Headless parcel prediction method and device, equipment and storage medium

The BP neural network model predicts and optimizes headless packages, solves the problem of headless package identification, improves the processing efficiency and service quality of logistics companies, and reduces operating costs.

CN119067536APending Publication Date: 2024-12-03SHANGHAI DONGPU INFORMATION TECH CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202411015105.4
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-07-26
Publication Date
2024-12-03

AI Technical Summary

Technical Problem

The headless package cannot be identified during express delivery due to label damage, information entry errors or sorting errors, resulting in a large amount of manpower and time for manual inquiry and verification, increasing operating costs and affecting logistics efficiency and service quality.

Method used

Through the headless package prediction method based on the BP neural network model, historical data is collected, key features are extracted, prediction model is trained, and real-time headless package is predicted, prediction results are output, and model parameters and adjustments are made according to the review results. Weight, improve prediction accuracy.

Benefits of technology

It realizes rapid response and efficient processing of headless packages, improves the service quality and customer satisfaction of logistics companies, reduces operating costs, and improves prediction accuracy and generalization capabilities through model optimization.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119067536A_ABST
    Figure CN119067536A_ABST
Patent Text Reader

Abstract

The invention relates to the field of logistics, and discloses a headless parcel prediction method, device and equipment and a storage medium, and the method is used for predicting a headless parcel through a headless parcel prediction model based on a BP neural network model, thereby improving the processing efficiency of the headless parcel. The method comprises the following steps: collecting historical headless parcels, and obtaining logistics information corresponding to the historical headless parcels; the processed headless parcels are screened out, and key features of the processed headless parcels are extracted; training a preset BP neural network model based on the extracted key features to obtain a headless parcel prediction model; acquiring a real-time headless parcel, and inputting the real-time headless parcel into the headless parcel prediction model to acquire a prediction result output by the headless parcel prediction model; and outputting a prediction result, obtaining a recheck result corresponding to the prediction result, and adjusting parameters and weights of the headless parcel prediction model according to the recheck result.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the field of logistics technology, and in particular to a method, device, equipment and storage medium for predicting headerless packages. Background Art

[0002] The problem of headerless packages, that is, the problem that packages cannot be identified during express delivery due to damaged labels, incorrect information entry or sorting errors, has brought many challenges to logistics companies.

[0003] Label damage: Express parcels may undergo multiple handling, loading and unloading and stacking during transportation, which may cause the parcel labels to wear, tear or fall off. Once the label is damaged, key information on the parcel, such as the recipient's name, address, contact information, etc., will be lost, resulting in inaccurate identification and delivery.

[0004] Information entry errors: During the collection, sorting and transshipment of express parcels, staff may enter information incorrectly due to improper operation or negligence. These errors may include incorrect recipient information, unknown or incorrect addresses, invalid contact information, etc., further increasing the risk of missing parcels.

[0005] Package missorting problem: During the express sorting process, the failure of sorting equipment or operational errors may cause packages to be mistakenly assigned to the wrong route or area, making these packages unable to be correctly identified during the subsequent delivery process.

[0006] The problem of missing packages forces logistics companies to invest a lot of manpower and time in manual inquiry and verification. Staff need to find the original information of the package through various channels and contact the recipient to confirm the delivery address and contact information. These cumbersome operations not only increase the operating costs of the company, but also seriously affect the logistics efficiency and service quality.

[0007] Therefore, the existing technology still needs to be improved and developed. Summary of the invention

[0008] The present invention provides a method, device, equipment and storage medium for predicting headless packages, which are used to predict headless packages through a headless package prediction model based on a BP neural network model, thereby improving the processing efficiency of headless packages.

[0009] A first aspect of the present invention provides a method for predicting a headless package, the method comprising: collecting historical headless packages and obtaining logistics information corresponding to the historical headless packages; based on the logistics information corresponding to the historical headless packages, screening out processed headless packages and extracting key features of the processed headless packages; training a preset BP neural network model based on the extracted key features to obtain a headless package prediction model; obtaining a real-time headless package, inputting the real-time headless package into the headless package prediction model to obtain a prediction result output by the headless package prediction model; outputting the prediction result, obtaining a review result corresponding to the prediction result, and adjusting the parameters and weights of the headless package prediction model according to the review result.

[0010] Optionally, in a first implementation of the first aspect of the present invention, collecting historical headerless packages and obtaining logistics information corresponding to the historical headerless packages includes:

[0011] Obtain historical abnormal packages in the order system, and filter out historical headerless packages from the historical abnormal packages; obtain order information, transportation information and abnormal information corresponding to the historical headerless packages according to the codes of the historical headerless packages; clean the acquired order information, transportation information and abnormal information, and integrate them into logistics information.

