Cost allocation optimization method and device, equipment and storage medium

Abstract

The invention relates to the technical field of logistics, and discloses a cost allocation optimization method, which comprises the steps of data fusion and acquisition, real-time access and fusion of multi-source logistics data, and construction of a dynamic allocation reference library containing entity relationship knowledge; accurate diagnosis is carried out, and real-time anomaly detection is carried out on a cost allocation result; dynamic optimization is carried out, an optimization process is modeled as a Markov decision process, and cost allocation parameters are adjusted according to a system state including allocation accuracy and business volume fluctuation; visualization and feedback: displaying a cost allocation path, receiving feedback of a user on an allocation result, and taking the feedback as a supervision signal to drive an optimization process; continuous learning is carried out, and model iteration is carried out in a mode of combining online learning and offline batch retraining. According to the invention, the dynamic self-adaptive optimization of the allocation strategy is realized; the rationality of the allocation result is improved; through a mode of combining online learning and offline retraining, long-term effectiveness and continuous optimization capability of the model in business scene change are guaranteed.

Description

Technical Field

[0001] This invention relates to the field of logistics information technology, and in particular to an optimization method, apparatus, equipment, and storage medium for cost allocation. Background Technology

[0002] The current financial allocation systems in the express delivery industry generally suffer from significant flaws. These systems are often independent, forming data silos, leading to a heavy reliance on manual processing and frequent human error resulting in low-quality financial data. Existing allocation methods mostly employ simple proportional allocation, such as allocating based on the weight or volume of packages. This simplistic approach fails to accurately reflect the complexities of routes and time-sensitive requirements, hindering the accurate reflection of actual resource consumption.

[0003] Traditional cost allocation processes are often handled in batches on a monthly or quarterly basis. These processes are lengthy, lack real-time responsiveness, and fail to provide immediate cost feedback for dynamic operations. They are also ill-suited for timely data feedback in daily operations, such as dynamic pricing and routing planning. Furthermore, cost allocation systems suffer from rigid architectures, weak adaptability, and a lack of flexibility. They cannot adapt to fluctuations in business volume or changes in network structure, such as during promotional seasons. The lack of transparency in the allocation process and poor traceability and interpretability of the results easily lead to disputes in internal performance evaluations and pose challenges for external audits and compliance checks.

[0004] Therefore, it is necessary to invent an optimized method, apparatus, equipment, and storage medium for cost allocation that can achieve scientific and reasonable allocation, improve allocation accuracy, enhance real-time decision-making, reduce operating and labor costs, and improve allocation transparency and compliance. Summary of the Invention

[0005] This invention provides an optimization method, apparatus, device, and storage medium for cost allocation, which enables scientific and reasonable allocation, improves allocation accuracy, enhances real-time decision-making, reduces operating and labor costs, and improves allocation transparency and compliance.

[0006] The first aspect of the present invention provides a method for optimizing cost allocation, the method comprising: Data fusion and collection: Real-time access and fusion of multi-source logistics data through a distributed network to build a dynamic allocation benchmark library containing entity relationship knowledge; Accurate diagnosis involves applying machine learning algorithms to detect anomalies in cost allocation results in real time and tracing them back along the data flow and business flow. Dynamic optimization, based on deep reinforcement learning algorithms, models the optimization process as a Markov decision process, and adjusts the cost allocation parameters according to the system state, including allocation accuracy and business volume fluctuations. Visualization and feedback: Display the cost allocation path, receive user feedback on the allocation results, and use the feedback as a monitoring signal to drive the optimization process; Continuous learning is employed, combining online learning with offline batch retraining to continuously optimize the machine learning model and iterate the model.

[0007] Optionally, in a first implementation of the first aspect of the present invention, the step of accessing and fusing multi-source logistics data in real time through a distributed network to construct a dynamic allocation benchmark library containing entity relationship knowledge includes: A distributed network with a microservice architecture is used to collect logistics data from multiple heterogeneous data sources in real time. The raw data collected is cleaned, standardized, and correlated and fused in real time using a stream processing engine. The fused data is injected into a dynamic allocation benchmark library, and an entity relationship network is constructed using knowledge graph technology.

