Return exception order interception method and device, equipment and storage medium
The method and device automate the detection and interception of abnormal return orders using data preprocessing and machine learning, addressing inefficiencies and errors in manual review, thereby enhancing logistics efficiency and reducing costs.
Patent Information
- Application Number
- CN202411202197.7
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-08-29
- Publication Date
- 2024-12-13
AI Technical Summary
The manual review method for handling abnormal return orders in logistics is inefficient, prone to errors, and increases operating costs due to human factors, affecting customer experience and company reputation.
A method and device using a rule engine, support vector machine model, and data preprocessing techniques to automatically intercept abnormal return orders, involving data cleaning, enhancement, partitioning, and model training for real-time detection and interception.
Automates the identification of abnormal return orders, improving efficiency, reducing unnecessary processing, and saving costs by minimizing errors and enhancing response speed.
Smart Images
Figure CN119130293A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of data processing technology, and in particular to a method, device, equipment and storage medium for intercepting abnormal returned items. Background Art
[0002] In the field of logistics, with the rapid development of e-commerce and the increasing diversification of consumer demand, the complexity and tediousness of cargo transportation, distribution and return processing are increasing day by day; this complexity is not only reflected in the diversity of cargo types, quantities and transportation routes, but also covers the information exchange and coordination of multiple participants such as suppliers, carriers, and end consumers; therefore, a large number of accounting records are generated in the logistics chain, including but not limited to freight settlement, warehousing costs, insurance claims, tariffs and value-added tax processing, etc., each of which requires accurate recording and accounting.
[0003] Faced with such a huge amount of data, the traditional manual review method seems to be inadequate. The manual review method is not only time-consuming and inefficient, but also prone to errors due to human factors, such as data entry errors, calculation errors or deviations in policy understanding. These errors will not only increase the operating costs of logistics companies, but may also damage customer experience and affect the reputation of logistics companies.
[0004] It can be seen that the existing technology still needs to be improved and enhanced. Summary of the invention
[0005] In order to overcome the shortcomings of the prior art, the purpose of the present invention is to provide a method for intercepting abnormal return orders, which can automatically intercept abnormal return orders, improve review efficiency, reduce unnecessary return processing procedures, and save review costs.
[0006] The first aspect of the present invention provides a method for intercepting abnormal return orders, including: obtaining historical return data, preprocessing the historical return data to obtain preprocessed historical data; using a rule engine to perform preliminary screening on the preprocessed historical data to obtain preliminary screened historical data; performing data enhancement processing and partitioning processing on the preliminary screened historical data to obtain a training set, a verification set and a test set; inputting the training set, the verification set and the test set into a model constructed based on a support vector machine for training to obtain an abnormal recognition model; obtaining real-time return data, preprocessing the real-time return data to obtain preprocessed real-time data; inputting the preprocessed real-time data into the abnormal recognition model to determine whether there is an abnormal return order, and if so, intercepting the identified abnormal return order; updating the rule engine and the abnormal recognition model according to a preset update plan.
[0007] Optionally, in a first implementation method of the first aspect of the present invention, the historical return data is obtained, and the historical return data is preprocessed to obtain preprocessed historical data, including: obtaining historical return data, the historical return data including multiple historical return orders; performing missing value processing, deduplication processing and denoising processing on each historical return order to obtain multiple denoised historical return orders; obtaining a preset integrity standard, using regular expressions to judge the integrity of multiple denoised historical return orders, and removing incomplete denoised historical return orders; using the Pandas library to convert the retained denoised historical return orders into a unified format, integrating the denoised historical return orders after the converted format, and obtaining preprocessed historical data.
[0008] Optionally, in a second implementation method of the first aspect of the present invention, the use of a rule engine to perform preliminary screening of the preprocessed historical data to obtain preliminary screened historical data includes: obtaining a preset classification target and a classification rule corresponding to the classification target; based on the acquired classification rule, using a rule engine to perform preliminary screening of the preprocessed historical data, retaining denoised historical return orders that meet the classification rules, and obtaining multiple preliminary screened historical return orders; using the OpenCV library to configure type labels for each preliminary screened historical return order, integrating the preliminary screened historical return orders with multiple configuration labels, and obtaining preliminary screened historical data.
[0009] Optionally, in a third implementation method of the first aspect of the present invention, the data enhancement processing and division processing of the initial screening historical data to obtain a training set, a verification set and a test set includes: using data enhancement technology to perform data enhancement processing on the initial screening historical data to obtain enhanced historical data; randomly shuffling the enhanced historical data, and dividing the enhanced historical data into a training set, a verification set and a test set according to a preset division ratio; verifying the balance and representativeness of the divided training set, verification set and test set by a statistical analysis method, and using the verified training set, verification set and test set for model training.