[0012] Optionally, in a second implementation method of the first aspect of the present invention, the processed headless packages are screened out based on the logistics information corresponding to the historical headless packages, and key features of the processed headless packages are extracted, including: determining the status of the historical headless packages based on the logistics information corresponding to the historical headless packages, the status of the historical headless packages including processed and pending; screening out the headless packages with a processed status, and parsing the logistics information of the processed headless packages to obtain multiple features of the headless packages; and using regular expressions to extract the key features of the headless packages from the multiple features.

[0013] Optionally, in a third implementation method of the first aspect of the present invention, the preset BP neural network model is trained based on the extracted key features to obtain a headerless package prediction model, including: constructing a preset BP neural network model; forming training samples based on the extracted key features, and marking the receiving address and recipient information of each training sample; using the training samples to train the preset BP neural network model to obtain a headerless package prediction model.

[0014] Optionally, in a fourth implementation method of the first aspect of the present invention, constructing a preset BP neural network model includes: using an input layer, a first hidden layer, a second hidden layer and an output layer to construct a basic architecture of the BP neural network model; determining that the activation function of the first hidden layer and the second hidden layer is a nonlinear activation function, and determining that the activation function of the output layer is a Softmax activation function; assigning initial values ​​to each connection weight and bias in the BP neural network model, and initializing the BP neural network model to obtain a preset neural network model.

[0015] Optionally, in a fifth implementation method of the first aspect of the present invention, the real-time headless package is obtained, and the real-time headless package is input into the headless package prediction model to obtain the prediction result output by the headless package prediction model, including: obtaining the real-time headless package and associated information; denoising the associated information to obtain denoised information; inputting the denoised information into the headless package prediction model to obtain the prediction result output by the headless package prediction model, and the prediction result includes the predicted receiving address and recipient information.

[0016] Optionally, in a sixth implementation method of the first aspect of the present invention, the outputting of the prediction result, obtaining a review result corresponding to the prediction result, and adjusting the parameters and weights of the headless package prediction model according to the review result include: outputting the prediction result, and obtaining a review result corresponding to the prediction result; comparing and analyzing the prediction result and the review result, and calculating the prediction error; analyzing the prediction error, and determining the model parameters and weights that need to be adjusted based on the analysis result, and using a back propagation algorithm to adjust the parameters and weights of the headless package prediction model according to the prediction error.

[0017] The second aspect of the present invention provides a device for predicting headless packages, including: a collection module, used to collect historical headless packages and obtain the logistics information corresponding to the historical headless packages; an extraction module, used to screen out processed headless packages based on the logistics information corresponding to the historical headless packages, and extract the key features of the processed headless packages; a training module, used to train a preset BP neural network model based on the extracted key features to obtain a headless package prediction model; a prediction module, used to obtain real-time headless packages, input the real-time headless packages into the headless package prediction model to obtain the prediction results output by the headless package prediction model; an adjustment module, used to output the prediction results, obtain the review results corresponding to the prediction results, and adjust the parameters and weights of the headless package prediction model according to the review results.

[0018] Optionally, in a first implementation method of the second aspect of the present invention, the collection module includes: a collection unit, used to acquire historical abnormal packages in the order system, and filter out historical headerless packages from the historical abnormal packages; a first acquisition unit, used to acquire the order information, transportation information and abnormal information corresponding to the historical headerless packages according to the encoding of the historical headerless packages; an integration unit, used to clean the acquired order information, transportation information and abnormal information, and integrate them into logistics information.

[0019] Optionally, in a second implementation of the second aspect of the present invention, the extraction module includes: a determination unit, used to determine the status of a historical headless package based on the logistics information corresponding to the historical headless package, the status of the historical headless package including processed and pending; a screening unit, used to screen out headless packages with a processed status, and parse the logistics information of the processed headless packages to obtain multiple features of the headless package; an extraction unit, used to extract key features of the headless package from the multiple features using regular expressions.

[0020] Optionally, in a third implementation of the second aspect of the present invention, the training module includes: a construction unit for constructing a preset BP neural network model; a labeling unit for forming training samples based on the extracted key features, and labeling the recipient address and recipient information of each training sample; a training unit for using the training samples to train the preset BP neural network model to obtain a headerless package prediction model.

[0021] Optionally, in a fourth implementation of the second aspect of the present invention, the prediction module includes: a second acquisition unit, used to acquire real-time headless package and related information; a noise reduction unit, used to denoise the related information to obtain noise reduction information; a prediction unit, used to input the noise reduction information into the headless package prediction model to obtain a prediction result output by the headless package prediction model, wherein the prediction result includes a predicted receiving address and recipient information.