[0008] Optionally, in a second implementation of the first aspect of the present invention, the accuracy diagnosis applies a machine learning algorithm to perform real-time anomaly detection on the cost allocation results and performs reverse tracing along the data flow and business flow, including: Establish a multi-dimensional accuracy evaluation system that includes both direct and indirect indicators; Apply machine learning algorithms to perform real-time anomaly detection on cost allocation results; Based on stream processing technology, each batch of allocation results is scanned in real time and a warning is issued immediately upon detection of anomalies. When an anomaly is detected, reverse tracing is performed along the data flow and business flow to accurately pinpoint the root cause of the problem.

[0009] Optionally, in a third implementation of the first aspect of the present invention, the dynamic optimization, based on a deep reinforcement learning algorithm, models the optimization process as a Markov decision process, and adjusts the cost allocation parameters according to the system state, including allocation accuracy and business volume fluctuations, including: The dynamic optimization process of cost allocation parameters is modeled as a Markov decision process; A deep reinforcement learning algorithm is used to learn the optimal parameters and adjust the strategy through continuous interaction between the agent and the environment; During the parameter optimization process, multiple business scenario factors such as business volume fluctuations, changes in business types, network structure changes, and cost structure changes are considered to ensure that the parameter adjustment strategy conforms to actual business constraints. The effect of each parameter adjustment is continuously tracked and evaluated, and the evaluation results are fed back to the deep reinforcement learning model as a reward signal.

[0010] Optionally, in a fourth implementation of the first aspect of the present invention, the visualization and feedback, which displays the cost allocation path, receives user feedback on the allocation results, and uses the feedback as a monitoring signal to drive the optimization process, includes: Visualize the cost allocation path and results using charts; Provide a user interface to receive user feedback on the allocation results; User feedback is transformed into supervisory signals and applied to the optimization process. Positive affirmation feedback is used as a positive reinforcement signal, and negative questioning feedback is used as a negative penalty signal, which are then input into the reward function of the deep reinforcement learning model. When a specific type of user feedback is received, a manual rule audit process is automatically triggered to check the current allocation rules.

[0011] Optionally, in a fifth implementation of the first aspect of the present invention, the continuous learning, through a combination of online learning and offline batch retraining, continuously optimizes the machine learning model and iterates the model, including: The machine learning model is optimized by combining online learning with offline batch retraining. Continuously monitor the model's input data distribution and prediction performance metrics, and automatically trigger the model retraining process when a decline in model performance or a change in data distribution is detected. Version control is implemented for training data, model code, and parameters to ensure the reproducibility of model iterations.

[0012] Optionally, in a sixth implementation of the first aspect of the present invention, the continuous learning, which involves continuously optimizing the machine learning model and iterating the model through a combination of online learning and offline batch retraining, further includes: The effectiveness evaluation of self-healing actions, user feedback, and information on changes in the business environment are used as feedback signals. The feedback signal is continuously input into the model training process.

[0013] A second aspect of the present invention provides a cost-sharing optimization apparatus, comprising: The feature database initialization module is used by the mobile terminal to create a feature database for storing multi-dimensional feature data of the printer, and to initialize the feature database. The initialization feature database module includes: The data collection unit is used to collect logistics data from multiple heterogeneous data sources in real time via a distributed network with a microservice architecture. The processing unit is used to perform real-time cleaning, standardization, and correlation fusion of the collected raw data using a stream processing engine; The building unit is used to inject the fused data into the dynamic allocation benchmark library and to build an entity relationship network using knowledge graph technology.

[0014] The connection priority determination module is used to determine the connection priority of the printer by calling a priority sorting algorithm to score the printers in the database when the automatic connection condition is triggered. The connection priority determination module includes: Establish a unit to build a multi-dimensional accuracy evaluation system that includes direct and indirect indicators; The detection unit is used to perform real-time anomaly detection on the cost allocation results using machine learning algorithms. The scanning unit is used to perform real-time and rapid scanning of each batch of allocation results based on stream processing technology, and to immediately issue an alert when an anomaly is detected. The traceability unit is used to trace back along the data flow and business flow when an anomaly is detected, so as to accurately locate the root cause of the problem.