[0010] Optionally, in a fourth implementation method of the first aspect of the present invention, the training set, validation set and test set are respectively input into a model constructed based on a support vector machine for training to obtain an abnormality recognition model, including: initializing the model constructed based on the support vector machine, setting the initial learning rate of the model, setting the kernel function of the model to a radial basis function, and setting the optimization function of the model to an SGD function; inputting the training set and validation set into the initialized model for training, and during the training process, adjusting the learning rate of the model by a cross-validation method, adjusting the parameters of the kernel function by a grid search method, and optimizing the parameters of the model according to the optimization function; based on the early stopping method, monitoring the performance of the model on the validation set, stopping the training when the performance on the validation set no longer improves, and obtaining a model to be tested; inputting the test set into the model to be tested for testing, and during the testing process, measuring the performance of the model on the test set according to preset evaluation indicators; adjusting the parameters of the model to be tested according to the performance evaluation results to obtain an abnormality recognition model.
[0011] Optionally, in a fifth implementation of the first aspect of the present invention, the real-time return data is obtained, and the real-time return data is preprocessed to obtain preprocessed real-time data, including: obtaining real-time return data, the real-time return data including multiple real-time return orders within a preselected time period; performing data cleaning processing on the multiple real-time return orders respectively to obtain multiple real-time cleaned return orders; using the Pandas library to perform format conversion processing on the multiple real-time cleaned return orders respectively, and integrating the multiple real-time cleaned return orders after format conversion to obtain preprocessed real-time data.
[0012] Optionally, in a sixth implementation method of the first aspect of the present invention, the pre-processed real-time data is input into the abnormal recognition model to determine whether there is an abnormal return order. If so, the identified abnormal return order is intercepted, including: inputting the pre-processed real-time data into the abnormal recognition model, and determining whether there is an abnormal return order based on the recognition result output by the abnormal recognition model; when there is an abnormal return order, intercepting the abnormal return order, and obtaining an abnormal warning text based on the abnormal type of the abnormal return order; filling the order information included in the abnormal return order into the obtained abnormal warning text to obtain the abnormal information content, and generating a warning sending instruction based on the preset sending configuration information and the abnormal information content.
[0013] The second aspect of the present invention provides a device for intercepting abnormal return orders, including: a first processing module, used to obtain historical return data, pre-process the historical return data, and obtain pre-processed historical data; a screening module, used to use a rule engine to perform preliminary screening on the pre-processed historical data, and obtain preliminary screening historical data; an enhancement and partitioning module, used to perform data enhancement processing and partition processing on the preliminary screening historical data, and obtain a training set, a verification set, and a test set; a training module, used to input the training set, the verification set, and the test set into a model constructed based on a support vector machine for training, and obtain an abnormal recognition model; a second processing module, used to obtain real-time return data, pre-process the real-time return data, and obtain pre-processed real-time data; an interception module, used to input the pre-processed real-time data into the abnormal recognition model to determine whether there is an abnormal return order, and if so, intercept the identified abnormal return order; an update module, used to update the rule engine and the abnormal recognition model according to a preset update plan.
[0014] Optionally, in a first implementation method of the second aspect of the present invention, the first processing module includes: a first acquisition unit, used to acquire historical return data, the historical return data including multiple historical return orders; a first processing unit, used to perform missing value processing, deduplication processing and denoising processing on each historical return order, to obtain multiple denoised historical return orders; a removal unit, used to obtain a preset integrity standard, use regular expressions to judge the integrity of multiple denoised historical return orders, and remove incomplete denoised historical return orders; a first conversion unit, used to use the Pandas library to convert the retained denoised historical return orders into a unified format, integrate the denoised historical return orders after the conversion format, and obtain preprocessed historical data.
[0015] Optionally, in a second implementation of the second aspect of the present invention, the screening module includes: a second acquisition unit, used to acquire a preset classification target and a classification rule corresponding to the classification target; a screening unit, used to use a rule engine to perform preliminary screening of the preprocessed historical data based on the acquired classification rules, retain the denoised historical return orders that meet the classification rules, and obtain multiple preliminary screened historical return orders; a configuration unit, used to use the OpenCV library to configure type labels for each preliminary screened historical return order, integrate the preliminary screened historical return orders with multiple configuration labels, and obtain preliminary screened historical data.
[0016] Optionally, in a third implementation of the second aspect of the present invention, the enhanced partitioning module includes: an enhancement unit, used to perform data enhancement processing on the initial screening historical data using data enhancement technology to obtain enhanced historical data; a partitioning unit, used to randomly shuffle the enhanced historical data, and divide the enhanced historical data into a training set, a verification set, and a test set according to a preset partitioning ratio; a verification unit, used to verify the balance and representativeness of the divided training set, verification set, and test set through a statistical analysis method, and use the verified training set, verification set, and test set for model training.