[0022] Optionally, in a fifth implementation of the second aspect of the present invention, the adjustment module includes: an output unit, used to output a prediction result and obtain a review result corresponding to the prediction result; a comparison unit, used to compare and analyze the prediction result and the review result, and calculate a prediction error; an adjustment unit, used to analyze the prediction error, and determine the model parameters and weights that need to be adjusted based on the analysis result, and use a back propagation algorithm to adjust the parameters and weights of the headless package prediction model based on the prediction error.

[0023] The third aspect of the present invention provides a headless package prediction device, comprising: a memory and at least one processor, the memory storing computer-readable instructions, the memory and the at least one processor being interconnected via lines; the at least one processor calling the computer-readable instructions in the memory so that the headless package prediction device executes each step of the headless package prediction method as described above.

[0024] A fourth aspect of the present invention provides a computer-readable storage medium, in which computer-readable instructions are stored. When the computer-readable storage medium is run on a computer, the computer executes each step of the headless package prediction method as described above.

[0025] In the technical solution provided by the present invention, a headless package prediction model is obtained by collecting historical data and training a BP neural network model, and the headless package prediction model is used to predict real-time headless packages, which can achieve rapid response to real-time data, so that logistics companies can handle headless packages more efficiently and accurately, improve service quality, reduce operating costs, and improve customer satisfaction; in addition, the solution obtains review results corresponding to the prediction results, and adjusts the parameters and weights of the headless package prediction model according to the review results, thereby improving the prediction accuracy and generalization ability of the model through continuous model adjustment and optimization. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] Figure 1 A first flow chart of a method for predicting a headerless package provided by an embodiment of the present invention;

[0027] Figure 2 A second flow chart of the method for predicting a headerless package provided by an embodiment of the present invention;

[0028] Figure 3 A third flow chart of the method for predicting a headerless package provided by an embodiment of the present invention;

[0029] Figure 4 A fourth flow chart of the method for predicting a headerless package provided by an embodiment of the present invention;

[0030] Figure 5 A fifth flow chart of the method for predicting a headerless package provided by an embodiment of the present invention;

[0031] Figure 6 A sixth flow chart of the method for predicting a headerless package provided by an embodiment of the present invention;

[0032] Figure 7 A schematic diagram of the structure of a headless package prediction device provided by an embodiment of the present invention;

[0033] Figure 8A schematic diagram of the structure of a headless package prediction device provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0034] The embodiment of the present invention provides a method, device, equipment and storage medium for predicting headless packages. The method is used to predict headless packages through a headless package prediction model based on a BP neural network model, thereby improving the processing efficiency of headless packages. The method includes: collecting historical headless packages and obtaining the logistics information corresponding to the historical headless packages; based on the logistics information corresponding to the historical headless packages, screening out processed headless packages and extracting the key features of the processed headless packages; training a preset BP neural network model based on the extracted key features to obtain a headless package prediction model; obtaining a real-time headless package, inputting the real-time headless package into the headless package prediction model to obtain a prediction result output by the headless package prediction model; outputting the prediction result, obtaining a review result corresponding to the prediction result, and adjusting the parameters and weights of the headless package prediction model according to the review result.

[0035] 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.

[0036] For ease of understanding, the specific process of the embodiment of the present invention is described below. Figure 1 A first embodiment of a method for predicting a package without a header in an embodiment of the present invention includes:

[0037] S101. Collect historical headerless packages and obtain logistics information corresponding to the historical headerless packages.

[0038] In this embodiment, historical headerless package data is collected from the order system, including the package's size, weight, transportation route, scanning records, etc.

[0039] In this embodiment, detailed logistics information of each historical headerless package is obtained, including but not limited to the origin of the package, the stations it passes through, the final processing result, etc.

[0040] In this embodiment, the collected data may also be cleaned and preprocessed to remove noise and outliers.

[0041] It is understandable that the execution subject of the present invention may be a headless package prediction device, or a terminal or a server, which is not limited here. The embodiment of the present invention is described by taking a server as the execution subject as an example.

[0042] S102. Based on the logistics information corresponding to the historical headerless packages, the processed headerless packages are screened out, and key features of the processed headerless packages are extracted.

[0043] In this embodiment, the successfully processed headerless packages can be screened out according to the processing result field in the logistics information.

[0044] In this embodiment, the logistics information corresponding to the historical headerless packages is analyzed to extract the features that have an important impact on the prediction of the processing results of the headerless packages, such as the size, weight, and transportation route of the package.

[0045] S103: training a preset BP neural network model based on the extracted key features to obtain a headerless package prediction model.

[0046] In this embodiment, the BP neural network model, namely the back propagation neural network (Backpropagation Neural Network), is trained to enable the BP neural network model to learn and simulate the rules of the headerless package processing process.

