes, determine the cargo dwell information of each reverse flow node, and distinguish the differences in the dwell information of different categories of goods at each reverse flow node according to the cargo dwell information and the information of the backlog of goods categories; Based on the complete information of the cargo traceability link, determine the traceability matching status of each category of goods in reverse flow, and obtain the node adaptation information of each category of goods in reverse flow based on the traceability matching status and the difference in the dwell time of different categories of goods at each reverse flow node. Based on node adaptation information, the average turnover time for different categories of goods to complete the entire reverse flow process is obtained. The average turnover time is then corrected based on the backlog relief information corresponding to the inventory backlog warning data to obtain the corrected average turnover time. The initial reverse processing efficiency is obtained based on the corrected average turnover time and the volume of goods reversed flow corresponding to the returned goods reverse flow data.

[0026] The reverse flow data of returned goods is used to extract information on the nodes of the reverse flow of goods. This information records each step of the return process, such as the nodes from the consumer's drop-off point to the regional distribution warehouse and then to the warehouse quality inspection point for clothing returns. The inventory backlog warning data is used to extract information on the category of backlogged goods, which means clarifying what category the current backlogged returns belong to, such as electronic devices or daily necessities. The traceability feedback data is used to extract information on the complete traceability chain of goods. This information refers to whether the entire chain record of goods from production to sales and then to return is complete. For example, can a food return be traced back to the production batch and logistics transportation batch?

[0027] Based on the information of the reverse flow nodes, the dwell time of goods at each reverse flow node is determined, such as a dwell time of 24 hours at a regional distribution warehouse and a dwell time of 12 hours at a warehouse quality inspection point. Combined with the information of the categories of backlogged goods, the differences in dwell time of different categories of goods at each node are distinguished. For example, electronic equipment typically dwells longer at the quality inspection node than daily necessities. This allows us to obtain information such as the difference in dwell time between electronic equipment and daily necessities at each node.

[0028] The traceability matching status of each category of goods in reverse flow is determined based on the completeness of the goods traceability chain information. For example, if the traceability chain of a batch of returned clothing is complete and can be matched with the production batch, its traceability matching status is complete; while if the traceability chain of a batch of returned toys lacks production batch information, its traceability matching status is incomplete. The node adaptation information for the reverse flow of each category of goods is obtained by combining the differences in the dwell time of different categories of goods at each node. For example, clothing goods with complete traceability matching status have a more reasonable dwell time at each node, resulting in higher node adaptation; while toys with incomplete traceability matching status have a longer dwell time at the distribution node, resulting in lower node adaptation.

[0029] The average turnover time for different categories of goods to complete the reverse flow process is calculated based on node adaptation information. For example, the average turnover time for clothing is 48 hours, and for toys it is 72 hours. The average turnover time is then corrected by incorporating backlog relief correlation information corresponding to inventory backlog warning data. This backlog relief correlation information is a coefficient representing the impact of the current backlog on processing efficiency. For instance, when the backlog exceeds a threshold, processing efficiency decreases, and the correction coefficient is set to 1.2. Corrected average turnover time = Average turnover time × Correction coefficient corresponding to backlog relief correlation information. If the average turnover time is 48 hours and the correction coefficient is 1.2, then the corrected average turnover time is 48 × 1.2 = 57.6 hours.

[0030] The initial reverse processing efficiency is calculated by combining the adjusted average turnover time and the corresponding reverse turnover volume of returned goods. Initial reverse processing efficiency = reverse turnover volume ÷ adjusted average turnover time. For example, if the reverse turnover volume of clothing goods is 200 pieces in a certain period, and the adjusted average turnover time is 57.6 hours, then the initial reverse processing efficiency is 200 ÷ 57.6 ≈ 3.47 pieces / hour.

[0031] This allows us to gradually extract information from the initial reverse logistics data, ultimately obtaining the initial reverse processing efficiency, which reflects the reverse logistics processing capability.

[0032] Based on the difference in the reasonableness of returned goods and the difference in reverse sorting suitability, the current reverse constraint feature data is obtained, which includes the following steps: Analyze the return reason correlation information corresponding to the reasonableness difference of the return reason, distinguish the correlation type between the return reason and the attributes of the goods themselves based on the return reason correlation information, and extract the difference fluctuation range information corresponding to the reasonableness difference of the return reason; The correctness of return reasons is determined based on the association type, and the strength of reasonableness constraints corresponding to each association type is determined based on the correctness and difference fluctuation range information. Analyze the sorting process association information corresponding to the reverse sorting fit difference, identify the matching gap between the sorting link and the reverse flow of goods based on the sorting process association information, and extract the change trend information of the difference corresponding to the reverse sorting fit difference. Based on the matching gap, the impact level of the gap is divided, and the fitness constraint strength corresponding to each impact level of the gap is determined according to the difference change trend information. The current reverse constraint feature data is obtained by processing the strength of the correlation rationality constraint, the strength of the fit constraint, and the level of the gap influence.