[0015] An automatic connection module is used to determine the target printer from the printers according to the connection priority of the printers and initiate an automatic connection; The automatic connection module includes: The modeling unit is used to model the dynamic optimization process of cost allocation parameters as a Markov decision process; The learning unit is used to learn the optimal parameters and adjust the strategy through continuous interaction between the agent and the environment using deep reinforcement learning algorithms. The determination unit is used to ensure that the parameter adjustment strategy conforms to the actual business constraints during the parameter optimization process, based on multi-dimensional business scenario factors such as business volume fluctuations, changes in business types, network structure changes, and cost structure changes. The evaluation unit is used to continuously track and evaluate the effect of each parameter adjustment, and feeds the evaluation results back to the deep reinforcement learning model as a reward signal.

[0016] The anomaly diagnosis and output module is used to enable manual connection mode if automatic connection fails, and to perform anomaly diagnosis and output the diagnosis results and solutions if manual connection fails. The anomaly diagnosis and output module includes: The display unit is used to visually represent the cost allocation path and results through charts and graphs. The receiving unit is used to provide a user interface and receive user feedback on the allocation results; The conversion unit is used to convert user feedback into supervisory signals and apply them to the optimization process. Positive confirmation feedback is used as a positive reinforcement signal, and negative questioning feedback is used as a negative penalty signal. These signals are then input into the reward function of the deep reinforcement learning model. When a specific type of user feedback is received, a manual rule audit process is automatically triggered to check the current allocation rules.

[0017] The feature update module is used to update the multi-dimensional feature data of the corresponding device in the printer feature database after the connection interaction with the printer is completed.

[0018] The updated feature module includes: Training units are used to optimize machine learning models by combining online learning with offline batch retraining. The monitoring unit is used to continuously monitor the model's input data distribution and prediction performance metrics. When a decline in model performance or a change in data distribution is detected, the model retraining process is automatically triggered. The version management unit is used to manage training data, model code, and parameters in a versioned manner, ensuring the reproducibility of model iterations.

[0019] The continuous learning, which combines online learning with offline batch retraining to continuously optimize the machine learning model and iterate the model, also includes: The effectiveness evaluation of self-healing actions, user feedback, and information on changes in the business environment are used as feedback signals. The feedback signal is continuously input into the model training process.

[0020] A third aspect of the present invention provides a cost-sharing optimization device, comprising a memory and at least one processor, wherein the memory stores computer-readable instructions; The at least one processor invokes the computer-readable instructions in the memory to perform the steps of the cost-allocation optimization method as described above.

[0021] A fourth aspect of the present invention provides a computer-readable storage medium storing computer-readable instructions, which, when executed by a processor, implement the steps of the cost-allocation optimization method described above.

[0022] The technical solution of this invention solves the problems of data silos and lag in traditional cost allocation; it utilizes machine learning algorithms to detect anomalies and trace back the allocation results, improving the transparency and accuracy of the allocation process; based on deep reinforcement learning, it models the optimization process as a Markov decision process, enabling the system to automatically adjust parameters according to business fluctuations and allocation accuracy, achieving dynamic adaptive optimization of the allocation strategy; a visual feedback mechanism forms a closed loop of system decision-making-human intervention-model iteration, improving the rationality of the allocation results; finally, by combining online learning and offline retraining, it ensures the long-term effectiveness and continuous optimization capability of the model in the face of changes in business scenarios. Attached Figure Description

[0023] Figure 1 A first flowchart of a cost allocation optimization method provided in an embodiment of the present invention; Figure 2 A second flowchart of the cost allocation optimization method provided in an embodiment of the present invention; Figure 3 A third flowchart of the cost allocation optimization method provided in the embodiments of the present invention; Figure 4 A fourth flowchart of the cost allocation optimization method provided in the embodiments of the present invention; Figure 5 A fifth flowchart of the cost allocation optimization method provided in the embodiments of the present invention; Figure 6 A sixth flowchart of the cost allocation optimization method provided in the embodiments of the present invention; Figure 7 A schematic diagram of the structure of the cost-sharing optimization device provided in an embodiment of the present invention; Figure 8 This is a schematic diagram of the structure of the cost-sharing optimization device provided in an embodiment of the present invention. Detailed Implementation

[0024] This invention provides a method, apparatus, device, and storage medium for optimizing cost allocation. The method is used to establish dynamic customer profiles and manage customers based on these dynamic customer profiles.