[0047] In this embodiment, the BP neural network model is designed with multiple hidden layers, which can learn the complex nonlinear relationship between features, so as to more accurately identify the key attributes of the headless parts. During the training process, optimization strategies such as stochastic gradient descent are used to ensure that the model can achieve the best performance within a reasonable time.

[0048] S104, obtaining a real-time headerless package, and inputting the real-time headerless package into a headerless package prediction model to obtain a prediction result output by the headerless package prediction model.

[0049] In this embodiment, headerless parcels, i.e., real-time headerless parcels, are obtained from the logistics system in real time, and key features of the real-time headerless parcels are extracted and input into the headerless parcel prediction model. The headerless parcel prediction model calculates and outputs prediction information, such as possible recipient addresses, based on the input features.

[0050] S105: output the prediction result, obtain the review result corresponding to the prediction result, and adjust the parameters and weights of the headerless package prediction model according to the review result.

[0051] In this embodiment, the prediction results output by the headerless package prediction model are output to relevant personnel or systems, and the manual review results of the prediction results, that is, the actual processing results, are obtained.

[0052] Compare the predicted results with the actual results, analyze the prediction errors, and adjust the model parameters and weights.

[0053] The present embodiment provides a method for predicting headless packages, which obtains a headless package prediction model by collecting historical data and training a BP neural network model, and uses the headless package prediction model to predict real-time headless packages, thereby enabling rapid response to real-time data, thereby enabling logistics companies to handle headless packages more efficiently and accurately, improve service quality, reduce operating costs, and increase customer satisfaction. In addition, the solution obtains review results corresponding to the prediction results, and adjusts the parameters and weights of the headless package prediction model according to the review results, thereby improving the prediction accuracy and generalization ability of the model through continuous model adjustment and optimization.

[0054] See also Figure 2 , a second embodiment of the method for predicting a package without a header in an embodiment of the present invention includes:

[0055] S201. Obtain historical abnormal packages in the order system, and filter out historical headerless packages from the historical abnormal packages.

[0056] In this embodiment, in the order system, the query function is used to filter out all abnormal package records within the historical time period. The filter can be performed by date range, status label (such as "abnormal") and other conditions. Then, according to the definition of headless items (such as missing or unrecognizable information on the waybill, etc.), the historical headless packages are filtered out from the abnormal package data. Finally, the filtered historical headless packages are exported to CSV, Excel or other suitable data format files for subsequent processing.

[0057] S202. Obtain order information, transportation information, and exception information corresponding to the historical headerless package according to the code of the historical headerless package.

[0058] In this embodiment, in the order system, each abnormal package has an independent package code. For each historical headerless package, the package code is used to query the corresponding order information in the order center or related system, including the sender, recipient, product details, etc.

[0059] Input the parcel code into the logistics tracking platform to query the parcel's transportation information, including delivery time, transportation route, transfer station, estimated arrival time, etc. And collect parcel-related abnormal information from the order center's logistics tracking platform, such as parcel appearance pictures, barcode / QR code fragments, delay reasons, damage conditions, customer feedback, etc.

[0060] S203: clean the acquired order information, transportation information and abnormal information, and integrate them into logistics information.

[0061] In this embodiment, the collected order information, transportation information and exception information are cleaned to remove duplicates, correct erroneous data, format date and time, etc.

[0062] In this embodiment, the cleaned data is integrated into logistics information, and each historical headerless package corresponds to a piece of logistics information including order information, transportation information and exception information.

[0063] In this embodiment, historical abnormal packages in the order center are obtained, and historical headerless packages are screened out from the historical abnormal packages. The order information, transportation information and abnormal information corresponding to the historical headerless packages are obtained according to the codes of the historical headerless packages. The obtained order information, transportation information and abnormal information are cleaned and integrated into logistics information, which can improve data quality.

[0064] See also Figure 3 A third embodiment of a method for predicting a package without a header in an embodiment of the present invention includes:

[0065] S301. Determine the status of the historical headerless package based on the logistics information corresponding to the historical headerless package. The status of the historical headerless package includes processed and pending.

[0066] In this embodiment, the status of the historical headerless package is processed, which means that the package has been successfully identified and processed in some way (such as manual identification, technical identification, etc.), while the status of the historical headerless package is pending, which means that the package has not yet been identified or processed.

[0067] In this embodiment, the logistics information corresponding to the historical headless packages is determined based on the collected logistics information and combined with preset rules or algorithms (such as checking the processing result field, processing time, etc.) to determine the status of each headless package and classify it as "processed" or "pending".

[0068] S302, filtering out the headerless packages whose status is processed, and parsing the logistics information of the processed headerless packages to obtain multiple features of the headerless packages.