[0033] This analysis examines the correlation information between the return reason and the actual situation, specifically the details of how the return reason relates to the actual circumstances. For example, if the return reason for an electronic product is a damaged screen, the corresponding correlation information is the relationship between the screen damage and the product's inherent quality. Based on this information, the type of correlation between the return reason and the product's inherent attributes can be distinguished. Common correlation types include direct correlation, indirect correlation, and no correlation. For instance, a return reason for clothing due to pilling fabric is directly related to the material properties of the product; while a return reason such as the consumer disliking the color is unrelated to the product's inherent attributes. Simultaneously, the analysis extracts the fluctuation range information corresponding to the return reason's reasonableness difference. This fluctuation range refers to the range of change of the reasonableness difference over a period of time; for example, the reasonableness difference for a certain type of goods might fluctuate between 0.2 and 0.5.

[0034] The correctiveness of return reasons under different association types is determined based on the association type. Correctiveness refers to whether the return reason requires adjustment of the reverse processing flow. For example, return reasons with direct association require stricter quality inspection and have higher correctiveness; return reasons without association do not require special adjustment and have lower correctiveness. The reasonableness constraint strength corresponding to each association type is determined by combining the difference fluctuation range information. Reasonableness constraint strength is an indicator that measures the degree to which the return reason constrains the reverse process. Reasonableness constraint strength = correctiveness coefficient × mean difference fluctuation range. The correctiveness coefficient is set according to the association type; for example, the correctiveness coefficient is 0.8 for direct association, 0.5 for indirect association, and 0.2 for no association. Taking the direct association type as an example, if the mean difference fluctuation range is 0.35, then the reasonableness constraint strength = 0.8 × 0.35 = 0.28.

[0035] This analysis examines the correlation information in the sorting process corresponding to the reverse sorting fit difference. This information refers to the relationship between the operational details of the sorting process and the reverse flow requirements. For example, the reverse sorting process for a certain daily necessities involves manual sorting, while the reverse flow requirement is rapid distribution. The corresponding correlation information is the matching status between manual sorting and rapid distribution requirements. Based on this information, the matching gap between the sorting process and the reverse flow requirements is identified. A matching gap refers to the part of the sorting process that cannot meet the flow requirements. For example, the speed of manual sorting cannot meet the timeliness requirements of rapid distribution, which constitutes a matching gap. Simultaneously, the trend information of the difference in reverse sorting fit difference is extracted. For instance, if the sorting fit difference for a certain type of goods is continuously increasing, it indicates that the matching gap is widening.

[0036] The impact level of gaps is categorized based on the matching gaps, according to the degree of impact on the reverse process. For example, a gap can be classified as having a slight, moderate, or severe impact. A 1-hour delay due to insufficient manual sorting speed is considered a moderate impact. The fit constraint strength corresponding to each gap impact level is determined by combining the difference trend information. The fit constraint strength is an indicator that measures the degree to which the sorting process constrains the reverse process. Fit constraint strength = gap impact level coefficient × slope of the difference trend. The gap impact level coefficient is set as 0.3 for slight impact, 0.6 for moderate impact, and 0.9 for severe impact. For example, for a moderate impact level, if the slope of the difference trend is 0.1, then the fit constraint strength = 0.6 × 0.1 = 0.06.

[0037] The strength of the reasonableness constraint, the strength of the fit constraint, and the level of the gap impact are processed. For example, the strength of the reasonableness constraint and the strength of the fit constraint are superimposed according to their weights. Combined with the labeling of the level of the gap impact, the current reverse constraint characteristic data can comprehensively reflect the reverse logistics constraint factors. The current reverse constraint characteristic data can intuitively reflect the degree of constraint of the return reason and the sorting process on the reverse flow.

[0038] The current inverse constraint feature data is obtained by processing the strength of the correlation rationality constraint, the strength of the fit constraint, and the level of the gap influence. The specific steps include: After associating the rationality constraint strength and the fit constraint strength, the constraint items that affect the reverse flow process are screened out; The current reverse constraint feature data is formed by combining the association weights of the association types and gap impact levels corresponding to each constraint item.

[0039] First, the rationality constraint strength and the fit constraint strength need to be correlated and integrated. Then, constraints that have a real impact on the reverse flow process need to be screened out. Constraints refer to factors that affect the efficiency of the reverse flow, such as constraints directly related to the rationality constraint strength or constraints with a medium or higher impact level in the fit constraint strength. The screening is based on a constraint strength threshold. For example, if the constraint strength threshold is set to 0.2, and a rationality constraint strength is 0.28 and a fit constraint strength is 0.06, then the constraint corresponding to 0.28 will be screened out, while the constraint corresponding to 0.06 will be excluded because it is below the threshold. This ensures that the retained constraints have a substantial impact on the reverse flow.

[0040] The current inverse constraint feature data is constructed by combining the association weights of the association types and gap impact levels corresponding to each filtered constraint item. The association weights are numerical values ​​set according to the importance of the association type and gap impact level; for example, the association weight for a direct association type is set to 0.7, and the association weight for a medium gap impact level is set to 0.6. For instance, if the association type corresponding to the rationality constraint item is direct, its association weight is 0.7, and the weighted constraint strength = constraint strength of the constraint item × corresponding association weight. Substituting the data, the weighted constraint strength of this rationality constraint item is 0.28 × 0.7 = 0.196.