[0025] The terms "first," "second," "third," "fourth," etc. (if present) in the specification, claims, and accompanying drawings of this invention are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" or "having" and any variations thereof are intended to cover a non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0026] For ease of understanding, the specific process of the embodiments of the present invention is described below. Please refer to [link / reference]. Figure 1 The first embodiment of a cost allocation optimization method according to the present invention includes: S101. Data fusion and collection: Real-time access and fusion of multi-source logistics data through a distributed network to construct a dynamic allocation benchmark library containing entity relationship knowledge. S102. Accuracy diagnosis: Apply machine learning algorithms to perform real-time anomaly detection on cost allocation results and trace back along the data flow and business flow. S103. Dynamic optimization: Based on deep reinforcement learning algorithms, the optimization process is modeled as a Markov decision process, and the cost allocation parameters are adjusted according to the system state, which includes allocation accuracy and business volume fluctuations. S104. Visualization and feedback: Display the cost allocation path, receive user feedback on the allocation results, and use the feedback as a monitoring signal to drive the optimization process. S105. Continuous learning: Through a combination of online learning and offline batch retraining, the machine learning model is continuously optimized and iterated.

[0027] This invention addresses the issues of data silos and lag in traditional cost allocation; it utilizes machine learning algorithms for anomaly detection and reverse tracing of allocation results, improving the transparency and accuracy of the allocation process; it models the optimization process as a Markov decision process based on deep reinforcement learning, enabling the system to automatically adjust parameters according to business fluctuations and allocation accuracy, achieving dynamic adaptive optimization of the allocation strategy; it forms a closed loop of system decision-making, manual intervention, and model iteration through a visual feedback mechanism, improving the rationality of the allocation results; and finally, it ensures the long-term effectiveness and continuous optimization capability of the model in the face of changing business scenarios by combining online learning with offline retraining.

[0028] Please see Figure 2 The second embodiment of the cost allocation optimization method in this invention includes the following step: real-time access and fusion of multi-source logistics data through a distributed network to construct a dynamic allocation benchmark library containing entity relationship knowledge. S201. A distributed network with a microservice architecture is used to collect logistics data from multiple heterogeneous data sources in real time. S202. Use a stream processing engine to clean, standardize, and correlate the collected raw data in real time. S203. Inject the fused data into the dynamic allocation benchmark library and use knowledge graph technology to construct an entity relationship network.

[0029] The embodiments of this invention ensure the elasticity and scalability of data access; utilize a stream processing engine to perform real-time cleaning, standardization, and correlation fusion of raw data, effectively improving data quality and consistency; and leverage knowledge graph technology to construct an entity relationship network, enhancing the semantic relevance and dynamic reasoning capabilities of the cost allocation benchmark database, providing a high-quality data foundation for subsequent accurate and interpretable allocation decisions.

[0030] Please see Figure 3A third embodiment of a cost allocation optimization method in this invention includes an accuracy diagnosis that applies machine learning algorithms to perform real-time anomaly detection on the cost allocation results and performs reverse tracing along the data flow and business flow, comprising: S301. Establish a multi-dimensional accuracy evaluation system that includes both direct and indirect indicators; S302. Apply machine learning algorithms to perform real-time anomaly detection on cost allocation results; S303. Based on stream processing technology, perform real-time rapid scanning of each batch of allocation results and issue an immediate warning upon detecting anomalies; S304. When an anomaly is detected, trace back along the data flow and business flow to accurately locate the root cause of the problem.

[0031] The embodiments of the present invention significantly improve the monitoring efficiency and timeliness of problem detection in the allocation process; based on stream processing technology, each batch of allocation results is quickly scanned to ensure that anomalies can be captured in time and trigger the early warning mechanism; when an anomaly is detected, the precise location capability of bidirectional tracing along the data flow and business flow effectively shortens the problem investigation time and reduces the cost of manual intervention.

[0032] Please see Figure 4 The fourth embodiment of a cost allocation optimization method in this invention involves dynamic optimization based on a deep reinforcement learning algorithm. The optimization process is modeled as a Markov decision process, and the cost allocation parameters are adjusted according to the system state, which includes allocation accuracy and business volume fluctuations. This includes: S401. Model the dynamic optimization process of cost allocation parameters as a Markov decision process; S402. Employ a deep reinforcement learning algorithm to learn the optimal parameter adjustment strategy through continuous interaction between the agent and the environment; S403. During the parameter optimization process, based on multi-dimensional business scenario factors such as business volume fluctuations, changes in business types, network structure changes, and cost structure changes, ensure that the parameter adjustment strategy conforms to actual business constraints. S404. Continuously track and evaluate the effect of each parameter adjustment, and feed the evaluation results back to the deep reinforcement learning model as a reward signal.

