Logistics order full-link signing timeliness monitoring method and device, equipment and medium

Abstract

The invention provides a logistics order full-link signing timeliness monitoring method and device, equipment and a medium, and the method comprises the steps: constructing a timeliness standard prediction model, and training the constructed timeliness standard prediction model; predicting the sign-off duration by using the trained aging standard prediction model; through order trajectory data collected in real time, according to an anomaly detection mechanism, correcting the actual signing duration to obtain the corrected actual signing duration; when the corrected actual signing duration is greater than the predicted signing duration, according to the multi-dimensional influence factors, performing principal component analysis dimensionality reduction, and extracting a core delay risk feature vector meeting a preset requirement; marking a corresponding historical order based on the extracted core delay risk feature vector meeting the preset requirement, and training an early warning grading model by using the marked historical order to obtain a trained early warning grading model; and carrying out delay risk classification and grading on the orders by utilizing the trained early warning grading model.

Description

Technical Field

[0001] The present invention relates to the field of logistics management technology, and specifically to a full-link collection and signing time monitoring method, device, equipment and medium for logistics orders. Background Art

[0002] In today's efficient business environment, logistics timeliness is of vital importance. Whether it is the fulfillment of e-commerce companies' delivery commitments to consumers or the smooth coordination of supply chains between companies, accurately controlling the time from logistics order collection to receipt, timely detecting delays and taking countermeasures are key factors in improving logistics service quality, reducing operating costs, and enhancing customer satisfaction. However, existing logistics timeliness monitoring often has problems such as inaccurate data statistics, rigid application of timeliness standards, and lack of dynamic flexibility in delay determination. It cannot adapt to complex and changing logistics scenarios and business needs. Therefore, there is an urgent need for a refined, intelligent and efficient full-link collection and receipt timeliness monitoring solution. Summary of the invention

[0003] The main purpose of the present invention is to solve the technical problems in the prior art of inaccurate statistics of logistics timeliness monitoring data, rigid application of timeliness standards, lack of dynamic flexibility in delay determination, and inability to adapt to complex and changeable logistics scenarios and business needs.

[0004] The first aspect of the present invention provides a method for monitoring the timeliness of a full-link collection and signing of a logistics order, comprising: Constructing a timeliness standard prediction model, and training the constructed timeliness standard prediction model to obtain a trained timeliness standard prediction model; Use the trained time efficiency standard prediction model to predict the visa collection time; Through the real-time collected order trajectory data, the actual pickup time is corrected according to the anomaly detection mechanism to obtain the corrected actual pickup time. When the corrected actual visa collection time is greater than the predicted visa collection time, the core delay risk feature vector that meets the preset requirements is extracted based on multi-dimensional influencing factors and dimensionality reduction through principal component analysis; Based on the extracted core delay risk feature vector that meets the preset requirements, the corresponding historical orders are labeled, and the early warning classification model is trained using the labeled historical orders to obtain the trained early warning classification model; Use the trained early warning classification model to classify and grade the delay risks of orders, and push the early warning classification model warning results to the client according to the preset requirements, so that the client can formulate corresponding adjustment strategies; The timeliness standard prediction model is a hybrid model constructed by integrating a neural network and a gradient boosting decision tree, and the signing time is predicted based on the constructed hybrid model; The early warning classification model is constructed based on logistic regression and support vector machine.

[0005] Optionally, in a first implementation of the first aspect of the present invention, constructing a timeliness standard prediction model and training the constructed timeliness standard prediction model to obtain a trained timeliness standard prediction model includes: Construct a timeliness standard prediction model based on neural network and gradient boosting decision tree; Acquire historical logistics order data under different conditions, and preprocess the acquired historical logistics order data to obtain preprocessed historical logistics order data; Construct training and test sets based on preprocessed historical logistics order data; The constructed timeliness standard prediction model is trained with the training set through multiple rounds of back propagation and gradient boosting iterations to obtain the trained timeliness standard prediction model; The test set is used to verify and evaluate the currently trained timeliness standard prediction model through the mean absolute error and root mean square error. When the evaluation fails to meet the preset requirements, the hyperparameters of the currently trained timeliness standard prediction model are adjusted, and the training is triggered repeatedly until the test set evaluation meets the preset requirements.

[0006] Optionally, in a second implementation of the first aspect of the present invention, the acquiring of historical logistics order data under different conditions, and preprocessing the acquired historical logistics order data to obtain the preprocessed historical logistics order data includes: Obtain historical logistics order data under different conditions; where different conditions include: different seasons, time periods, shipping and receiving areas, cargo types, and different dimensions of logistics carrier information; the historical logistics order data includes: the longitude and latitude of the shipping city, the longitude and latitude of the receiving city, the order timestamp, the weight and volume of the cargo, the carrier code, and the full-link stipulated pickup time; The historical logistics order data obtained under different conditions are processed including removing outliers, supplementing missing values, one-hot encoding of categorical variables and normalization of numerical variables to obtain the processed historical logistics order data.