[0041] The weighted constraint strengths of all selected constraints are integrated, and the association type and gap impact level information corresponding to each constraint are labeled to form the current reverse constraint feature data. For example, the integrated feature data includes the weighted strength of the rationality constraint (0.196) and the direct association information. This information can comprehensively reflect the strength and attributes of each effective constraint factor in the reverse flow process.

[0042] Based on the current reverse constraint feature data and the correlation characteristics between the historical return reason rationality difference, the historical reverse sorting fit difference, and the historical chain delay error, the first predicted reverse error value is obtained, specifically including the following steps: After classifying the differences in the rationality of historical return reasons, the differences in historical reverse sorting adaptability, and the historical chain delay error based on their common origin, a combination of historical features is formed. For each set of historical feature combinations, the correlation between the fluctuation range of the reasonableness difference of historical return reasons, the change gradient of the historical reverse sorting adaptability difference, and the change direction and amplitude of the corresponding historical chain delay error is determined, thus obtaining the historical correlation influence feature set. The current reverse constraint feature data is decomposed into the fluctuation feature of the current return reason reasonableness difference and the change feature of the current reverse sorting suitability difference. The current return reason reasonableness difference fluctuation feature and the current reverse sorting suitability difference change feature are compared with the corresponding historical features in the historical correlation influence feature set to determine the target historical correlation influence feature set that matches the current feature trend. The deviation between the current return reason reasonableness difference fluctuation feature and the corresponding historical fluctuation feature in the target historical feature set, as well as the deviation between the current reverse sorting suitability difference change feature and the corresponding historical change feature in the target historical feature set, are calculated to form the feature difference quantification result. Based on the correlation pattern between historical features and historical chain delay error 1 in the target historical correlation influence feature set, and combined with the feature difference quantification results, the deviation correction of historical chain delay error 1 corresponding to the target historical features is performed; the initial prediction reverse error value is generated by comprehensively considering the impact of the corrected historical chain delay error 1 and the feature difference quantification results on the current reverse constraint feature data. The first prediction inverse error value is obtained by calibrating the initial prediction inverse error value using the fit of historical feature combinations.

[0043] First, the historical return reason reasonableness difference, historical reverse sorting adaptability difference, and historical chain-end delay error are classified according to their common origin, forming historical feature combinations. Common origin classification means grouping historical return reason reasonableness difference, historical reverse sorting adaptability difference, and historical chain-end delay error corresponding to the same time period and the same product category into one group. For example, the historical return reason reasonableness difference, historical reverse sorting adaptability difference, and historical chain-end delay error corresponding to electronic products in a certain quarter are grouped into one historical feature combination. Such combinations can comprehensively reflect the reverse process characteristics and error relationships under a certain scenario.

[0044] For each set of historical feature combinations, the correlation between the fluctuation range of the historical return reason reasonableness difference, the gradient of the historical reverse sorting fit difference, and the direction and magnitude of the corresponding historical chain delay error is determined, thus obtaining the historical correlation impact feature set. The fluctuation range refers to the difference between the maximum and minimum values ​​of the reasonableness difference within the time period of that combination. For example, in a combination, the reasonableness difference fluctuates from 0.2 to 0.5, with a fluctuation range of 0.3. The gradient refers to the rate of change of the sorting fit difference over time. For example, if the fit difference decreases from 0.4 to 0.1 in 2 days, the gradient is -0.15 / day. The direction of change of the historical chain delay error refers to whether the error increases or decreases, and the magnitude refers to the specific numerical value of the error change. For example, if the error increases from 1 hour to 3 hours, the direction is increase, and the magnitude is 2 hours. When the fluctuation range of the reasonableness difference increases and the gradient of the fit difference is negative, the chain delay error will increase. This type of correlation constitutes the content of the historical correlation impact feature set.

[0045] The current reverse constraint feature data is decomposed into the fluctuation feature of the current return reason reasonableness difference and the change feature of the current reverse sorting fit difference. The fluctuation feature refers to the fluctuation range of the current reasonableness difference, such as the current reasonableness difference fluctuating between 0.3 and 0.6; the change feature refers to the change trend of the current fit difference, such as the current fit difference decreasing at a rate of 0.1 / day. These two types of current features are compared with the corresponding historical features in the historical correlation influence feature set to determine the target historical correlation influence feature set that matches the trend of the current features. For example, if the current fluctuation feature is consistent with the fluctuation amplitude trend of a certain historical combination, and the current change feature is consistent with the change gradient trend of that historical combination, then the feature set corresponding to that historical combination is the target historical correlation influence feature set.

[0046] The deviation between the current feature and the target historical feature is calculated to form the feature difference quantification result. The deviation of the current return reason reasonableness difference fluctuation feature = current fluctuation amplitude - target historical fluctuation amplitude; the deviation of the current reverse sorting fit difference change feature = current change gradient - target historical change gradient. If the current fluctuation amplitude is 0.3 and the target historical fluctuation amplitude is 0.2, then the reasonableness difference fluctuation feature deviation = 0.3 - 0.2 = 0.1; if the current change gradient is -0.1 / day and the target historical change gradient is -0.15 / day, then the fit difference change feature deviation = -0.1 - (-0.15) = 0.05.