[0033] In this embodiment of the invention, the dynamic optimization process of cost allocation parameters is modeled as a Markov decision process, and a deep reinforcement learning algorithm is used to achieve continuous interactive learning between the agent and the environment, enabling the system to autonomously adapt to complex and ever-changing business environments. During the parameter optimization process, multiple factors such as fluctuations in business volume, changes in business type, changes in network structure, and changes in cost structure are comprehensively considered to ensure that the parameter adjustment strategy meets both the requirements for allocation accuracy and the constraints of actual business. By continuously tracking and evaluating the effect of parameter adjustment and feeding the results back to the model as a reward signal, a closed-loop optimization mechanism is formed, which improves the adaptive capability and long-term operational stability of the cost allocation system.

[0034] Please see Figure 5 The fifth embodiment of a cost allocation optimization method in this invention includes visualization and feedback, which displays the cost allocation path, receives user feedback on the allocation results, and uses the feedback as a monitoring signal to drive the optimization process, comprising: S501. Display the cost allocation path and results through visual charts; S502. Provide a user interface to receive user feedback on the allocation results; S503. User feedback is transformed into a supervisory signal and applied to the optimization process. Positive confirmation feedback is used as a positive reinforcement signal, and negative questioning feedback is used as a negative penalty signal. These signals are then input into the reward function of the deep reinforcement learning model. When a specific type of user feedback is received, a manual rule audit process is automatically triggered to check the current allocation rules.

[0035] In this embodiment of the invention, the cost allocation path and results are intuitively displayed through visual charts, enhancing the transparency and interpretability of the allocation process. Feedback is received through a user interface, and positive confirmations are transformed into positive reinforcement signals, while negative questions are transformed into negative penalty signals, driving the optimization process of the deep reinforcement learning model and realizing a closed-loop optimization mechanism of human-machine collaboration. For specific types of user feedback, a rule audit process is automatically triggered, ensuring the rationality and compliance of business rules while improving the system's adaptability to complex scenarios.

[0036] Please see Figure 6 The sixth embodiment of a cost-allocation optimization method in this invention includes continuous learning, which continuously optimizes the machine learning model and iterates the model by combining online learning with offline batch retraining, comprising: S601. Optimize machine learning models by combining online learning with offline batch retraining; S602. Continuously monitor the model's input data distribution and prediction performance metrics. When a decline in model performance or a change in data distribution is detected, automatically trigger the model retraining process. S603. Implement version control for training data, model code, and parameters to ensure the reproducibility of model iterations.

[0037] In some embodiments, the continuous learning, which combines online learning with offline batch retraining to continuously optimize the machine learning model and perform model iteration, further includes: The effectiveness evaluation of self-healing actions, user feedback, and information on changes in the business environment are used as feedback signals. The feedback signal is continuously input into the model training process.

[0038] This invention combines online learning with offline batch retraining to enable the model to respond promptly to changes in the business environment while maintaining long-term stability; it continuously monitors the distribution of input data and prediction performance metrics, and automatically triggers the retraining process when performance degradation or data drift is detected, effectively preventing model degradation; and it ensures that the model iteration process is traceable and reproducible through version management of training data, code, and parameters.

[0039] The above describes the optimization method for cost allocation in the embodiments of the present invention. The following describes the apparatus in the embodiments of the present invention. Please refer to [link / reference]. Figure 7 The implementation methods of the cost-sharing optimization device in this invention include: The feature database initialization module 701 is used by the mobile terminal to create a feature database for storing multi-dimensional feature data of the printer, and to initialize the feature database. The connection priority determination module 702 is used to, when the automatic connection condition is triggered, call the priority sorting algorithm to score the priority of the printers in the database and determine the connection priority of the printer. The automatic connection module 703 is used to determine the target printer from the printers according to the connection priority of the printers and initiate an automatic connection; The anomaly diagnosis and output module 704 is used to enable manual connection mode if automatic connection fails, and to perform anomaly diagnosis and output diagnosis results and solutions if manual connection fails. The feature update module 705 is used to update the multi-dimensional feature data of the corresponding device in the printer feature database after completing the connection interaction with the printer.