[0017] Optionally, in a fourth implementation method of the second aspect of the present invention, the training module includes: an initial unit, used to initialize a model built based on a support vector machine, set the initial learning rate of the model, set the kernel function of the model to a radial basis function, and set the optimization function of the model to an SGD function; a training unit, used to input a training set and a validation set into the initialized model for training, during the training process, adjusting the learning rate of the model by a cross-validation method, adjusting the parameters of the kernel function by a grid search method, and optimizing the parameters of the model according to the optimization function; a verification unit, used to monitor the performance of the model on the validation set based on an early stopping method, stopping training when the performance on the validation set no longer improves, and obtaining a model to be tested; a testing unit, used to input a test set into the model to be tested for testing, during the testing process, measuring the performance of the model on the test set according to preset evaluation indicators; an adjustment unit, used to adjust the parameters of the model to be tested according to the performance evaluation results, and obtain an abnormality recognition model.
[0018] Optionally, in a fifth implementation of the second aspect of the present invention, the second processing module includes: a third acquisition unit, used to acquire real-time return data, the real-time return data including multiple real-time return orders within a preselected time period; a second processing unit, used to perform data cleaning processing on the multiple real-time return orders respectively to obtain multiple real-time cleaning return orders; a second conversion unit, used to use the Pandas library to perform format conversion processing on the multiple real-time cleaning return orders respectively, and integrate the multiple real-time cleaning return orders after format conversion to obtain preprocessed real-time data.
[0019] Optionally, in a sixth implementation of the second aspect of the present invention, the interception module includes: an identification unit, used to input the pre-processed real-time data into an abnormal identification model, and determine whether there is an abnormal return order based on the identification result output by the abnormal identification model; an interception unit, used to intercept the abnormal return order when there is an abnormal return order, and obtain an abnormal warning text based on the abnormal type of the abnormal return order; an early warning unit, used to fill the order information included in the abnormal return order into the obtained abnormal warning text, obtain the abnormal information content, and generate an early warning sending instruction based on the preset sending configuration information and the abnormal information content.
[0020] The third aspect of the present invention provides a device for intercepting abnormal return orders, which includes: a memory and at least one processor, wherein the memory stores instructions; at least one of the processors calls the instructions in the memory to enable the device to execute each step of the abnormal return order interception method described above.
[0021] A fourth aspect of the present invention provides a computer-readable storage medium, on which instructions are stored. When the instructions are executed by a processor, the various steps of the method for intercepting abnormal return orders described in any one of the above items are implemented.
[0022] In the technical solution of the present invention, historical return data is obtained for preprocessing and preliminary screening to obtain preliminary screening historical data; data enhancement processing and partitioning processing are performed on the preliminary screening historical data to obtain a training set, a validation set and a test set; the training set, the validation set and the test set are respectively input into a model built based on a support vector machine for training to obtain an abnormality recognition model; real-time return data is obtained and preprocessed, and the preprocessed real-time data is input into the abnormality recognition model to determine whether there is an abnormal return order, and if so, the identified abnormal return order is intercepted; according to a preset update plan, the rule engine and the abnormality recognition model are updated; the method disclosed in the present application can automatically intercept abnormal return orders, improve audit efficiency, reduce unnecessary return processing procedures, and save audit costs. BRIEF DESCRIPTION OF THE DRAWINGS
[0023] Figure 1 A first flow chart of a data processing method provided by an embodiment of the present invention;
[0024] Figure 2 A second flow chart of the data processing method provided by an embodiment of the present invention;
[0025] Figure 3 A third flow chart of the data processing method provided by an embodiment of the present invention;
[0026] Figure 4 A fourth flow chart of the data processing method provided by an embodiment of the present invention;
[0027] Figure 5 A fifth flow chart of the data processing method provided by the embodiment of the present invention;
[0028] Figure 6 A sixth flow chart of the data processing method provided by an embodiment of the present invention;
[0029] Figure 7 A seventh flow chart of the data processing method provided by the embodiment of the present invention;
[0030] Figure 8 A schematic diagram of the structure of a data processing device provided by an embodiment of the present invention;
[0031] Fig. 9 A schematic diagram of the structure of a data processing device provided in an embodiment of the present invention. DETAILED DESCRIPTION
[0032] The present invention provides a method, device, equipment and storage medium for intercepting abnormal return orders. In the present invention, 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 interchangeable 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.
[0033] For ease of understanding, the specific process of the embodiment of the present invention is described below. Figure 1 An embodiment of the method for intercepting abnormal return orders in an embodiment of the present invention includes:
[0034] 101. Obtain historical return data, pre-process the historical return data, and obtain pre-processed historical data;
[0035] In this embodiment, by collecting and preprocessing historical return data, the basic data quality for subsequent analysis is ensured, providing reliable data support for subsequent steps.
[0036] 102. Use the rule engine to preliminarily screen the preprocessed historical data to obtain the preliminarily screened historical data;
[0037] In this embodiment, a rule engine is used to perform preliminary screening on the preprocessed historical data, effectively eliminating invalid data or data that does not meet specific conditions, thereby improving the efficiency and accuracy of data processing.