[0069] In this embodiment, the headless packages with the status of processed are screened out from the historical headless packages classified as processed and unprocessed, and the processed headless packages are parsed, and the parsed information is organized into multiple features.

[0070] In this embodiment, the multiple features of the headerless package include: delivery time, transit station, package appearance picture, barcode / QR code fragment, delivery address, weight, size, and receipt address.

[0071] S303. Use regular expressions to extract key features of the headerless package from multiple features.

[0072] In this embodiment, the key features include a picture of the package appearance, a barcode / QR code fragment, a shipping address, a weight, a size, and a signing address.

[0073] In this embodiment, a corresponding regular expression is designed for each key feature that needs to be extracted.

[0074] In this embodiment, the extracted key features may also be organized into a structured data format (such as JSON, CSV, etc.).

[0075] In this embodiment, the status of the historical headless packages is determined based on the logistics information corresponding to the historical headless packages. According to the status of the historical headless packages, the processed headless packages can be quickly screened out, and the logistics information of the processed headless packages can be parsed to obtain multiple features of the headless packages. Then, regular expressions are used to extract the key features of the headless packages from the multiple features, so that the key features can be extracted more accurately.

[0076] See also Figure 4 A fourth embodiment of a method for predicting a package without a header in an embodiment of the present invention includes:

[0077] S401, constructing a preset BP neural network model.

[0078] In this embodiment, a preset BP neural network model is constructed, specifically including: using an input layer, a first hidden layer, a second hidden layer and an output layer to construct a basic architecture of the BP neural network model; determining that the activation functions of the first hidden layer and the second hidden layer are nonlinear activation functions, and determining that the activation function of the output layer is a Softmax activation function; assigning initial values ​​to each connection weight and bias in the BP neural network model, and initializing the BP neural network model to obtain a preset neural network model.

[0079] In this embodiment, the first hidden layer extracts features from the input data. The second hidden layer is used to further extract and combine features to increase the complexity of the model.

[0080] In this embodiment, the number of nodes in the input layer will be equal to the number of key features, the number of nodes in the first hidden layer can be less than that in the input layer to facilitate the extraction of useful features, and the number of nodes in the second hidden layer is less than or equal to that in the first hidden layer.

[0081] S402: construct training samples based on the extracted key features, and annotate the recipient address and recipient information of each training sample.

[0082] In this embodiment, the key features corresponding to each headerless package except the receiving address and the recipient information are organized into a training sample, and the receiving address and the recipient information of each training sample are labeled.

[0083] S403: Use the training samples to train the preset BP neural network model to obtain a headerless package prediction model.

[0084] In this embodiment, the training samples are divided into a training set and a data set. The training set is used to train the preset BP neural network model, and the test set is used to train and test the trained BP neural network model. The parameters of the trained BP neural network model are adjusted according to the test results to obtain a headerless package prediction model.

[0085] When using the training set to train the preset BP neural network model, the parameters in the training process are pre-set, such as learning rate, number of iterations, batch size, etc. Then the training set is input into the preset BP neural network, and the output of each neuron is calculated through forward propagation until the output layer, and the loss function (such as mean square error, cross entropy loss, etc.) is calculated based on the prediction results and annotation information of the output layer. Finally, according to the gradient of the loss function, the weight and bias of the network are adjusted through the back propagation algorithm to reduce the loss.

[0086] Iterative training is performed during the training process, that is, the process of forward propagation, loss calculation and back propagation is repeated until the preset number of iterations is reached or the loss meets certain conditions.

[0087] In this embodiment, training samples are formed based on the extracted key features, and the delivery address and recipient of each training sample are marked. The training samples are used to train the preset BP neural network model to obtain a headless package prediction model. The model is constructed according to the specific characteristics of the headless package, which can more accurately capture the complexity and diversity of package processing. Moreover, by training a large amount of historical data, the model can learn the potential relationship between the key features and the delivery address and recipient, thereby improving the accuracy of the prediction.

[0088] See also Figure 5 A fifth embodiment of a method for predicting a package without a header in an embodiment of the present invention includes:

[0089] S501, obtaining real-time headerless packages and related information.

[0090] In this embodiment, the headerless package currently being processed and related information are obtained through the real-time interface or database query of the logistics system.

[0091] The associated information is any information such as the unique identifier of the package (such as the waybill number), size, weight, shipping time, shipping location, appearance picture of the package, transportation route and shipping location.

[0092] S502: De-noise the associated information to obtain de-noised information.

[0093] In this embodiment, the associated information is cleaned and preprocessed to remove duplicate, erroneous or irrelevant information, and then data noise reduction techniques such as filtering, smoothing, outlier detection and correction are applied to reduce noise and interference in the data. After noise reduction processing, high-quality noise reduction information directly related to the headerless package is obtained, and the noise reduction information will be used as the input of the prediction model to generate prediction results.