[0040] In some embodiments, the initialization feature database module 701 includes: The acquisition unit 7011 is used to access logistics data from multiple heterogeneous data sources in a real-time manner through a distributed network with a microservice architecture. The processing unit 7012 is used to perform real-time cleaning, standardization, and correlation fusion of the collected raw data using a stream processing engine; The construction unit 7013 is used to inject the fused data into the dynamic allocation benchmark library and to build an entity relationship network using knowledge graph technology.

[0041] The embodiments of this invention ensure the elasticity and scalability of data access; utilize a stream processing engine to perform real-time cleaning, standardization, and correlation fusion of raw data, effectively improving data quality and consistency; and leverage knowledge graph technology to construct an entity relationship network, enhancing the semantic relevance and dynamic reasoning capabilities of the cost allocation benchmark database, providing a high-quality data foundation for subsequent accurate and interpretable allocation decisions.

[0042] In some embodiments, the connection priority determination module 702 includes: Unit 7021 is established to create a multi-dimensional accuracy evaluation system that includes both direct and indirect indicators. The detection unit 7022 is used to perform real-time anomaly detection on the cost allocation results using machine learning algorithms. The scanning unit 7023 is used to perform real-time rapid scanning of each batch of allocation results based on stream processing technology, and to immediately issue an alert when an anomaly is detected. The traceability unit 7024 is used to trace back along the data flow and business flow when an anomaly is detected, so as to accurately locate the root cause of the problem.

[0043] The embodiments of the present invention significantly improve the monitoring efficiency and timeliness of problem detection in the allocation process; based on stream processing technology, each batch of allocation results is quickly scanned to ensure that anomalies can be captured in time and trigger the early warning mechanism; when an anomaly is detected, the precise location capability of bidirectional tracing along the data flow and business flow effectively shortens the problem investigation time and reduces the cost of manual intervention.

[0044] In some embodiments, the automatic connection module 703 includes: Modeling unit 7031 is used to model the dynamic optimization process of cost allocation parameters as a Markov decision process; Learning unit 7032 is used to learn the optimal parameter adjustment strategy through continuous interaction between the agent and the environment using a deep reinforcement learning algorithm. The determination unit 7033 is used to ensure that the parameter adjustment strategy conforms to the actual business constraints during the parameter optimization process, based on multi-dimensional business scenario factors such as business volume fluctuations, changes in business types, network structure changes, and cost structure changes. Evaluation unit 7034 is used to continuously track and evaluate the effect of each parameter adjustment, and feed the evaluation results back to the deep reinforcement learning model as a reward signal.

[0045] In this embodiment of the invention, the dynamic optimization process of cost allocation parameters is modeled as a Markov decision process, and a deep reinforcement learning algorithm is used to achieve continuous interactive learning between the agent and the environment, enabling the system to autonomously adapt to complex and ever-changing business environments. During the parameter optimization process, multiple factors such as fluctuations in business volume, changes in business type, changes in network structure, and changes in cost structure are comprehensively considered to ensure that the parameter adjustment strategy meets both the requirements for allocation accuracy and the constraints of actual business. By continuously tracking and evaluating the effect of parameter adjustment and feeding the results back to the model as a reward signal, a closed-loop optimization mechanism is formed, which improves the adaptive capability and long-term operational stability of the cost allocation system.

[0046] In some embodiments, the anomaly diagnosis and output module 704 includes: Display unit 7041 is used to visually represent the cost allocation path and results through charts. The receiving unit 7042 is used to provide a user interface and receive user feedback on the allocation results; The conversion unit 7043 is used to convert user feedback into a supervision signal and apply it to the optimization process. Positive confirmation feedback is used as a positive reinforcement signal, and negative questioning feedback is used as a negative penalty signal, which is input into the reward function of the deep reinforcement learning model. When a specific type of user feedback is received, a manual rule audit process is automatically triggered to check the current allocation rules.