[0007] Optionally, in a third implementation of the first aspect of the present invention, the predicted collection time is corrected according to the anomaly detection mechanism using the order trajectory data collected in real time to obtain the corrected collection time, including: Collect order track data in real time based on the information system of each link of logistics; wherein, the information system of each link of logistics includes: PDA at the collection end, GPS of transportation vehicle, warehouse management WMS and distribution handheld terminal; The Kalman filter algorithm is used to integrate GPS positioning and site scanning time to correct the time error caused by signal drift and equipment delay, and the actual signature collection time is obtained after correction; The isolation forest algorithm is used to identify anomalies in the actual pickup time. The weighted average method is used to correct the current corrected actual pickup time based on the historical average time for the same region and type of goods during the same period and the current logistics congestion index to obtain the corrected actual pickup time.

[0008] Optionally, in a fourth implementation of the first aspect of the present invention, when the corrected actual pickup time is greater than the predicted pickup time, based on multi-dimensional influencing factors, principal component analysis is performed to reduce the dimension, and a core delay risk feature vector that meets preset requirements is extracted, including: When the corrected actual pickup time is greater than the predicted pickup time, based on multi-dimensional influencing factors including real-time weather data of the logistics process, the full load rate of the transport vehicle and the saturation of the deliveryman's daily tasks, principal component analysis is used to reduce the dimension and extract the core delay risk feature vector that meets the preset requirements.

[0009] Optionally, in a fifth implementation of the first aspect of the present invention, the corresponding historical orders are labeled based on the extracted core delay risk feature vector that meets preset requirements, and the labeled historical orders are used to train the early warning classification model to obtain the trained early warning classification model, including: Construct an early warning classification model based on logistic regression and support vector machine; Based on the extracted core delay risk feature vectors that meet the preset requirements, the corresponding historical orders are labeled, and the labeled historical orders are used to train the early warning classification model to obtain the trained early warning classification model; Using the early warning classification model to classify and grade the current delay risk of the order; wherein the classification includes delay or normal; the classification includes mild, moderate or severe delay; The warning results of the warning classification model are pushed to the client according to the preset requirements.

[0010] Optionally, in a sixth implementation of the first aspect of the present invention, the method further includes: Deploy the performance monitoring module of the timeliness standard prediction model to track the prediction in real time, including accuracy, MAE, and RMSE indicators. When the indicators deteriorate, trigger retrospective analysis including the source of data deviation, overfitting or underfitting of the timeliness standard prediction model, and formulate corresponding tuning strategies.

[0011] The second aspect of the present invention provides a full-link signing time monitoring device for logistics orders, comprising: Constructing a timeliness standard prediction model module, which is used to construct a timeliness standard prediction model, and train the constructed timeliness standard prediction model to obtain a trained timeliness standard prediction model; The visa collection time prediction module is used to predict the visa collection time using the trained time efficiency standard prediction model; The actual pickup time correction module is used to correct the actual pickup time based on the order trajectory data collected in real time and the anomaly detection mechanism to obtain the corrected actual pickup time; The delay risk extraction module is used to extract the core delay risk feature vector that meets the preset requirements based on multi-dimensional influencing factors and principal component analysis and dimensionality reduction when the actual revised visa collection time is greater than the predicted visa collection time. The warning classification module is a construction module, which is used to mark the corresponding historical orders based on the extracted core delay risk feature vectors that meet the preset requirements, and train the warning classification model using the marked historical orders to obtain the trained warning classification model; The classification warning module is used to classify and grade the delay risks of orders using the trained warning classification model; The timeliness standard prediction model is a hybrid model constructed by integrating a neural network and a gradient boosting decision tree, and the signing time is predicted based on the constructed hybrid model; The early warning classification model is constructed based on logistic regression and support vector machine.

[0012] Optionally, in a first implementation of the second aspect of the present invention, the module for constructing a timeliness standard prediction model includes: Construct a timeliness standard prediction model based on neural network and gradient boosting decision tree; Acquire historical logistics order data under different conditions, and preprocess the acquired historical logistics order data to obtain preprocessed historical logistics order data; Construct training and test sets based on preprocessed historical logistics order data; The constructed timeliness standard prediction model is trained with the training set through multiple rounds of back propagation and gradient boosting iterations to obtain the trained timeliness standard prediction model; The test set is used to verify and evaluate the currently trained timeliness standard prediction model through the mean absolute error and root mean square error. When the evaluation fails to meet the preset requirements, the hyperparameters of the currently trained timeliness standard prediction model are adjusted, and the training is triggered repeatedly until the test set evaluation meets the preset requirements.

[0013] Optionally, in a second implementation of the second aspect of the present invention, the acquiring of historical logistics order data under different conditions, and preprocessing the acquired historical logistics order data to obtain the preprocessed historical logistics order data includes: Obtain historical logistics order data under different conditions; where different conditions include: different seasons, time periods, shipping and receiving areas, cargo types, and different dimensions of logistics carrier information; the historical logistics order data includes: the longitude and latitude of the shipping city, the longitude and latitude of the receiving city, the order timestamp, the weight and volume of the cargo, the carrier code, and the full-link stipulated pickup time; The historical logistics order data obtained under different conditions are processed including removing outliers, supplementing missing values, one-hot encoding of categorical variables and normalization of numerical variables to obtain the processed historical logistics order data.