[0038] 103. Perform data enhancement and partitioning on the initial screening historical data to obtain a training set, a validation set, and a test set;
[0039] In this embodiment, sufficient training sets, validation sets, and test sets are obtained through data enhancement and partitioning processing, ensuring that the anomaly recognition model built based on the support vector machine can be fully trained and verified.
[0040] 104. Input the training set, the validation set, and the test set into the model constructed based on the support vector machine for training respectively, so as to obtain an anomaly recognition model;
[0041] 105. Acquire real-time returned data, pre-process the real-time returned data, and obtain pre-processed real-time data;
[0042] In this embodiment, the pre-processed real-time return data is input into the abnormality recognition model, which realizes the real-time detection and interception of abnormal return orders, and effectively prevents potential risks and problems.
[0043] 106. Input the pre-processed real-time data into the abnormality recognition model to determine whether there is an abnormal return order. If so, intercept the identified abnormal return order;
[0044] 107. Update the rule engine and the anomaly recognition model according to a preset update plan;
[0045] In this embodiment, by setting an update plan, the model can be continuously updated, thereby ensuring the adaptability and accuracy of the anomaly recognition model.
[0046] In this embodiment, the update plan can be updated once on the last day of each month; before updating, the currently running rule engine and anomaly recognition model are backed up so that they can be quickly restored to a stable state when problems arise during the update process; when updating, the update content is obtained, and the update content includes the latest version of the rule engine and anomaly recognition model code or configuration file; based on the preset test verification process and the obtained update content, the rule engine and anomaly recognition model are updated, the updated rule engine and anomaly recognition model are deployed to the interception system, and the updated system performance and anomaly recognition accuracy are monitored.
[0047] The present application discloses a method for intercepting abnormal return orders, which comprises the following steps: obtaining historical return data for preprocessing and preliminary screening to obtain preliminary screening historical data; performing data enhancement processing and partitioning processing on the preliminary screening historical data to obtain a training set, a validation set and a test set; inputting the training set, the validation set and the test set into a model constructed based on a support vector machine for training to obtain an abnormal recognition model; obtaining real-time return data and performing preprocessing, and inputting the preprocessed real-time data into the abnormal recognition model to determine whether there is an abnormal return order, and if so, intercepting the identified abnormal return order; updating the rule engine and the abnormal recognition model according to a preset update plan; the method disclosed in the present application can realize audit automation, and can automatically intercept abnormal return orders, improve audit efficiency, reduce the occurrence rate of erroneous payments, and can reduce unnecessary return processing procedures, saving audit costs.
[0048] See also Figure 2 The second embodiment of the method for intercepting abnormal returned documents in the embodiment of the present invention includes:
[0049] 201. Obtain historical return data, where the historical return data includes multiple historical return orders;
[0050] 202. Perform missing value processing, duplicate removal processing, and denoising processing on each historical return order to obtain multiple denoised historical return orders;
[0051] In this embodiment, missing values in historical return orders can be processed by filling missing values; then, duplicate order records can be deleted to deduplicate historical return orders; finally, statistical methods are used to identify outliers in historical return orders to denoise historical return orders.
[0052] 203. Obtain a preset integrity standard, use regular expressions to determine the integrity of multiple denoised historical return orders, and remove incomplete denoised historical return orders;
[0053] In this embodiment, the integrity standard can be used to confirm whether fields such as order number, customer information, and return reason are complete; according to the integrity standard, a corresponding regular expression is written for each field that must exist, and these expressions will be used to match and verify the data in the order; all denoised historical return orders are traversed, and regular expressions are applied one by one for matching checks. If a field fails to match the regular expression, it means that the order is incomplete in this field, and the matched incomplete denoised historical return orders are removed.
[0054] 204. Use the Pandas library to convert the retained denoised historical return orders into a unified format, integrate the denoised historical return orders after the conversion format, and obtain the preprocessed historical data;
[0055] In this embodiment, historical return orders may be uniformly converted into JSON format.
[0056] See also Figure 3 A third embodiment of the method for intercepting abnormal returned documents in the embodiment of the present invention includes:
[0057] 301. Obtain a preset classification target and a classification rule corresponding to the classification target;
[0058] 302. Based on the acquired classification rules, a rule engine is used to preliminarily screen the preprocessed historical data, and denoised historical return orders that meet the classification rules are retained to obtain a plurality of preliminarily screened historical return orders;
[0059] In this embodiment, the defined classification rules are input into the rule engine, which specifically involves writing rule scripts, using the graphical interface of the rule engine, or configuring rules through an API; the rule engine is run to match and filter the pre-processed historical return order data. The rule engine will automatically identify and retain the denoised historical return orders that meet the classification conditions according to the set rules.