[0047] In this embodiment of the invention, the cost allocation path and results are intuitively displayed through visual charts, enhancing the transparency and interpretability of the allocation process. Feedback is received through a user interface, and positive confirmations are transformed into positive reinforcement signals, while negative questions are transformed into negative penalty signals, driving the optimization process of the deep reinforcement learning model and realizing a closed-loop optimization mechanism of human-machine collaboration. For specific types of user feedback, a rule audit process is automatically triggered, ensuring the rationality and compliance of business rules while improving the system's adaptability to complex scenarios.

[0048] In some embodiments, the updated feature module 705 includes: Training unit 7051 is used to optimize machine learning models by combining online learning with offline batch retraining. The monitoring unit 7052 is used to continuously monitor the model's input data distribution and prediction performance metrics. When a decline in model performance or a change in data distribution is detected, the model retraining process is automatically triggered. Version management unit 7053 is used to version control training data, model code and parameters to ensure the reproducibility of model iteration.

[0049] In some embodiments, the continuous learning, which combines online learning with offline batch retraining to continuously optimize the machine learning model and perform model iteration, further includes: The effectiveness evaluation of self-healing actions, user feedback, and information on changes in the business environment are used as feedback signals. The feedback signal is continuously input into the model training process.

[0050] This invention combines online learning with offline batch retraining to enable the model to respond promptly to changes in the business environment while maintaining long-term stability; it continuously monitors the distribution of input data and prediction performance metrics, and automatically triggers the retraining process when performance degradation or data drift is detected, effectively preventing model degradation; and it ensures that the model iteration process is traceable and reproducible through version management of training data, code, and parameters.

[0051] Figure 7 The structure of the cost allocation optimization device shown does not constitute a limitation on the cost allocation optimization device, and can implement the steps of the cost allocation optimization method provided in the above-described method embodiments.

[0052] above Figure 7 The cost-sharing optimization device in this embodiment of the invention is described in detail from the perspective of modular functional entities. The cost-sharing optimization equipment in this embodiment of the invention is described in detail from the perspective of hardware processing.

[0053] Figure 8 This is a schematic diagram of the structure of a cost-sharing optimization device provided in an embodiment of the present invention. The device 800 can vary significantly due to different configurations or performance characteristics. It may include one or more central processing units (CPUs) 810 (e.g., one or more processors) and a memory 820, and one or more storage media 830 (e.g., one or more mass storage devices) for storing application programs 833 or data 832. The memory 820 and storage media 830 can be temporary or persistent storage. The program stored in the storage media 830 may include one or more modules (not shown), each module including a series of instruction operations on the device 800. Furthermore, the processor 810 may be configured to communicate with the storage media 830 and execute the series of instruction operations in the storage media on the device 800.

[0054] Device 800 may also include one or more power supplies 840, one or more wired or wireless network interfaces 850, one or more input / output interfaces 860, and / or one or more operating systems 831, such as Windows Server, Mac OS X, Unix, Linux, FreeBSD, etc.

[0055] This invention also provides a computer-readable storage medium, which can be a non-volatile computer-readable storage medium or a volatile computer-readable storage medium. The computer-readable storage medium stores instructions that, when executed on a computer, cause the computer to perform the steps of a cost-sharing optimization method.

[0056] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working process of the system, device, or unit described above can be referred to the corresponding process in the foregoing method embodiments, and will not be repeated here.

[0057] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0058] The above-described embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A cost allocation optimization method, characterized by, The cost allocation optimization method comprises the following steps: Data fusion and collection: real-time access and fusion of multi-source logistics data through a distributed network to build a dynamic allocation benchmark library containing entity relationship knowledge; Accuracy diagnosis: applying machine learning algorithms to perform real-time anomaly detection on cost allocation results and tracing back along data flow and business flow; Dynamic optimization: based on deep reinforcement learning algorithms, modeling the optimization process as a Markov decision process, adjusting cost allocation parameters according to system states containing allocation accuracy and business volume fluctuations; Visualization and feedback: displaying cost allocation paths and receiving user feedback on allocation results, using the feedback as a supervisory signal to drive the optimization process; Continuous learning: continuously optimizing the machine learning model through a combination of online learning and offline batch retraining, and iterating the model.