[0047] Based on the correlation between historical features and historical chain delay error 1 in the target historical feature set, and combined with the feature difference quantification results, the deviation correction is applied to the historical chain delay error 1 corresponding to the target historical features. Assuming the historical chain delay error 1 in the target historical combination is 2 hours, the correlation is that for every 0.1 increase in fluctuation amplitude, the error increases by 0.5 hours; for every 0.05 increase in change gradient, the error decreases by 0.2 hours. Therefore, the corrected historical chain delay error 1 = target historical chain delay error 1 + (reasonableness difference fluctuation feature deviation × 0.5) + (fitness difference change feature deviation × -0.2). Substituting the data, we get the corrected historical chain delay error 1 = 2 + (0.1 × 0.5) + (0.05 × -0.2) = 2 + 0.05 - 0.01 = 2.04 hours. By combining the historical chain delay error and the quantification results of feature differences, the impact of the corrected historical chain delay error on the current inverse constraint feature data is used to generate an initial prediction inverse error value. For example, by combining the weight of feature differences, the corrected error is adjusted to 2.1 hours, thus obtaining the initial prediction inverse error value.

[0048] The first prediction inverse error value is obtained by calibrating the initial prediction inverse error value using the fit of historical feature combinations. Fit refers to the degree of matching between the current feature and the target historical features. For example, if the fit is 0.9, the first prediction inverse error value = initial prediction inverse error value × fit + initial prediction inverse error value × (1 - fit) × adjustment coefficient (the adjustment coefficient is set to 0.8 based on the actual scenario). Substituting the data, we get the first prediction inverse error value = 2.1 × 0.9 + 2.1 × 0.1 × 0.8 = 1.89 + 0.168 = 2.058 hours.

[0049] The current error impact degree is obtained based on the proportion of time spent on various types of data in the preprocessed and reverse-processed data. This includes the following steps: If the proportion of quality inspection review time in the preprocessed reverse processing data is greater than the time threshold and the proportion of refund process connection time is greater than the time proportion threshold, then the preprocessed reverse processing data is determined to be negatively correlated feature data affecting reverse processing efficiency, and the first error affects the value-added reverse degree is output. If the proportion of quality inspection review time in the preprocessed reverse processing data is less than or equal to the time threshold and / or the proportion of refund process connection time is less than or equal to the time proportion threshold, then the preprocessed reverse processing data is determined to be positively correlated feature data of reverse processing efficiency, and the second error impact reduction reverse degree is output. The first error affects the inverse degree of added value, and the second error affects the inverse degree of subtraction. These two factors combine to form the inverse degree of the current error.

[0050] First, it's necessary to clarify the proportion of time spent on quality inspection and review, and the proportion of time spent on refund process integration, within the preprocessing and reverse engineering data. The quality inspection and review time proportion refers to the percentage of time spent on this step within the total reverse engineering time. For example, if the total reverse engineering time is 10 hours, and quality inspection and review takes 4 hours, then the quality inspection and review time proportion is 40%. The refund process integration time proportion refers to the percentage of time spent on the refund process integration steps within the total time. For example, if the refund process integration time is 3 hours, then its proportion is 30%. For instance, we can set the time threshold at 35% and the time proportion threshold at 25%.

[0051] Next, the relationship between these two percentages and the threshold is analyzed to determine the type of impact of preprocessed reverse processing data on reverse processing efficiency. If the percentage of time spent on quality inspection and review is greater than the time threshold, and the percentage of time spent on the refund process connection is greater than the time percentage threshold, then this data is determined to be negatively correlated characteristic data affecting reverse processing efficiency. That is, excessively high time consumption in these two stages will reduce reverse processing efficiency, and the corresponding output first error impact value-added reverse degree is set. For example, if the percentage of time spent on quality inspection and review is 40%, which is greater than the time threshold of 35%, and the percentage of time spent on the refund process connection is 30%, which is greater than the time percentage threshold of 25%, then the output first error impact value-added reverse degree can be set to 0.6. The larger the value, the stronger the negative impact on efficiency.

[0052] If the time spent on quality inspection and review is less than or equal to the time threshold, or the time spent on the refund process is less than or equal to the time percentage threshold, then this data is considered a positively correlated feature affecting reverse processing efficiency. This means the time spent on these two steps is reasonable and helps improve reverse processing efficiency. The corresponding output is the second error impact reduction in reverse degree. For example, if the time spent on quality inspection and review is 30% less than the 35% time threshold, even if the time spent on the refund process is 30% greater than the 25% time percentage threshold, this condition is still met. In this case, the output second error impact reduction in reverse degree can be set to -0.3, with a negative value indicating a positive impact on efficiency.

[0053] The current error impact inverse degree is obtained by combining the first error impact increment inverse degree and the second error impact decrement inverse degree. Current error impact inverse degree = first error impact increment inverse degree + second error impact decrement inverse degree. If the first error impact increment inverse degree is 0.6 and the second error impact decrement inverse degree is -0.3, then the current error impact inverse degree = 0.6 + (-0.3) = 0.3. This value reflects the overall impact of the current preprocessed reverse-processed data on the reverse-processing efficiency.