[0014] Optionally, in a third implementation of the second aspect of the present invention, the actual signature collection time correction module includes: Collect order track data in real time based on the information system of each link of logistics; wherein, the information system of each link of logistics includes: PDA at the collection end, GPS of transportation vehicle, warehouse management WMS and distribution handheld terminal; The Kalman filter algorithm is used to integrate GPS positioning and site scanning time to correct the time error caused by signal drift and equipment delay, and the actual signature collection time is obtained after correction; The isolation forest algorithm is used to identify anomalies in the actual pickup time. The weighted average method is used to correct the current corrected actual pickup time based on the historical average time for the same region and type of goods during the same period and the current logistics congestion index to obtain the corrected actual pickup time.

[0015] Optionally, in a fourth implementation of the second aspect of the present invention, the delay risk refining module includes: When the corrected actual pickup time is greater than the predicted pickup time, based on multi-dimensional influencing factors including real-time weather data of the logistics process, the full load rate of the transport vehicle and the saturation of the deliveryman's daily tasks, principal component analysis is used to reduce the dimension and extract the core delay risk feature vector that meets the preset requirements.

[0016] Optionally, in a fifth implementation of the second aspect of the present invention, the warning classification module construction module includes: Construct an early warning classification model based on logistic regression and support vector machine; Based on the extracted core delay risk feature vectors that meet the preset requirements, the corresponding historical orders are labeled, and the labeled historical orders are used to train the early warning classification model to obtain the trained early warning classification model; Using the early warning classification model to classify and grade the current delay risk of the order; wherein the classification includes delay or normal; the classification includes mild, moderate or severe delay; The warning results of the warning classification model are pushed to the client according to the preset requirements.

[0017] Optionally, in the sixth implementation method of the second aspect of the present invention, the device also includes: a performance monitoring module, which is used to track and predict in real time, including accuracy, MAE and RMSE indicators. When the indicators deteriorate, it triggers a retrospective analysis including the source of data deviation, the overfitting or underfitting condition of the timeliness standard prediction model, and formulates a corresponding tuning strategy.

[0018] The third aspect of the present invention provides a computer-readable storage medium storing a computer program, characterized in that when the computer program is executed by a processor, the steps of the full-link collection and signing time monitoring method for logistics orders described above are implemented.

[0019] The fourth aspect of the present invention provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that when the computer program is executed by the processor, the steps of the full-link collection and signing time monitoring method for logistics orders described above are implemented.

[0020] Compared with the prior art, the present invention has the following beneficial effects: 1. The present invention constructs a hybrid model that integrates a neural network and a gradient boosted decision tree (GBDT), which can better capture the characteristics of complex nonlinear relationships and structured data, thereby improving the prediction accuracy of the signing time; 2. The present invention combines the real-time collected order trajectory data with the Kalman filter algorithm to further correct the time error caused by equipment delay and signal drift, ensuring the calculation accuracy of the actual signature collection time; 3. The present invention uses the isolation forest algorithm to detect abnormalities in the actual signing time, which can quickly identify abnormal times caused by various reasons, such as warehouse explosions, bad weather, etc., and make corrections in time to ensure the authenticity and reliability of the data; 4. The present invention extracts the core delay risk feature vector through principal component analysis (PCA), which can comprehensively consider multiple influencing factors, such as weather, transportation vehicle status, delivery personnel task saturation, etc., so as to more comprehensively evaluate the delay risk and help the operation team make more scientific decisions; 5. The early warning classification model constructed by the present invention can classify and grade the delay risks of orders, and push corresponding early warning information according to the risk level, ensuring that the operation team can respond to potential delay problems in a timely manner and improve operational efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] Other features, objects and advantages of the present invention will become more apparent from the detailed description of non-limiting embodiments made with reference to the following drawings: Figure 1A first flow chart of a method for monitoring the timeliness of logistics orders in the entire chain provided in an embodiment of the present invention.

[0022] Figure 2 A second flow chart of the full-link collection and signing time monitoring method for logistics orders provided in an embodiment of the present invention.

[0023] Figure 3 A structural schematic diagram of a full-link collection and signing time monitoring device for logistics orders provided in an embodiment of the present invention.

[0024] Figure 4 A schematic diagram of the structure of an electronic device provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0025] The embodiment of the present invention provides a full-link collection and signing time monitoring method, device, equipment and medium for logistics orders, constructs a time standard prediction model, and trains the constructed time standard prediction model to obtain a trained time standard prediction model; uses the trained time standard prediction model to predict the collection and signing time; through the real-time collected order trajectory data, according to the anomaly detection mechanism, the actual collection and signing time is corrected to obtain the corrected actual collection and signing time; when the corrected actual collection and signing time is greater than the predicted collection and signing time, according to the multi-dimensional influencing factors, the principal component analysis is reduced in dimension to extract the core delay risk feature direction that meets the preset requirements. quantity; based on the extracted core delay risk feature vector that meets the preset requirements, the corresponding historical orders are labeled, and the early warning classification model is trained using the labeled historical orders to obtain the trained early warning classification model; the order delay risk is classified and graded using the trained early warning classification model, and the early warning classification model warning results are pushed to the client according to the preset requirements, so that the client can formulate corresponding adjustment strategies; the timeliness standard prediction model is a hybrid model constructed by integrating neural network and gradient boosting decision tree, and the signing time is predicted based on the constructed hybrid model; the early warning classification model is a warning classification model based on logistic regression and support vector machine. The present invention solves the technical problems in the prior art of inaccurate statistics of logistics timeliness monitoring data, rigid application of timeliness standards, lack of dynamic flexibility in delay judgment, and inability to adapt to complex and changing logistics scenarios and business needs.