2. The cost apportionment optimization method of claim 1, wherein, The dynamic allocation benchmark library containing entity relationship knowledge is built by real-time access and fusion of multi-source logistics data through a distributed network, which comprises the following steps: Distributed network using microservices architecture to access logistics data from multiple heterogeneous data sources in real-time collection mode; Using a stream processing engine to perform real-time cleaning, standardization and correlation fusion on the collected raw data; Injecting the fusion-processed data into the dynamic allocation benchmark library and using knowledge graph technology to build an entity relationship network.

3. The cost apportionment optimization method of claim 1, wherein, The accuracy diagnosis applies machine learning algorithms to perform real-time anomaly detection on cost allocation results and traces back along data flow and business flow, which comprises the following steps: Establish a multi-dimensional accuracy evaluation system containing direct and indirect indicators; Apply machine learning algorithms to perform real-time anomaly detection on cost allocation results; Based on stream processing technology, perform real-time and rapid scanning on each batch of allocation results, and immediately issue an early warning when an anomaly is detected; When an anomaly is detected, trace back along the data flow and business flow to accurately locate the root cause.

4. The cost apportionment optimization method of claim 3, wherein, The dynamic optimization based on deep reinforcement learning algorithms models the optimization process as a Markov decision process, adjusts cost allocation parameters according to system states containing allocation accuracy and business volume fluctuations, which comprises the following steps: Modeling the dynamic optimization process of cost allocation parameters as a Markov decision process; Using deep reinforcement learning algorithms to learn the optimal parameter adjustment strategy through continuous interaction between the agent and the environment; During the parameter optimization process, according to the multi-dimensional business scenario factors such as business volume fluctuations, business type changes, network structure changes and cost structure changes, ensure that the parameter adjustment strategy meets the actual business constraints; Continuously track and evaluate the effect of each parameter adjustment, and feed the evaluation results back to the deep reinforcement learning model as reward signals.

5. The cost apportionment optimization method of claim 3, wherein, The visualization and feedback display cost allocation paths and receive user feedback on allocation results, using the feedback as a supervisory signal to drive the optimization process, which comprises the following steps: Display cost allocation paths and results through visual charts; Provide a user interaction interface to receive user feedback on allocation results; Convert user feedback into supervisory signals and apply them to the optimization process, positive confirmation feedback as positive reinforcement signals, negative feedback as negative punishment signals, input into the reward function of the deep reinforcement learning model; When receiving a specific type of user challenge feedback, an automatic rule audit process is triggered to check the current allocation rule.

6. The cost apportionment optimization method of claim 5, wherein, The continuous learning continuously optimizes the machine learning model and iterates the model by combining online learning and offline batch retraining, including: The machine learning model is optimized by combining online learning and offline batch retraining; When detecting that the model performance decreases or the data distribution changes, the model retraining process is automatically triggered; The training data, model code and parameters are versioned to ensure the reproducibility of model iteration.

7. The cost apportionment optimization method of claim 6, wherein, The continuous learning continuously optimizes the machine learning model and iterates the model by combining online learning and offline batch retraining, and also includes: Determine the effect evaluation of self-healing action, user feedback and business environment change information as feedback signal; The feedback signal is continuously input to the model training process.

8. A cost allocation optimization apparatus, characterized by, It includes: The feature database module is initialized to create a feature database for storing multi-dimensional feature data of the printer on the mobile terminal, and the feature database is initialized; The connection priority determination module is used to call a priority sorting algorithm to score the printers in the database and determine the connection priority of the printers when the automatic connection condition is triggered; The automatic connection module is used to determine the target printer from the printers according to the connection priority of the printers and initiate automatic connection; The abnormal diagnosis and output module is used to enable manual connection mode if automatic connection fails, and perform abnormal diagnosis and output diagnosis results and solutions if manual connection fails; The feature updating module is used to update the multi-dimensional feature data of the corresponding equipment in the printer feature database after completing the connection interaction with the printer.

9. A cost allocation optimization device characterized by comprising: It includes a memory and at least one processor, and the memory stores computer readable instructions; The at least one processor invokes the computer readable instructions in the memory to perform the steps of the cost allocation optimization method according to any one of claims 1-7.

10. A computer-readable storage medium having stored thereon computer-readable instructions, wherein, The computer readable instructions are executed by the processor to realize the steps of the cost allocation optimization method according to any one of claims 1-7.

Images