[0054] The second predicted inverse error value is obtained based on the correlation characteristics between the current error's impact on the inverse degree, the historical error's impact on the inverse degree, and the historical chain delay error. This process includes the following steps: The alignment result is obtained by performing time-series alignment processing on the historical error impact inverse degree and the historical chain delay error using the same time dimension division standard; Based on the alignment results, the feature changes corresponding to the two time-series features of historical error impact on reverse degree and historical chain delay error are filtered to obtain the association mapping relationship set; Hierarchical filtering retains the set of association mapping relationships that meet the requirements to form a valid set of association mapping relationships; Extract the current temporal features that affect the inverse degree of the current error and calculate the feature change of each temporal node. Match the feature change of each temporal node with the historical error increase / decrease impact feature change of the effective association mapping relationship set to determine the target historical association mapping relationship. Based on the historical correlation mapping relationship of the target, the change of the two features of the corresponding historical chain delay error is used as the reference value of the delay error change. The delay error change reference value is fused in time series to obtain the delay error change trend. The second predicted reverse error value is obtained based on the delay error change trend and the overall change of the current error affecting the reverse degree.

[0055] The historical error impact inverse degree and historical chain delay error 2 are time-series aligned using the same time dimension division standard. The time dimension division standard can be based on days or batches. For example, if the division is based on days, the historical error impact inverse degree for each day is matched with the historical chain delay error 2 for that day to obtain the alignment result. For instance, matching the historical error impact inverse degree for each day within a week with the historical chain delay error 2 for that day forms the time-series aligned data for that week.

[0056] Based on the alignment results, the feature changes corresponding to the historical error impact on the inverse degree and the historical chain delay error are filtered to obtain a set of association mapping relationships. Feature change refers to the change in data between adjacent time-series nodes. For example, if the historical error impact on the inverse degree is 0.3 on a certain day and 0.5 the next day, its feature change is 0.2; correspondingly, if the historical chain delay error is 2 hours on a certain day and 2.5 hours the next day, its feature change is 0.5 hours. These corresponding feature changes are combined; for example, a historical error impact on the inverse degree change of 0.2 corresponds to a historical chain delay error change of 0.5 hours, thus forming a set of association mapping relationships. Multiple such sets constitute the association mapping relationship set.

[0057] The process involves stratified filtering to retain only the sets of association mappings that meet the required correlation degree, thus forming a valid set of association mappings. The correlation degree can be determined by calculating the correlation coefficient of the feature changes. For example, if a correlation coefficient threshold of 0.7 is set, and a set of association mappings has a correlation coefficient of 0.8, it is retained because it is higher than the threshold, while relationships with correlation coefficients lower than the threshold are discarded. The final set of relationships remaining is the valid set of association mappings.

[0058] Extract the current temporal features of the impact of the current error on the inverse degree, and calculate the feature change at each temporal node. For example, if the current error impact on the inverse degree is 0.2, 0.4, and 0.5 for three consecutive days, then the feature change from day one to day two is 0.2, and the feature change from day two to day three is 0.1. Match these current feature changes with the historical error increase / decrease impact feature changes in the effective association mapping relationship set to determine the target historical association mapping relationship. For example, if the current feature change is 0.2, and it matches the historical chain delay error change of 0.5 hours corresponding to the historical error impact on the inverse degree change of 0.2 in the effective set, this relationship is the target historical association mapping relationship.

[0059] Based on the historical correlation mapping relationship of the target, the change in the second feature of the corresponding historical chain delay error is used as the reference value for delay error change. Taking the above matching relationship as an example, the change in the second feature of the corresponding historical chain delay error of 0.5 hours is the reference value for delay error change. Subsequently, time series fusion is performed on the reference value for delay error change to obtain the trend of delay error change. Time series fusion can be to calculate the mean or trend slope of continuous reference values. For example, if three consecutive matched reference values ​​are 0.5 hours, 0.4 hours, and 0.6 hours, their change trend is fluctuating upward.

[0060] The second predicted inverse error value is obtained based on the trend of the delay error and the overall change in the inverse degree affected by the current error. The overall change in the inverse degree affected by the current error refers to the total change in data within the current time series. For example, if the inverse degree affected by the current error changes from an initial 0.2 to a final 0.5, the overall change is 0.3. The second predicted inverse error value = the historical chain delay error corresponding to the initial value of the inverse degree affected by the current error + (the overall change in the inverse degree affected by the current error × the slope of the delay error trend). Assuming that the historical chain delay error corresponding to the initial value of the inverse degree affected by the current error is 0.2, and the slope of the delay error trend is 2, that is, for every change of 1 in the inverse degree affected by the error, the delay error changes by 2 hours, then the second predicted inverse error value = 2 + (0.3 × 2) = 2.6 hours.

[0061] The actual reverse processing efficiency is obtained by analyzing the initial reverse processing efficiency based on the first and second predicted reverse error values, specifically including the following steps: Extract the error impact dimension features corresponding to the first and second predicted inverse error values, and clarify the constraint class impact dimension corresponding to the first predicted inverse error value and the process time class impact dimension corresponding to the second predicted inverse error value based on the error impact dimension features; Based on the weight differences between the constraint-type influence dimension and the process time-type influence dimension, the first and second predicted reverse error values ​​are assigned hierarchically to obtain the weighted first and second predicted reverse error values. The weighted first and second predicted reverse error values ​​are correlated and verified with the initial reverse processing efficiency, and error correction directions that are compatible with the initial reverse processing efficiency are selected. Based on the error correction direction, the first predicted reverse error value and the second predicted reverse error value are applied sequentially to the initial reverse processing efficiency for correction, and then the actual reverse processing efficiency is obtained.