[0094] In this embodiment, rule matching, data mining or machine learning algorithms can also be used to identify and filter out information that is not directly related to the prediction target, and retain key information that is useful for prediction.

[0095] S503: Input the noise reduction information into the headerless package prediction model to obtain a prediction result output by the headerless package prediction model, where the prediction result includes a predicted delivery address and recipient information.

[0096] In this embodiment, the noise reduction information is organized into a format suitable for prediction model input, such as a vector, matrix or a specific data structure. Then, the organized noise reduction information is input into the trained headless package prediction model. The prediction model learns and infers the noise reduction information through a complex calculation and analysis process to generate a prediction result, which includes the predicted recipient address and recipient information.

[0097] In this embodiment, the prediction results will serve as an important basis for logistics companies to handle headerless packages, helping to quickly locate recipients and delivery addresses.

[0098] For example, a real-time headless package A is collected. The real-time headless package A only has a tracking code ABC12345 and no recipient information. Through the logistics system, the following information associated with the tracking code is obtained:

[0099] Tracking code: ABC12345

[0100] Place of shipment: Wuhou District, Chengdu

[0101] Shipping date: June 15, 2023

[0102] Package size: length 50cm, width 40cm, height 30cm

[0103] Package weight: 3kg

[0104] Transportation route: It starts from Chengdu, passes through Chongqing, and is currently in the sorting center in Beijing;

[0105] When predicting the real-time headerless package A, first reduce the noise of the associated information, specifically retaining "Wuhou District, Chengdu City" and converting it into a code or format recognizable by the model; standardize the package size and weight, such as converting the size into a relative size (such as the ratio of length / width / height), and converting the weight into a standardized value of a standard weight unit (such as kilograms); simplify it to the starting point (Wuhou District, Chengdu City) and the current location (Beijing sorting center), and ignore the intermediate sites.

[0106] Next, the denoised information is input into the headless package prediction model for prediction. The input layer of the headless package prediction model receives the denoised information, such as the mailing place code, and the standardized values ​​of the package size and weight; the hidden layer extracts and combines high-level representations of the input features through nonlinear transformations such as weighted summation and activation functions (such as ReLU, Sigmoid, or Tanh); the output layer generates prediction results, including the predicted delivery address (such as "No. XX, XX Street, Dongcheng District, Beijing") and possible recipient information (such as "Mr. Wang").

[0107] In this embodiment, the headless packages and their related information are acquired in real time, and are input into the prediction model after noise reduction processing, and finally accurate prediction results are obtained. Its data-driven and machine learning-based method not only improves the efficiency and accuracy of headless package processing, but also reduces operating costs.

[0108] See also Figure 6 A sixth embodiment of a method for predicting a package without a header in an embodiment of the present invention includes:

[0109] S601: Output the prediction result, and obtain the review result corresponding to the prediction result.

[0110] In this embodiment, the prediction result is usually some representation of the recipient address (such as province, city, district, street, etc.).

[0111] The reviewer will evaluate the accuracy of the forecast results based on his / her professional knowledge and practical experience and give the correct delivery address.

[0112] S602: Compare and analyze the prediction results and the review results, and calculate the prediction error.

[0113] In this embodiment, the prediction results are compared with the review results item by item to identify the errors and deviations in the prediction. The prediction error is calculated based on the comparison results. The prediction error can be measured by a variety of indicators, such as absolute error (distance of address deviation), relative error (prediction accuracy), mean square error (MSE), etc.

[0114] S603, analyzing the prediction error, and determining the model parameters and weights that need to be adjusted according to the analysis results, and using the back propagation algorithm to adjust the parameters and weights of the headerless package prediction model according to the prediction error.

[0115] In this embodiment, based on the error analysis results, it is evaluated which model parameters and weights contribute most to the error. This usually involves a sensitivity analysis of the model parameters to determine which parameters have the most significant impact on the model performance.

[0116] In this embodiment, based on the evaluation results of the parameters and weights, hyperparameters such as the learning rate and the regularization coefficient are adjusted, or certain weights are reinitialized.

[0117] In this embodiment, the gradient of the loss function for each parameter in the network is calculated based on the prediction error and the loss function. This step is the core of the back propagation algorithm, which uses the chain rule to calculate the gradient layer by layer. Finally, based on the calculated gradient, the gradient descent method or its variants (such as Adam, RMSprop, etc.) are used to update the parameters and weights of the model.

[0118] In this embodiment, the model algorithm can also be optimized, such as adjusting the loss function, regularization strategy, etc.

[0119] In this embodiment, through manual review and error-based model adjustment, the prediction error can be gradually reduced, and the accuracy of the headless package prediction model in the headless package prediction task can be improved. Moreover, using the review results as a supervision signal can help the headless package prediction model better learn the essential characteristics of the data, thereby improving its performance on unseen data.