[0026] The terms "first", "second", "third", "fourth", etc. (if any) in the specification and claims of the present invention and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchanged where appropriate, so that the embodiments described herein can be implemented in an order other than that illustrated or described herein. In addition, the terms "including" or "having" and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units that are clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0027] For ease of understanding, the specific process of the embodiment of the present invention is described below. Figure 1 The first embodiment of the full-link signing time monitoring method for logistics orders in the embodiment of the present invention includes: 101. Constructing a timeliness standard prediction model, and training the constructed timeliness standard prediction model to obtain a trained timeliness standard prediction model; In this embodiment, a timeliness standard prediction model is constructed based on a hybrid model that integrates a neural network and a gradient boosting decision tree (GBDT). The neural network is good at capturing complex nonlinear relationships and learning hidden patterns of various factors in logistics. GBDT has strong ability to summarize structured data and excellent interpretability.

[0028] Obtain historical logistics order data under different conditions, covering different seasons, time periods, shipping and receiving areas (detailed to the district and county level), cargo types, logistics carriers and other dimensional information, and preprocess the acquired historical logistics order data to obtain the preprocessed historical logistics order data; among them, the preprocessing of the acquired historical logistics order data includes cleaning the data, eliminating outliers (such as records where the duration deviates from the normal range due to equipment failure), using multiple filling methods for missing values, and supplementing them according to feature correlation fitting; uniquely encoding categorical variables (such as cargo type: fresh food, electronics, daily necessities, etc.), and normalizing numerical variables (weight, distance, etc.) to lay a solid foundation for model training.

[0029] Based on the pre-processed historical logistics order data, the training set and test set are constructed in an 8:2 ratio. The longitude and latitude of the sending city, the longitude and latitude of the receiving city, the order timestamp, the weight and volume of the goods, the carrier code, etc. are used as input features, and the full-link stipulated collection and signing time is used as the output label. After multiple rounds of back propagation and gradient boosting iterative training, the model hyperparameters (such as the number of neural network layers, the number of nodes, the depth of the GBDT tree, etc.) are optimized to minimize the mean absolute error (MAE) and root mean square error (RMSE) of the test set, and obtain the timeliness standard prediction model.

[0030] 102. Use the trained time efficiency standard prediction model to predict the visa collection time; 103. Through the order trajectory data collected in real time, according to the abnormality detection mechanism, the actual pickup time is corrected to obtain the corrected actual pickup time; In this embodiment, order track data is collected in real time based on the information system of each link of logistics; wherein, the information system of each link of logistics includes: PDA at the collection end, GPS of the transport vehicle, warehouse management WMS and distribution handheld terminal; The Kalman filter algorithm is used to integrate GPS positioning and site scanning time, correct the time error caused by signal drift and equipment delay, lock the first collection time and the storage time corresponding to the first normal signing scan, and ensure the calculation accuracy of the actual collection time of the entire link; The isolation forest algorithm is used to identify anomalies in the actual pickup time. The algorithm constructs multiple random binary trees and measures the "degree of outliers" of the data points according to their depth in the tree. For the time data judged as abnormal (such as temporary storage turnover due to warehouse overflow, which causes the signature delay to be far longer than normal), the weighted average method is used to correct the current corrected actual pickup time based on the historical average time of the same region and type of goods in the same period and the current logistics congestion index (obtained by analyzing the busyness of transportation routes through real-time traffic big data), to obtain the corrected actual pickup time, thereby ensuring that the time truly reflects the normal operation time efficiency.

[0031] 104. When the actual revised visa collection time is greater than the predicted visa collection time, the core delay risk feature vector that meets the preset requirements is extracted based on multi-dimensional influencing factors and dimensionality reduction through principal component analysis; In this embodiment, the actual collection time of the entire order chain, the predicted timeliness standard, the real-time weather data of the logistics process (calling the meteorological API to obtain severe weather conditions such as precipitation, snowfall, fog, etc. in the passing area), the full load rate of the transport vehicle (combined with the on-board weighing and volume sensor data statistics), the deliveryman’s daily task saturation and other multi-dimensional characteristics are integrated, and the principal component analysis (PCA) is used for dimensionality reduction to refine the core delay risk feature vector, filter out redundant information, and highlight the key influencing factors.

[0032] 105. Label the corresponding historical orders based on the extracted core delay risk feature vector that meets the preset requirements, and use the labeled historical orders to train the early warning classification model to obtain the trained early warning classification model; In this embodiment, a combination of logistic regression and support vector machine (SVM) model is used, and the feature vector after dimension reduction is used as input to construct a delay risk binary classification (delay / normal) and multi-level (mild, moderate, severe delay) discrimination model. The historical order annotated sample training is used to adjust the SVM kernel function (radial basis function, etc.) and logistic regression coefficient, optimize the model recall rate and precision rate, so that the early warning can capture potential delayed orders in advance, and push the corresponding disposal strategy to the operation team according to the risk level.