[0062] First, we extract the error impact dimension features corresponding to the first and second predicted reverse error values ​​to clarify their respective impact dimension types. Error impact dimension features refer to the core direction of the reverse process where the error acts. The first predicted reverse error value corresponds to a constraint-type impact dimension, which is the error impact caused by return reasons and sorting adaptation constraints. The second predicted reverse error value corresponds to a process time-type impact dimension, which is the error impact caused by the time spent in quality inspection and refund connection processes. For example, if the first predicted reverse error value is due to an unreasonable return reason, it belongs to the constraint-type dimension; if the second predicted reverse error value is due to excessive quality inspection time, it belongs to the process time-type dimension.

[0063] The two prediction error values ​​are assigned stratified values ​​based on the weight differences of the two types of influence dimensions, resulting in weighted first and second prediction inverse error values. The weight differences are set according to the actual impact of the two types of dimensions on inverse efficiency; for example, the weight of the constraint-related influence dimension is set to 0.6, and the weight of the process time-related influence dimension is set to 0.4. Assuming the first prediction inverse error value is 2.058 hours and the second prediction inverse error value is 2.6 hours, then the weighted first prediction inverse error value = 2.058 × 0.6 ≈ 1.235 hours, and the weighted second prediction inverse error value = 2.6 × 0.4 = 1.04 hours.

[0064] The weighted first and second predicted reverse error values ​​are correlated and verified with the initial reverse processing efficiency to determine the appropriate error correction direction. Correlation verification determines the direction of the error's impact on the initial testing efficiency. For example, if the initial reverse processing efficiency is 3.47 pieces / hour, and the error is due to process blockage leading to an overestimation of efficiency, the correction direction is to reduce the initial testing efficiency; if the error is due to data statistical bias leading to an underestimation of efficiency, the correction direction is to increase the initial testing efficiency. Both types of errors represent efficiency losses caused by process or constraint factors; therefore, the correction direction is to reduce the initial testing efficiency.

[0065] Based on the error correction direction, the first and second predicted reverse error values ​​are sequentially applied to correct the initial reverse processing efficiency, yielding the actual reverse processing efficiency. Actual reverse processing efficiency = Initial reverse processing efficiency ÷ (1 + Weighted first predicted reverse error value ÷ Baseline time + Weighted second predicted reverse error value ÷ Baseline time), where the baseline time is the standard time for the reverse process, for example, set to 24 hours. Substituting the data, we get the actual reverse processing efficiency = 3.47 ÷ (1 + 1.235 ÷ 24 + 1.04 ÷ 24) ≈ 3.47 ÷ (1 + 0.051 + 0.043) ≈ 3.47 ÷ 1.094 ≈ 3.17 pieces / hour. This method yields the actual reverse processing efficiency, reflecting the true capability of the reverse process.

[0066] The processing result is output based on the actual reverse processing efficiency, specifically including the following steps: Determine the output direction by matching the corresponding optimization requirements based on the actual reverse processing efficiency level; Based on the output direction, the constraint-type impact information and process time-type impact information related to the actual reverse processing efficiency are integrated to generate basic output information including efficiency level, influencing factors and optimization direction. The basic output information is hierarchically organized, and key contents are sorted according to optimization priority to form a structured processing result; The structured processing results are adapted to the corresponding output format based on different application scenarios of warehouse logistics management to obtain the final processing result.

[0067] By matching the actual reverse processing efficiency level with the corresponding optimization requirements, the output direction is determined. The actual reverse processing efficiency level is usually categorized by processing capacity; for example, a standard of 3 pieces / hour or more is considered excellent, 2-3 pieces / hour is good, and below 2 pieces / hour is considered needing optimization. Taking an actual reverse processing efficiency of 3.17 pieces / hour as an example, its level is good, and the corresponding optimization requirement is to improve process flow efficiency. Therefore, the output direction is determined to be process optimization-oriented.

[0068] Based on the output direction, basic output information is generated by integrating constraint-based impact information related to the actual reverse processing efficiency and process time-related impact information. Constraint-based impact information refers to the specific impact of factors such as the reasonableness of return reasons and sorting suitability constraints. For example, if unrelated return reasons account for 30%, it increases redundant sorting steps. Process time-related impact information refers to the impact of time-consuming steps such as quality inspection and refund coordination. For example, if quality inspection review accounts for 40% of the time, exceeding the threshold by 5%, this information is integrated with the efficiency level (good) and optimization direction (optimizing redundant sorting steps and quality inspection processes) to form basic output information that includes efficiency level, influencing factors, and optimization direction.

[0069] The basic output information is hierarchically organized, and key content is prioritized for optimization to form a structured processing result. The optimization priority is determined based on the degree of influence of factors on efficiency. For example, the impact weight of excessively high time consumption in quality inspection review on efficiency is 0.6, and the impact weight of a high proportion of unrelated return reasons is 0.4. Therefore, optimizing the quality inspection process and reducing the time consumption of quality inspection review is the first priority, and optimizing the return reason review mechanism and reducing the proportion of unrelated returns is the second priority. After prioritizing in this way, a clear and structured processing result is formed.