[0120] The above describes the method for predicting a package without a header in an embodiment of the present invention. The following describes the device in an embodiment of the present invention. Figure 7 , the implementation of the headless package prediction device in the embodiment of the present invention includes:

[0121] The collection module 701 is used to collect historical headerless packages and obtain the logistics information corresponding to the historical headerless packages;

[0122] An extraction module 702 is used to filter out processed headless packages based on the logistics information corresponding to the historical headless packages, and extract key features of the processed headless packages;

[0123] The training module 703 is used to train the preset BP neural network model based on the extracted key features to obtain a prediction model for headerless packages;

[0124] Prediction module 704, used to obtain real-time headless package, and input the real-time headless package into the headless package prediction model to obtain the prediction result output by the headless package prediction model;

[0125] The adjustment module 705 is used to output the prediction result, obtain the review result corresponding to the prediction result, and adjust the parameters and weights of the headless package prediction model according to the review result.

[0126] In this embodiment, the collection module 701 includes: a collection unit 7011, which is used to obtain historical abnormal packages in the order system, and filter out historical headerless packages from the historical abnormal packages; a first acquisition unit 7012, which is used to obtain the order information, transportation information and abnormal information corresponding to the historical headerless packages according to the encoding of the historical headerless packages; an integration unit 7013, which is used to clean the acquired order information, transportation information and abnormal information, and integrate them into logistics information.

[0127] In this embodiment, the extraction module 702 includes: a determination unit 7021, which is used to determine the status of a historical headless package based on the logistics information corresponding to the historical headless package, and the status of the historical headless package includes processed and pending; a screening unit 7022, which is used to screen out headless packages with a processed status, and parse the logistics information of the processed headless packages to obtain multiple features of the headless package; an extraction unit 7023, which is used to use regular expressions to extract key features of the headless package from the multiple features.

[0128] In this embodiment, the training module 703 includes: a construction unit 7031, which is used to construct a preset BP neural network model; a labeling unit 7032, which is used to form training samples based on the extracted key features, and label the recipient address and recipient information of each training sample; a training unit 7033, which is used to use the training samples to train the preset BP neural network model to obtain a headerless package prediction model.

[0129] In this embodiment, the prediction module 704 includes: a second acquisition unit 7041, used to obtain real-time headless package and related information; a noise reduction unit 7042, used to denoise the related information to obtain noise reduction information; a prediction unit 7043, used to input the noise reduction information into the headless package prediction model to obtain the prediction result output by the headless package prediction model, and the prediction result includes the predicted receiving address and recipient information.

[0130] In this embodiment, the adjustment module 705 includes: an output unit 7051, which is used to output the prediction result and obtain the review result corresponding to the prediction result; a comparison unit 7052, which is used to compare and analyze the prediction result and the review result, and calculate the prediction error; an adjustment unit 7053, which is used to analyze the prediction error, and determine the model parameters and weights that need to be adjusted according to the analysis result, and use the back propagation algorithm to adjust the parameters and weights of the headless package prediction model according to the prediction error.

[0131] In this embodiment, by collecting historical data and training the BP neural network model, a headless package prediction model is obtained, and the headless package prediction model is used to predict real-time headless packages, which can achieve a rapid response to real-time data, so that logistics companies can handle headless packages more efficiently and accurately, improve service quality, reduce operating costs, and improve customer satisfaction; in addition, the solution obtains review results corresponding to the prediction results, and adjusts the parameters and weights of the headless package prediction model according to the review results, thereby improving the prediction accuracy and generalization ability of the model through continuous model adjustment and optimization.

[0132] above Figure 7 The headless package prediction device in the embodiment of the present invention is described in detail from the perspective of modular functional entities, and the headless package prediction device in the embodiment of the present invention is described in detail from the perspective of hardware processing.

[0133] Figure 8 800 may have relatively large differences due to different configurations or performances, and may include one or more processors (central processing units, CPU) 810 (for example, one or more processors) and memory 820, and one or more storage media 830 (for example, one or more mass storage devices) storing application programs 833 or data 832. Among them, the memory 820 and the storage medium 830 may be short-term storage or permanent storage. The program stored in the storage medium 830 may include one or more modules (not shown), and each module may include a series of instruction operations in the device 800. Furthermore, the processor 810 may be configured to communicate with the storage medium 830 to execute a series of instruction operations in the storage medium on the device 800.

[0134] The device 800 may also include one or more power supplies 840, one or more wired or wireless network interfaces 850, one or more input and output interfaces 860, and / or one or more operating systems 831, such as Windows Serve, Mac OS X, Unix, Linux, FreeBSD, etc.