[0033] 106. Use the trained early warning classification model to classify and grade the delay risks of orders, and push the early warning results of the early warning classification model to the client according to the preset requirements, so that the client can formulate corresponding adjustment strategies.

[0034] In this embodiment, multi-channel intelligent warning push: deeply integrated with enterprise office software (such as DingTalk, Enterprise WeChat) and SMS platforms, differentiated warning message templates are set according to the delay risk level, severe delays are directly pushed to senior management and emergency dispatch teams, with detailed time analysis charts attached; mild and moderate delays are pushed to corresponding business posts (customer service pre-notifies customers, dispatch optimizes subsequent delivery), to achieve information reach and collaborative and efficient response.

[0035] Closed-loop data analysis feedback: Analyze the distribution of delayed order characteristics by business cycle (week / month), dig into the underlying causes (shortcomings in regional logistics infrastructure, gaps in personnel skills, etc.) and draw a visual dashboard; drive process reengineering and resource reallocation based on the analysis results, such as adding transportation capacity and optimizing warehouse layout in areas with high delays, and feed back improvement measures into model training to continuously iterate and improve monitoring and operational efficiency.

[0036] This embodiment can significantly improve the accuracy, response speed and overall efficiency of logistics operations by building an efficient timeliness monitoring and early warning system, bringing better operational management results and customer experience to the enterprise.

[0037] See also Figure 2 , a second embodiment of the full-link signing time monitoring method for logistics orders in an embodiment of the present invention includes: 201. Constructing a timeliness standard prediction model, and training the constructed timeliness standard prediction model to obtain a trained timeliness standard prediction model; 202. Use the trained time efficiency standard prediction model to predict the visa collection time; 203. Using the order trajectory data collected in real time, according to the abnormality detection mechanism, the actual pickup time is corrected to obtain the corrected actual pickup time; 204. When the actual revised visa collection time is greater than the predicted visa collection time, the core delay risk feature vector that meets the preset requirements is extracted based on multi-dimensional influencing factors and dimensionality reduction through principal component analysis; 205. Based on the extracted core delay risk feature vector that meets the preset requirements, the corresponding historical orders are labeled, and the early warning classification model is trained using the labeled historical orders to obtain the trained early warning classification model; 206. Use the trained early warning classification model to classify and grade the delay risks of the orders, and push the early warning results of the early warning classification model to the client according to the preset requirements, so that the client can formulate corresponding adjustment strategies; 207. Deploy the performance monitoring module of the timeliness standard prediction model to track the prediction in real time, including accuracy, MAE and RMSE indicators. When the indicators deteriorate, trigger the retrospective analysis including the source of data deviation, overfitting or underfitting of the timeliness standard prediction model, and formulate corresponding tuning strategies; In this embodiment, a model performance monitoring module is deployed to track indicators such as prediction accuracy, MAE, and RMSE in real time. When indicators deteriorate (such as the accuracy is lower than the set threshold), the module automatically back-analyzes the source of data deviation and model overfitting or underfitting, and triggers corresponding tuning strategies (re-sampling data, updating feature sets, and fine-tuning model structures) according to the problem type (data drift, feature failure, etc.).

[0038] At the same time, the timeliness standards are updated adaptively: new order data (manually reviewed and confirmed) are regularly (weekly or monthly) included in the training set, and the timeliness standard prediction model is updated through incremental learning to adapt to changes such as logistics network expansion and business strategy adjustments (opening of new routes, upgrade of delivery timeliness), ensuring that the timeliness standards are in line with operational dynamics.

[0039] The timeliness standard prediction model is a hybrid model constructed by integrating a neural network and a gradient boosting decision tree, and the signing time is predicted based on the constructed hybrid model; The early warning classification model is constructed based on logistic regression and support vector machine.

[0040] The above describes the full-link collection and signing time monitoring method for logistics orders in an embodiment of the present invention. The following describes the full-link collection and signing time monitoring device for logistics orders in an embodiment of the present invention. Please refer to Figure 3 In one embodiment of the present invention, a device for monitoring the timeliness of a full-link collection and signing of a logistics order includes: The module 301 for constructing a timeliness standard prediction model is used to construct a timeliness standard prediction model and train the constructed timeliness standard prediction model to obtain a trained timeliness standard prediction model; In this embodiment, the module 301 for constructing a timeliness standard prediction model includes: A timeliness standard prediction model is constructed based on a hybrid model that integrates neural networks and gradient boosting decision trees (GBDT). Neural networks are good at capturing complex nonlinear relationships and learning hidden patterns of various logistics factors. GBDT has strong ability to summarize structured data and excellent interpretability.