[0060] 303. Use the OpenCV library to configure type labels for each of the initial screening historical return orders, integrate the initial screening historical return orders after multiple configuration labels, and obtain the initial screening historical data;
[0061] In this embodiment, natural language processing technology is first used to extract feature keywords from each initially screened historical return order, and then the initially screened historical return orders are classified based on the mentioned feature keywords and a pre-built classifier, and the classification results are attached to each initially screened historical return order as type labels; the type labels include normal return and abnormal return. When the label is abnormal return, the corresponding abnormal reasons are also configured, such as "multiple returns to the same address", "frequent returns of high-value items", etc.
[0062] See also Figure 4 The fourth embodiment of the method for intercepting abnormal returned documents in the embodiment of the present invention includes:
[0063] 401. Perform data enhancement processing on the initial screening historical data using data enhancement technology to obtain enhanced historical data;
[0064] In this embodiment, the data enhancement technology of noise addition and data synthesis is combined to generate new samples.
[0065] 402. Randomly shuffle the enhanced historical data, and divide the enhanced historical data into a training set, a validation set, and a test set according to a preset division ratio;
[0066] In this embodiment, the `random.shuffle()` function is used to randomly shuffle the enhanced historical data, and then 70% of the data is used as a training set, 15% as a validation set, and 15% as a test set. According to the determined ratio, the amount of data that each set should contain is calculated; using the randomly shuffled data set, the training set, validation set, and test set are divided in sequence from the beginning of the enhanced historical data according to the calculated data amount.
[0067] 403. Verify the balance and representativeness of the divided training set, validation set, and test set by statistical analysis methods, and use the verified training set, validation set, and test set for model training;
[0068] In this embodiment, a descriptive statistical analysis is performed on each data set, including calculating the number of samples in each category. Then, the chi-square test method is used to evaluate whether there are significant differences in the distribution of each category in different data sets (training set, validation set, test set). If the difference is not significant, it means that the data set division has a good balance. A correlation analysis is performed on the features in the data set to ensure that the feature distributions in the training set, validation set, and test set are similar, so as to ensure that the performance of the model on different data sets is consistent.
[0069] See also Figure 5 The fifth embodiment of the method for intercepting abnormal returned documents in the embodiment of the present invention includes:
[0070] 501. Initialize the model built based on the support vector machine, set the initial learning rate of the model, set the kernel function of the model to the radial basis function, and set the optimization function of the model to the SGD function;
[0071] In this embodiment, the kernel function is selected as the Radial Basis Function (RBF), which provides great flexibility. The RBF kernel can map data to a higher-dimensional space, thereby solving nonlinear problems. This is particularly useful when processing complex data patterns. The RBF kernel function has good generalization ability and can accurately predict unknown data. The optimization function adopts the Stochastic Gradient Descent (SGD) method, which makes the model more efficient when training on large-scale data sets. SGD updates the model parameters by using only one or a small batch of samples each time, thereby speeding up the convergence speed and reducing memory consumption.
[0072] 502. Input the training set and the validation set into the initialized model for training. During the training process, adjust the learning rate of the model by a cross-validation method, adjust the parameters of the kernel function by a grid search method, and optimize the parameters of the model according to the optimization function.
[0073] In this embodiment, using the cross-validation method to adjust the learning rate helps to find the optimal learning rate, avoiding too fast a learning rate that causes the model to fail to converge, or too slow a learning rate that causes the training time to be too long; the grid search method adjusts the kernel function parameters to ensure that the model can find the optimal decision boundary in the feature space; finally, optimizing the model parameters according to the optimization function can further improve the accuracy and efficiency of the model; through the combined effect of two adjustment steps and one optimization step, the model is continuously optimized during the training process, and ultimately achieves better prediction results.
[0074] 503. Based on the early stopping method, the performance of the model is monitored on the validation set. When the performance on the validation set no longer improves, the training is stopped to obtain the model to be tested.
[0075] 504. Input the test set into the model to be tested for testing. During the testing process, the performance of the model on the test set is measured according to a preset evaluation index.
[0076] In this embodiment, the evaluation indicators may include accuracy and mean square error.
[0077] 505. Adjust the parameters of the model to be tested according to the performance evaluation result to obtain an anomaly recognition model.
[0078] See also Figure 6 The sixth embodiment of the method for intercepting abnormal returned documents in the embodiment of the present invention includes:
[0079] 601. Acquire real-time return data, where the real-time return data includes a plurality of real-time return orders within a preselected time period;
[0080] In this embodiment, the preselected time period may be the previous day, that is, the acquired real-time return data is a combination of all real-time return orders received the previous day.
[0081] 602. Perform data cleaning on the multiple real-time return orders respectively to obtain multiple real-time cleaned return orders;
[0082] In this embodiment, the missing values in the real-time return order are processed by filling the missing values; then the real-time return order is deduplicated by deleting duplicate order records; finally, a statistical method is used to identify outliers in the real-time return order to denoise the real-time return order; after the missing value processing, deduplication processing and denoising processing, the data of the real-time return order is completely cleaned.