[0135] An embodiment of 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 headless package prediction method.

[0136] 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.

[0137] 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.

[0138] 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 predicting a package without a header, characterized in that: The method for predicting a headerless package includes: Collect historical headerless packages and obtain the logistics information corresponding to the historical headerless packages; Based on the logistics information corresponding to the historical headerless packages, the processed headerless packages are screened out, and key features of the processed headerless packages are extracted; Based on the extracted key features, the preset BP neural network model is trained to obtain the prediction model of headerless package; Acquire a real-time headless package, and input the real-time headless package into the headless package prediction model to obtain a prediction result output by the headless package prediction model; The prediction result is output, and a review result corresponding to the prediction result is obtained, and the parameters and weights of the headless package prediction model are adjusted according to the review result.

2. The method for predicting a package without a header according to claim 1, characterized in that: The collecting of historical headerless packages and obtaining logistics information corresponding to the historical headerless packages includes: Obtain historical abnormal packages in the order system, and filter out historical headerless packages from the historical abnormal packages; Acquire order information, transportation information and abnormal information corresponding to the historical headerless package according to the code of the historical headerless package; The acquired order information, transportation information and exception information are cleaned and integrated into logistics information.

3. The method for predicting a package without a header according to claim 1, characterized in that: The method of screening out processed headless packages based on the logistics information corresponding to the historical headless packages and extracting key features of the processed headless packages includes: Determine the status of the historical headless package based on the logistics information corresponding to the historical headless package, where the status of the historical headless package includes processed and pending; Filter out the headerless packages that are in the processed state, parse the logistics information of the processed headerless packages, and obtain multiple features of the headerless packages; Regular expressions are used to extract key features of the headerless package from the multiple features.

4. The method for predicting a package without a header according to claim 1, characterized in that: The preset BP neural network model is trained based on the extracted key features to obtain a headless package prediction model, including: Construct a preset BP neural network model; Based on the extracted key features, training samples are formed, and the recipient address and recipient information of each training sample are annotated; The preset BP neural network model is trained using the training samples to obtain a headerless package prediction model.

5. The method for predicting a package without a header according to claim 4, characterized in that: The construction of the preset BP neural network model includes: Construct the basic architecture of BP neural network model using input layer, first hidden layer, second hidden layer and output layer; Determine that the activation functions of the first hidden layer and the second hidden layer are nonlinear activation functions, and determine that the activation function of the output layer is a Softmax activation function; Assign initial values ​​to each connection weight and bias in the BP neural network model, and initialize the BP neural network model to obtain a preset neural network model.

6. The method for predicting a package without a header according to claim 1, characterized in that: The step of obtaining a real-time headless package and inputting the real-time headless package into the headless package prediction model to obtain a prediction result output by the headless package prediction model includes: Get real-time headerless packages and related information; Denoising the associated information to obtain denoised information; The noise reduction information is input into the headless package prediction model to obtain a prediction result output by the headless package prediction model, wherein the prediction result includes a predicted receiving address and recipient information.

7. The method for predicting a package without a header according to claim 1, characterized in that: The outputting of the prediction result, obtaining a review result corresponding to the prediction result, and adjusting the parameters and weights of the headless package prediction model according to the review result include: Outputting a prediction result and obtaining a review result corresponding to the prediction result; Comparing and analyzing the prediction result and the review result, and calculating the prediction error; The prediction error is analyzed, and the model parameters and weights that need to be adjusted are determined based on the analysis results, and a back propagation algorithm is used to adjust the parameters and weights of the headless package prediction model based on the prediction error.

8. A device for predicting a package without a header, characterized in that: include: A collection module, used to collect historical headerless packages and obtain logistics information corresponding to the historical headerless packages; An extraction module, used to screen out processed headless packages based on the logistics information corresponding to the historical headless packages, and extract key features of the processed headless packages; A training module is used to train a preset BP neural network model based on the extracted key features to obtain a prediction model for headerless packages; A prediction module, used for obtaining a real-time headless package, and inputting the real-time headless package into the headless package prediction model to obtain a prediction result output by the headless package prediction model; An adjustment module is used to output the prediction result, obtain a review result corresponding to the prediction result, and adjust the parameters and weights of the headless package prediction model according to the review result.

9. A headless package prediction device, characterized in that: comprising a memory and at least one processor, wherein the memory has computer-readable instructions stored therein; The at least one processor calls the computer-readable instructions in the memory to execute the various steps of the headless package prediction method as described in any one of claims 1-7.

10. A computer-readable storage medium having computer-readable instructions stored thereon, characterized in that: When the computer-readable instructions are executed by a processor, the various steps of the headless package prediction method as described in any one of claims 1-7 are implemented.