[0070] Based on different application scenarios in warehouse logistics management, the structured processing results are adapted to the corresponding output format. For example, for warehouse operation scenarios, the output format is an operation process optimization checklist, which clarifies the simplified steps in the quality inspection process; for management decision-making scenarios, the output format is an efficiency optimization report, which includes data statistics and priority suggestions; and for financial settlement scenarios, the output format is a refund process optimization plan, which focuses on shortening the refund process time. Through this scenario adaptation, processing results that can be directly applied to different warehouse management stages are obtained.

[0071] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.

[0072] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments or some parts of the embodiments.

[0073] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; 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; and these 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 warehousing and logistics data, characterized in that, The method includes the following steps: Real-time collection of data on returned goods, reverse flow, inventory backlog warnings, and traceability feedback in the warehouse yields preliminary reverse logistics data; real-time collection of data on the time spent on quality inspection and verification of goods during the reverse flow process, as well as the time required for the refund process, yields pre-processed reverse processing data. The initial reverse processing efficiency was obtained based on the initial reverse logistics data. Based on the difference in the reasonableness of the reasons for the return of stored goods and the difference in the adaptability of reverse sorting, the current reverse constraint feature data is obtained. Based on the current reverse constraint feature data and the correlation and variation characteristics between the historical difference in the reasonableness of the reasons for the return, the historical difference in the adaptability of reverse sorting and the historical chain delay error, the first predicted reverse error value is obtained. The current error impact degree is obtained based on the proportion of various time consumptions in the preprocessed reverse processing data. The second predicted reverse error value is obtained based on the correlation and variation characteristics between the current error impact degree, the historical error impact degree, and the historical chain delay error. The actual reverse processing efficiency is obtained by analyzing the initial reverse processing efficiency based on the first predicted reverse error value and the second predicted reverse error value. Output the processing results based on the actual reverse processing efficiency.

2. The warehousing and logistics data processing method according to claim 1, characterized in that, The initial reverse logistics efficiency is determined based on the initial reverse processing data, specifically including the following steps: Extract the goods reverse flow node information corresponding to the return reverse flow data in the initial reverse logistics data, extract the backlog goods category information corresponding to the inventory backlog warning data in the initial reverse logistics data, and extract the complete goods traceability link information corresponding to the traceability feedback data in the initial reverse logistics data; Based on the information of the reverse flow nodes, determine the cargo dwell information of each reverse flow node, and distinguish the differences in the dwell information of different categories of goods at each reverse flow node according to the cargo dwell information and the information of the backlog of goods categories; Based on the complete information of the cargo traceability link, determine the traceability matching status of each category of goods in reverse flow, and obtain the node adaptation information of each category of goods in reverse flow based on the traceability matching status and the difference in the dwell time of different categories of goods at each reverse flow node. Based on node adaptation information, the average turnover time for different categories of goods to complete the entire reverse flow process is obtained. The average turnover time is then corrected based on the backlog relief information corresponding to the inventory backlog warning data to obtain the corrected average turnover time. The initial reverse processing efficiency is obtained based on the corrected average turnover time and the volume of goods reversed flow corresponding to the returned goods reverse flow data.

3. The warehousing and logistics data processing method according to claim 1, characterized in that, The reverse flow data of returns includes return type, return path nodes, and return liability determination results; The inventory backlog warning data includes the amount of returned goods backlog and the inventory turnover cycle.

4. The warehousing and logistics data processing method according to claim 1, characterized in that, Based on the difference in the reasonableness of returned goods and the difference in reverse sorting suitability, the current reverse constraint feature data is obtained, which includes the following steps: Analyze the return reason correlation information corresponding to the reasonableness difference of the return reason, distinguish the correlation type between the return reason and the attributes of the goods themselves based on the return reason correlation information, and extract the difference fluctuation range information corresponding to the reasonableness difference of the return reason; The correctness of return reasons is determined based on the association type, and the strength of reasonableness constraints corresponding to each association type is determined based on the correctness and difference fluctuation range information. Analyze the sorting process association information corresponding to the reverse sorting fit difference, identify the matching gap between the sorting link and the reverse flow of goods based on the sorting process association information, and extract the change trend information of the difference corresponding to the reverse sorting fit difference. Based on the matching gap, the impact level of the gap is divided, and the fitness constraint strength corresponding to each impact level of the gap is determined according to the difference change trend information. The current reverse constraint feature data is obtained by processing the strength of the correlation rationality constraint, the strength of the fit constraint, and the level of the gap influence.

5. The warehousing and logistics data processing method according to claim 4, characterized in that, The current inverse constraint feature data is obtained by processing the strength of the correlation rationality constraint, the strength of the fit constraint, and the level of the gap influence. The specific steps include: After associating the rationality constraint strength and the fit constraint strength, the constraint items that affect the reverse flow process are screened out; The current reverse constraint feature data is formed by combining the association weights of the association types and gap impact levels corresponding to each constraint item.