[0083] 603. Use the Pandas library to perform format conversion processing on multiple real-time cleaning and return orders respectively, integrate the real-time cleaning and return orders after multiple format conversions, and obtain pre-processed real-time data;
[0084] In this embodiment, the real-time cleaning return order can be converted into JSON format.
[0085] See also Figure 7 The seventh embodiment of the method for intercepting abnormal returned documents in the embodiment of the present invention includes:
[0086] 701. Input the pre-processed real-time data into the abnormality recognition model, and determine whether there is an abnormal return order according to the recognition result output by the abnormality recognition model;
[0087] The abnormality identification result includes the type of order, which includes normal and abnormal. When the type is abnormal, it also includes the reason for the abnormality.
[0088] 702. When there is an abnormal return order, intercept the abnormal return order, and obtain the abnormal warning text according to the abnormal type of the abnormal return order;
[0089] In this embodiment, the abnormal warning text contains necessary placeholders so that specific order information can be inserted during the actual warning.
[0090] 703. Fill the order information included in the abnormal return order into the acquired abnormal warning text to obtain the abnormal information content, and generate a warning sending instruction based on the preset sending configuration information and the abnormal information content;
[0091] In this embodiment, the detailed information of the abnormal return order (such as order number, customer information, return reason, etc.) is filled into the corresponding position of the warning text template to form a complete abnormal information content; according to the preset sending configuration information (including recipient list, sending channel, sending time, etc.), the system will generate a warning sending instruction; finally, the warning sending instruction is executed to send the abnormal information content to relevant managers or service personnel via email, SMS, application push or other communication methods, thereby improving the response speed and accuracy of handling abnormal return orders.
[0092] The above describes the method for intercepting abnormal return orders in the embodiment of the present invention. The following describes the device for intercepting abnormal return orders in the embodiment of the present invention. Figure 8 , an embodiment of the device for intercepting abnormal return orders in an embodiment of the present invention includes:
[0093] The first processing module 801 is used to obtain historical return data, pre-process the historical return data, and obtain pre-processed historical data; the screening module 802 is used to use the rule engine to perform preliminary screening on the pre-processed historical data to obtain the initial screening historical data; the enhancement and division module 803 is used to perform data enhancement processing and division processing on the initial screening historical data to obtain a training set, a verification set and a test set; the training module 804 is used to input the training set, the verification set and the test set into a model built based on a support vector machine for training to obtain an anomaly recognition model; the second processing module 805 is used to obtain real-time return data, pre-process the real-time return data, and obtain pre-processed real-time data; the interception module 806 is used to input the pre-processed real-time data into the anomaly recognition model to determine whether there is an abnormal return order. If so, intercept the identified abnormal return order; the update module 807 is used to update the rule engine and the anomaly recognition model according to a preset update plan.
[0094] In this embodiment, the first processing module 801 includes: a first acquisition unit 8011, which is used to acquire historical return data, and the historical return data includes multiple historical return orders; a first processing unit 8012, which is used to perform missing value processing, deduplication processing and denoising processing on each historical return order, so as to obtain multiple denoised historical return orders; a removal unit 8013, which is used to obtain a preset integrity standard, use regular expressions to judge the integrity of multiple denoised historical return orders, and remove incomplete denoised historical return orders; a first conversion unit 8014, which is used to use the Pandas library to convert the retained denoised historical return orders into a unified format, integrate the denoised historical return orders after the conversion format, and obtain preprocessed historical data.
[0095] In this embodiment, the screening module 802 includes: a second acquisition unit 8021, which is used to obtain a preset classification target and a classification rule corresponding to the classification target; a screening unit 8022, which is used to use a rule engine to perform preliminary screening on the pre-processed historical data based on the acquired classification rules, retain the denoised historical return orders that meet the classification rules, and obtain multiple preliminary screened historical return orders; a configuration unit 8023, which is used to use the OpenCV library to configure type labels for each preliminary screened historical return order, integrate the preliminary screened historical return orders with multiple configuration labels, and obtain preliminary screened historical data.
[0096] In this embodiment, the enhanced partitioning module 803 includes: an enhancement unit 8031, which is used to perform data enhancement processing on the initial screening historical data using data enhancement technology to obtain enhanced historical data; a partitioning unit 8032, which is used to randomly shuffle the enhanced historical data, and divide the enhanced historical data into a training set, a validation set, and a test set according to a preset partitioning ratio; a verification unit 8033, which is used to verify the balance and representativeness of the divided training set, validation set, and test set through a statistical analysis method, and use the verified training set, validation set, and test set for model training.