[0041] Obtain historical logistics order data under different conditions, covering different seasons, time periods, shipping and receiving areas (detailed to the district and county level), cargo types, logistics carriers and other dimensional information, and preprocess the acquired historical logistics order data to obtain the preprocessed historical logistics order data; among them, the preprocessing of the acquired historical logistics order data includes cleaning the data, eliminating outliers (such as records where the duration deviates from the normal range due to equipment failure), using multiple filling methods for missing values, and supplementing them according to feature correlation fitting; uniquely encoding categorical variables (such as cargo type: fresh food, electronics, daily necessities, etc.), and normalizing numerical variables (weight, distance, etc.) to lay a solid foundation for model training.

[0042] Based on the pre-processed historical logistics order data, the training set and test set are constructed in an 8:2 ratio. The longitude and latitude of the sending city, the longitude and latitude of the receiving city, the order timestamp, the weight and volume of the goods, the carrier code, etc. are used as input features, and the full-link stipulated collection and signing time is used as the output label. After multiple rounds of back propagation and gradient boosting iterative training, the model hyperparameters (such as the number of neural network layers, the number of nodes, the depth of the GBDT tree, etc.) are optimized to minimize the mean absolute error (MAE) and root mean square error (RMSE) of the test set, and obtain the timeliness standard prediction model.

[0043] The signature collection time prediction module 302 is used to predict the signature collection time using the trained time efficiency standard prediction model; The actual pickup time correction module 303 is used to correct the actual pickup time according to the order trajectory data collected in real time and the abnormality detection mechanism to obtain the corrected actual pickup time; In this embodiment, the actual signature collection time correction module 303 includes: Collect order track data in real time based on the information system of each link of logistics; wherein, the information system of each link of logistics includes: PDA at the collection end, GPS of transportation vehicle, warehouse management WMS and distribution handheld terminal; The Kalman filter algorithm is used to integrate GPS positioning and site scanning time, correct the time error caused by signal drift and equipment delay, lock the first collection time and the storage time corresponding to the first normal signing scan, and ensure the calculation accuracy of the actual collection time of the entire link; The isolation forest algorithm is used to identify anomalies in the actual pickup time. The algorithm constructs multiple random binary trees and measures the "degree of outliers" of the data points according to their depth in the tree. For the time data judged as abnormal (such as temporary storage turnover due to warehouse overflow, which causes the signature delay to be far longer than normal), the weighted average method is used to correct the current corrected actual pickup time based on the historical average time of the same region and type of goods in the same period and the current logistics congestion index (obtained by analyzing the busyness of transportation routes through real-time traffic big data), to obtain the corrected actual pickup time, thereby ensuring that the time truly reflects the normal operation time efficiency.

[0044] The delay risk extraction module 304 is used to extract the core delay risk feature vector that meets the preset requirements based on multi-dimensional influencing factors and principal component analysis and dimensionality reduction when the corrected actual visa collection time is greater than the predicted visa collection time; In this embodiment, the refining delay risk module 304 includes: The actual collection time of the entire order chain, the predicted timeliness standard, the real-time weather data of the logistics process (calling the meteorological API to obtain severe weather conditions such as precipitation, snowfall, fog, etc. in the passing area), the full load rate of the transport vehicle (combined with the statistics of on-board weighing and volume sensor data), the saturation of the delivery staff's daily tasks and other multi-dimensional characteristics are integrated, and the principal component analysis (PCA) is used to reduce the dimension, refine the core delay risk feature vector, filter out redundant information, and highlight the key influencing factors.

[0045] The warning classification module construction module 305 is used to mark the corresponding historical orders based on the extracted core delay risk feature vectors that meet the preset requirements, and train the warning classification model using the marked historical orders to obtain the trained warning classification model; In this embodiment, the warning classification module construction module 305 includes: A combination of logistic regression and support vector machine (SVM) model is used, and the feature vector after dimension reduction is used as input to construct a delay risk binary classification (delay / normal) and multi-level (mild, moderate, severe delay) discrimination model. The historical order annotated sample training is used to adjust the SVM kernel function (radial basis function, etc.) and logistic regression coefficient, optimize the model recall rate and precision rate, so that the early warning can capture potential delayed orders in advance, and push the corresponding disposal strategy to the operation team according to the risk level.

[0046] The classification warning module 306 is used to classify and grade the delay risks of orders using the trained warning classification model; In this embodiment, the classification warning module 306 includes: Multi-channel intelligent early warning push: deeply integrated with enterprise office software (such as DingTalk, WeChat for Enterprise) and SMS platforms, differentiated early warning message templates are set according to the delay risk level, and severe delays are directly pushed to senior management and emergency dispatch teams, with detailed time analysis charts attached; mild and moderate delays are pushed to corresponding business posts (customer service pre-notifies customers, dispatch optimizes subsequent delivery), achieving information reach and collaborative and efficient response.

[0047] Closed-loop data analysis feedback: Analyze the distribution of delayed order characteristics by business cycle (week / month), dig into the underlying causes (shortcomings in regional logistics infrastructure, gaps in personnel skills, etc.) and draw a visual dashboard; drive process reengineering and resource reallocation based on the analysis results, such as adding transportation capacity and optimizing warehouse layout in areas with high delays, and feed back improvement measures into model training to continuously iterate and improve monitoring and operational efficiency.