6. The warehousing and logistics data processing method according to claim 5, characterized in that, Based on the current reverse constraint feature data and the correlation characteristics between the historical return reason rationality difference, the historical reverse sorting fit difference, and the historical chain delay error, the first predicted reverse error value is obtained, specifically including the following steps: After classifying the differences in the rationality of historical return reasons, the differences in historical reverse sorting adaptability, and the historical chain delay error based on their common origin, a combination of historical features is formed. For each set of historical feature combinations, the correlation between the fluctuation range of the reasonableness difference of historical return reasons, the change gradient of the historical reverse sorting adaptability difference, and the change direction and amplitude of the corresponding historical chain delay error is determined, thus obtaining the historical correlation influence feature set. The current reverse constraint feature data is decomposed into the fluctuation feature of the current return reason reasonableness difference and the change feature of the current reverse sorting suitability difference. The current return reason reasonableness difference fluctuation feature and the current reverse sorting suitability difference change feature are compared with the corresponding historical features in the historical correlation influence feature set to determine the target historical correlation influence feature set that matches the current feature trend. The deviation between the current return reason reasonableness difference fluctuation feature and the corresponding historical fluctuation feature in the target historical feature set, as well as the deviation between the current reverse sorting suitability difference change feature and the corresponding historical change feature in the target historical feature set, are calculated to form the feature difference quantification result. Based on the correlation pattern between historical features and historical chain delay error 1 in the target historical correlation influence feature set, and combined with the feature difference quantification results, the deviation correction of historical chain delay error 1 corresponding to the target historical features is performed; the initial prediction reverse error value is generated by comprehensively considering the impact of the corrected historical chain delay error 1 and the feature difference quantification results on the current reverse constraint feature data. The first prediction inverse error value is obtained by calibrating the initial prediction inverse error value using the fit of historical feature combinations.

7. The warehousing and logistics data processing method according to claim 6, characterized in that, The current error impact degree is obtained based on the proportion of time spent on various types of data in the preprocessed and reverse-processed data. This includes the following steps: If the proportion of quality inspection review time in the preprocessed reverse processing data is greater than the time threshold and the proportion of refund process connection time is greater than the time proportion threshold, then the preprocessed reverse processing data is determined to be negatively correlated feature data affecting reverse processing efficiency, and the first error affects the value-added reverse degree is output. If the proportion of quality inspection review time in the preprocessed reverse processing data is less than or equal to the time threshold and / or the proportion of refund process connection time is less than or equal to the time proportion threshold, then the preprocessed reverse processing data is determined to be positively correlated feature data of reverse processing efficiency, and the second error impact reduction reverse degree is output. The first error affects the inverse degree of added value, and the second error affects the inverse degree of subtraction. These two factors combine to form the inverse degree of the current error.

8. The warehousing and logistics data processing method according to claim 7, characterized in that, The second predicted inverse error value is obtained based on the correlation characteristics between the current error's impact on the inverse degree, the historical error's impact on the inverse degree, and the historical chain delay error. This process includes the following steps: The alignment result is obtained by performing time-series alignment processing on the historical error impact inverse degree and the historical chain delay error using the same time dimension division standard; Based on the alignment results, the feature changes corresponding to the two time-series features of historical error impact on reverse degree and historical chain delay error are filtered to obtain the association mapping relationship set; Hierarchical filtering retains the set of association mapping relationships that meet the requirements to form a valid set of association mapping relationships; Extract the current temporal features that affect the inverse degree of the current error and calculate the feature change of each temporal node. Match the feature change of each temporal node with the historical error increase / decrease impact feature change of the effective association mapping relationship set to determine the target historical association mapping relationship. Based on the historical correlation mapping relationship of the target, the change of the two features of the corresponding historical chain delay error is used as the reference value of the delay error change. The delay error change reference value is fused in time series to obtain the delay error change trend. The second predicted reverse error value is obtained based on the delay error change trend and the overall change of the current error affecting the reverse degree.

9. The warehousing and logistics data processing method according to claim 8, characterized in that, The actual reverse processing efficiency is obtained by analyzing the initial reverse processing efficiency based on the first and second predicted reverse error values, specifically including the following steps: Extract the error impact dimension features corresponding to the first and second predicted inverse error values, and clarify the constraint class impact dimension corresponding to the first predicted inverse error value and the process time class impact dimension corresponding to the second predicted inverse error value based on the error impact dimension features; Based on the weight differences between the constraint-type influence dimension and the process time-type influence dimension, the first and second predicted reverse error values ​​are assigned hierarchically to obtain the weighted first and second predicted reverse error values. The weighted first and second predicted reverse error values ​​are correlated and verified with the initial reverse processing efficiency, and error correction directions that are compatible with the initial reverse processing efficiency are selected. Based on the error correction direction, the first predicted reverse error value and the second predicted reverse error value are applied sequentially to the initial reverse processing efficiency for correction, and then the actual reverse processing efficiency is obtained.

10. The warehousing and logistics data processing method according to claim 1, characterized in that, The processing result is output based on the actual reverse processing efficiency, specifically including the following steps: Determine the output direction by matching the corresponding optimization requirements based on the actual reverse processing efficiency level; Based on the output direction, the constraint-type impact information and process time-type impact information related to the actual reverse processing efficiency are integrated to generate basic output information including efficiency level, influencing factors and optimization direction. The basic output information is hierarchically organized, and key contents are sorted according to optimization priority to form a structured processing result; The structured processing results are adapted to the corresponding output format based on different application scenarios of warehouse logistics management to obtain the final processing result.

Images