[0097] In this embodiment, the training module 804 includes: an initial unit 8041, which is used to initialize the model built based on the support vector machine, set the initial learning rate of the model, set the kernel function of the model to the radial basis function, and set the optimization function of the model to the SGD function; a training unit 8042, which is used to input the training set and the verification set into the initialized model for training. During the training process, the learning rate of the model is adjusted by the cross-validation method, the parameters of the kernel function are adjusted by the grid search method, and the parameters of the model are optimized according to the optimization function; a verification unit 8043, which is used to monitor the performance of the model on the verification set based on the early stopping method, and stop the training when the performance on the verification set no longer improves to obtain a model to be tested; a testing unit 8044, which is used to input the test set into the model to be tested for testing. During the testing process, the performance of the model on the test set is measured according to a preset evaluation indicator; an adjustment unit 8045, which is used to adjust the parameters of the model to be tested according to the performance evaluation results to obtain an abnormality recognition model.
[0098] In this embodiment, the second processing module 805 includes: a third acquisition unit 8051, which is used to obtain real-time return data, and the real-time return data includes multiple real-time return orders within a preselected time period; a second processing unit 8052, which is used to perform data cleaning processing on multiple real-time return orders respectively to obtain multiple real-time cleaning return orders; a second conversion unit 8053, which is used to use the Pandas library to perform format conversion processing on multiple real-time cleaning return orders respectively, and integrate the real-time cleaning return orders after multiple format conversion to obtain preprocessed real-time data.
[0099] In this embodiment, the interception module 806 includes: an identification unit 8061, which is used to input the pre-processed real-time data into the abnormal identification model, and determine whether there is an abnormal return order according to the identification result output by the abnormal identification model; an interception unit 8062, which is used to intercept the abnormal return order when there is an abnormal return order, and obtain an abnormal warning text according to the abnormal type of the abnormal return order; an early warning unit 8063, which is used to fill the order information included in the abnormal return order into the obtained abnormal warning text, obtain the abnormal information content, and generate an early warning sending instruction based on the preset sending configuration information and the abnormal information content.
[0100] above Figure 8 The abnormal return order interception device in the embodiment of the present invention is described in detail from the perspective of modular functional entities. The abnormal return order interception device in the embodiment of the present invention is described in detail from the perspective of hardware processing.
[0101] Fig. 9 It is a structural schematic diagram of a return exception interception device provided in an embodiment of the present invention. The return exception interception device 900 may have relatively large differences due to different configurations or performances, and may include one or more processors (central processing units, CPU) 910 (for example, one or more processors) and a memory 920, and one or more storage media 930 (for example, one or more mass storage devices) storing application programs 933 or data 932. Among them, the memory 920 and the storage medium 930 can be short-term storage or permanent storage. The program stored in the storage medium 930 may include one or more modules (not shown in the figure), and each module may include a series of instruction operations in the return exception interception device 900. Furthermore, the processor 910 can be configured to communicate with the storage medium 930, and execute a series of instruction operations in the storage medium 930 on the return exception interception device 900 to implement the steps of the return exception interception method provided in the above-mentioned method embodiments.
[0102] The abnormal return interception device 900 may also include one or more power supplies 940, one or more wired or wireless network interfaces 950, one or more input and output interfaces 960, and / or one or more operating systems 931, such as Windows Server, Mac OS X, Unix, Linux, FreeBSD, etc. Those skilled in the art will appreciate that Fig. 9 The structure of the abnormal return order interception device shown does not constitute a limitation of the abnormal return order interception device, and may include more or fewer components than shown in the figure, or a combination of certain components, or a different arrangement of components.
[0103] 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 the method for intercepting abnormal return orders.
[0104] 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.
[0105] 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 all 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 a number of instructions to enable a computer device (which can be a personal computer, a server, or a 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 codes.
[0106] Finally, it should be noted that the above description is only a preferred example of the present invention and is not intended to limit the present invention. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art can still modify the technical solutions described in the aforementioned embodiments or replace some of the technical features therein by equivalents. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.
Claims
1. A method for intercepting abnormal returned items, characterized in that: include: Obtain historical return data, pre-process the historical return data, and obtain pre-processed historical data; Use the rule engine to preliminarily screen the preprocessed historical data to obtain the pre-screened historical data; Perform data enhancement and partitioning on the historical data of initial screening to obtain training sets, validation sets, and test sets; The training set, validation set, and test set are respectively input into the model built based on the support vector machine for training to obtain an anomaly recognition model; Acquire real-time return data, pre-process the real-time return data, and obtain pre-processed real-time data; Input the pre-processed real-time data into the anomaly recognition model to determine whether there are abnormal return orders. If so, intercept the identified abnormal return orders; According to a preset update plan, the rule engine and the anomaly recognition model are updated.
2. The method for intercepting abnormal returned items according to claim 1, characterized in that: The obtaining of historical return data and preprocessing of the historical return data to obtain preprocessed historical data includes: Acquire historical return data, where the historical return data includes multiple historical return orders; Perform missing value processing, duplicate removal processing and denoising processing on each historical return order to obtain multiple denoised historical return orders; Obtain the preset integrity standard, use regular expressions to judge the integrity of multiple denoised historical return orders respectively, and remove incomplete denoised historical return orders; The Pandas library is used to convert the retained denoised historical return orders into a unified format, and the denoised historical return orders after the conversion format are integrated to obtain the preprocessed historical data.