[0048] Performance monitoring module 307, used for real-time tracking of predictions including accuracy, MAE and RMSE indicators. When the indicators deteriorate, it triggers retrospective analysis including the source of data deviation, overfitting or underfitting of the timeliness standard prediction model, and formulates corresponding tuning strategies; In this embodiment, the prediction accuracy, MAE, RMSE and other indicators are tracked in real time. When the indicators deteriorate (such as the accuracy is lower than the set threshold), the source of data deviation and model overfitting or underfitting are automatically analyzed back-tracing, and the corresponding tuning strategy (resampling data, updating feature sets, fine-tuning model structure) is triggered according to the problem type (data drift, feature failure, etc.).

[0049] At the same time, the timeliness standards are updated adaptively: new order data (manually reviewed and confirmed) are regularly (weekly or monthly) included in the training set, and the timeliness standard prediction model is updated through incremental learning to adapt to changes such as logistics network expansion and business strategy adjustments (opening of new routes, upgrade of delivery timeliness), ensuring that the timeliness standards are in line with operational dynamics.

[0050] above Figure 3 The full-link collection and signing time monitoring device for logistics orders in the embodiment of the present invention is described in detail from the perspective of modular functional entities. The electronic device in the embodiment of the present invention is described in detail from the perspective of hardware processing.

[0051] Figure 47 is a schematic diagram of the structure of an electronic device provided by an embodiment of the present invention. The electronic device 700 may have relatively large differences due to different configurations or performances, and may include one or more processors (central processing units, CPU) 710 (for example, one or more processors) and a memory 720, and one or more storage media 730 (for example, one or more mass storage devices) storing application programs 733 or data 732. Among them, the memory 720 and the storage medium 730 can be temporary storage or permanent storage. The program stored in the storage medium 730 may include one or more modules (not shown in the figure), and each module may include a series of instruction operations in the electronic device 700. Furthermore, the processor 710 may be configured to communicate with the storage medium 730 to execute a series of instruction operations in the storage medium 730 on the electronic device 700.

[0052] The electronic device 700 may also include one or more power supplies 740, one or more wired or wireless network interfaces 750, one or more input and output interfaces 750, and / or one or more operating systems 731, such as Windows Serve, Mac OS X, Unix, Linux, FreeBSD, etc. It will be appreciated by those skilled in the art that Figure 4 The structure of the electronic device shown does not constitute a limitation on the electronic device, and may include more or less components than shown in the figure, or combine some components, or arrange the components differently.

[0053] The present invention also provides a computer-readable storage medium, which may be a non-volatile computer-readable storage medium or a volatile computer-readable storage medium. Instructions are stored in the computer-readable storage medium. When the instructions are executed on a computer, the computer executes the steps of a method for monitoring the timeliness of a full-link collection and signing of logistics orders.

[0054] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working process of the above-described system, device, or unit can refer to the corresponding process in the aforementioned method embodiment and will not be repeated here.

[0055] If the integrated unit is implemented in the form of 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 is essentially or the part that contributes to the prior art or the whole or part of the technical solution can be embodied in the form of a software product. The computer software product is stored in a storage medium, including several instructions to enable a computer device (which can be a personal computer, server, or network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM), random access memory (RAM), disk or optical disk and other media that can store program code.

[0056] As described above, the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that the technical solutions described in the aforementioned embodiments may still be modified, or some of the technical features thereof may be replaced by equivalents. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for monitoring the timeliness of the whole-link signing of logistics orders, characterized in that: include: Constructing a timeliness standard prediction model, and training the constructed timeliness standard prediction model to obtain a trained timeliness standard prediction model; Use the trained time efficiency standard prediction model to predict the visa collection time; Through the real-time collected order trajectory data, the actual pickup time is corrected according to the anomaly detection mechanism to obtain the corrected actual pickup time. When the corrected actual visa collection time is greater than the predicted visa collection time, the core delay risk feature vector that meets the preset requirements is extracted based on multi-dimensional influencing factors and dimensionality reduction through principal component analysis; Based on the extracted core delay risk feature vector that meets the preset requirements, the corresponding historical orders are labeled, and the early warning classification model is trained using the labeled historical orders to obtain the trained early warning classification model; Use the trained early warning classification model to classify and grade the delay risks of orders, and push the early warning classification model warning results to the client according to the preset requirements, so that the client can formulate corresponding adjustment strategies; The timeliness standard prediction model is a hybrid model constructed by integrating a neural network and a gradient boosting decision tree, and the signing time is predicted based on the constructed hybrid model; The early warning classification model is constructed based on logistic regression and support vector machine.

2. The full-link signing time monitoring method for logistics orders according to claim 1 is characterized in that: The step of constructing a timeliness standard prediction model and training the constructed timeliness standard prediction model to obtain a trained timeliness standard prediction model includes: Construct a timeliness standard prediction model based on neural network and gradient boosting decision tree; Acquire historical logistics order data under different conditions, and preprocess the acquired historical logistics order data to obtain preprocessed historical logistics order data; Construct training and test sets based on preprocessed historical logistics order data; The constructed timeliness standard prediction model is trained with the training set through multiple rounds of back propagation and gradient boosting iterations to obtain the trained timeliness standard prediction model; The test set is used to verify and evaluate the currently trained timeliness standard prediction model through the mean absolute error and root mean square error. When the evaluation fails to meet the preset requirements, the hyperparameters of the currently trained timeliness standard prediction model are adjusted, and the training is triggered repeatedly until the test set evaluation meets the preset requirements.