3. The method for intercepting abnormal return orders according to claim 2, characterized in that: The rule engine is used to preliminarily screen the preprocessed historical data to obtain the preliminarily screened historical data, including: Obtaining a preset classification target and a classification rule corresponding to the classification target; Based on the obtained classification rules, the rule engine is used to preliminarily screen the preprocessed historical data, retain the denoised historical return orders that meet the classification rules, and obtain multiple preliminarily screened historical return orders; The OpenCV library is used to configure type labels for each historical return order of the initial screening, and the historical return orders of the initial screening with multiple configuration labels are integrated to obtain the initial screening historical data.
4. The method for intercepting abnormal returned items according to claim 1, characterized in that: The data enhancement and division processing is performed on the initial screening historical data to obtain a training set, a validation set and a test set, including: The data enhancement technology is used to perform data enhancement processing on the initial screening historical data to obtain enhanced historical data; The enhanced historical data is randomly shuffled and divided into a training set, a validation set, and a test set according to a preset division ratio; The balance and representativeness of the divided training set, validation set, and test set are verified by statistical analysis methods, and the verified training set, validation set, and test set are used for model training.
5. The method for intercepting abnormal returned items according to claim 1, characterized in that: The training set, the validation set and the test set are respectively input into the model constructed based on the support vector machine for training to obtain an abnormality recognition model, including: Initialize the model built based on the support vector machine, set the initial learning rate of the model, set the kernel function of the model to the radial basis function, and set the optimization function of the model to the SGD function; The training set and validation set are input into the initialized model for training. During the training process, the learning rate of the model is adjusted by the cross-validation method, the parameters of the kernel function are adjusted by the grid search method, and the parameters of the model are optimized according to the optimization function. Based on the early stopping method, the performance of the model is monitored on the validation set. When the performance on the validation set no longer improves, the training is stopped to obtain the model to be tested. The test set is input into the model to be tested for testing. During the testing process, the performance of the model on the test set is measured according to the preset evaluation indicators; The parameters of the model to be tested are adjusted according to the performance evaluation results to obtain the anomaly recognition model.
6. The method for intercepting abnormal returned items according to claim 1, characterized in that: The step of obtaining the real-time returned data and preprocessing the real-time returned data to obtain preprocessed real-time data includes: Acquire real-time return data, where the real-time return data includes a plurality of real-time return orders within a preselected time period; Perform data cleaning on multiple real-time return orders respectively to obtain multiple real-time cleaned return orders; The Pandas library is used to perform format conversion on multiple real-time cleaning and return orders respectively, and the real-time cleaning and return orders after multiple format conversions are integrated to obtain pre-processed real-time data.
7. The method for intercepting abnormal returned items according to claim 1, characterized in that: The pre-processed real-time data is input into the abnormality identification model to determine whether there is an abnormal return order. If so, intercepting the identified abnormal return order includes: Input the pre-processed real-time data into the anomaly recognition model, and determine whether there is an abnormal return order based on the recognition result output by the anomaly recognition model; When there is an abnormal return order, intercept the abnormal return order and obtain the abnormal warning text according to the abnormal type of the abnormal return order; The order information included in the abnormal return order is filled into the acquired abnormal warning text to obtain the abnormal information content, and a warning sending instruction is generated based on the preset sending configuration information and the abnormal information content.
8. A device for intercepting abnormal returned items, characterized in that: include: The first processing module is used to obtain historical return data, pre-process the historical return data, and obtain pre-processed historical data; A screening module is used to perform preliminary screening on the preprocessed historical data using a rule engine to obtain preliminary screened historical data; The enhancement and partitioning module is used to perform data enhancement and partitioning processing on the initial screening historical data to obtain the training set, validation set and test set; A training module is used to input the training set, the validation set and the test set into the model built based on the support vector machine for training, so as to obtain an anomaly recognition model; The second processing module is used to obtain real-time returned data and pre-process the real-time returned data to obtain pre-processed real-time data; The interception module is used to input the pre-processed real-time data into the abnormality recognition model to determine whether there are abnormal return orders. If so, the identified abnormal return orders are intercepted; An update module is used to update the rule engine and the anomaly recognition model according to a preset update plan.
9. A device for intercepting abnormal return orders, characterized in that: The abnormal return order interception device comprises: a memory and at least one processor, wherein the memory stores instructions; At least one of the processors calls the instructions in the memory to enable the abnormal return interception device to execute each step of the abnormal return interception method as described in any one of claims 1-7.
10. A computer-readable storage medium having instructions stored thereon, characterized in that: When the instruction is executed by the processor, each step of the method for intercepting abnormal return orders as described in any one of claims 1 to 7 is implemented.