3. The full-link signing time monitoring method for logistics orders according to claim 1 is characterized in that: The obtaining of historical logistics order data under different conditions and preprocessing the obtained historical logistics order data to obtain the preprocessed historical logistics order data include: Obtain historical logistics order data under different conditions; where different conditions include: different seasons, time periods, shipping and receiving areas, cargo types, and different dimensions of logistics carrier information; the historical logistics order data includes: the longitude and latitude of the shipping city, the longitude and latitude of the receiving city, the order timestamp, the weight and volume of the cargo, the carrier code, and the full-link stipulated pickup time; The historical logistics order data obtained under different conditions are processed including data cleaning, outlier removal, missing value supplementation, categorical variable unique hot encoding and numerical variable normalization to obtain the processed historical logistics order data.

4. The full-link signing time monitoring method for logistics orders according to claim 1 is characterized in that: The order trajectory data collected in real time is used to correct the predicted pickup time according to the anomaly detection mechanism to obtain the corrected pickup time, including: Collect order track data in real time based on the information system of each link of logistics; wherein, the information system of each link of logistics includes: PDA at the collection end, GPS of transportation vehicle, warehouse management WMS and distribution handheld terminal; The Kalman filter algorithm is used to integrate GPS positioning and site scanning time to correct the time error caused by signal drift and equipment delay, and the actual signature collection time is obtained after correction; The isolation forest algorithm is used to identify anomalies in the actual pickup time. The weighted average method is used to correct the current corrected actual pickup time based on the historical average time for the same region and type of goods during the same period and the current logistics congestion index to obtain the corrected actual pickup time.

5. The full-link signing time monitoring method for logistics orders according to claim 1 is characterized in that: When the corrected actual visa collection time is greater than the predicted visa collection time, the core delay risk feature vector that meets the preset requirements is extracted through principal component analysis and dimensionality reduction based on multi-dimensional influencing factors, including: When the corrected actual pickup time is greater than the predicted pickup time, based on multi-dimensional influencing factors including real-time weather data of the logistics process, the full load rate of the transport vehicle and the saturation of the deliveryman's daily tasks, principal component analysis is used to reduce the dimension and extract the core delay risk feature vector that meets the preset requirements.

6. The full-link signing time monitoring method for logistics orders according to claim 1 is characterized in that: The method of labeling corresponding historical orders based on the extracted core delay risk feature vector that meets preset requirements, and using the labeled historical orders to train the early warning classification model to obtain the trained early warning classification model includes: Construct an early warning classification model based on logistic regression and support vector machine; Based on the extracted core delay risk feature vectors that meet the preset requirements, the corresponding historical orders are labeled, and the labeled historical orders are used to train the early warning classification model to obtain the trained early warning classification model; Using the early warning classification model to classify and grade the current delay risk of the order; wherein the classification includes delay or normal; the classification includes mild, moderate or severe delay; The warning results of the warning classification model are pushed to the client according to the preset requirements.

7. The full-link signing time monitoring method for logistics orders according to claim 1 is characterized in that: The method further comprises: Deploy the performance monitoring module of the timeliness standard prediction model to track the prediction in real time, including accuracy, MAE, and RMSE indicators. When the indicators deteriorate, trigger retrospective analysis including the source of data deviation, overfitting or underfitting of the timeliness standard prediction model, and formulate corresponding tuning strategies.

8. A full-link signing time monitoring device for logistics orders, characterized in that: include: Constructing a timeliness standard prediction model module, which is used to construct a timeliness standard prediction model, and train the constructed timeliness standard prediction model to obtain a trained timeliness standard prediction model; The visa collection time prediction module is used to predict the visa collection time using the trained time efficiency standard prediction model; The actual pickup time correction module is used to correct the actual pickup time based on the order trajectory data collected in real time and the anomaly detection mechanism to obtain the corrected actual pickup time; The delay risk extraction module is used to extract the core delay risk feature vector that meets the preset requirements based on multi-dimensional influencing factors and principal component analysis and dimensionality reduction when the actual revised visa collection time is greater than the predicted visa collection time. The warning classification module is a construction module, which is used to mark the corresponding historical orders based on the extracted core delay risk feature vectors that meet the preset requirements, and train the warning classification model using the marked historical orders to obtain the trained warning classification model; The classification warning module is used to classify and grade the delay risks of orders using the trained warning classification model, and push the warning results of the warning classification model to the client according to the preset requirements, so that the client can formulate corresponding adjustment strategies; The timeliness standard prediction model is a hybrid model constructed by integrating a neural network and a gradient boosting decision tree, and the signing time is predicted based on the constructed hybrid model; The early warning classification model is constructed based on logistic regression and support vector machine.

9. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by the processor, the steps of the full-link collection and signing time monitoring method for the logistics order described in any one of claims 1 to 7 are implemented.

10. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the computer program is executed by the processor, the steps of the full-link collection and signing time monitoring method for the logistics order described in any one of claims 1 to 7 are implemented.

Images