A method, device and medium for single asset cost allocation and aggregation analysis
By combining pre-trained cost prediction models with asset status information and clustering, target predicted costs are generated, solving the problem of cost accounting distortion caused by ignoring asset status differences in traditional methods, and achieving cost analysis with high accuracy and flexibility.
Patent Information
- Application Number
- CN202511138227.7
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2025-08-14
- Publication Date
- 2026-01-20
- Estimated Expiration
- 2045-08-14
AI Technical Summary
Traditional methods of allocating costs for individual assets ignore differences in asset status, leading to distorted cost accounting, poor accuracy, and insufficient flexibility in complex scenarios, making it difficult to meet the needs of large-scale, highly dynamic asset cost management.
By employing a pre-trained cost prediction model combined with asset status information, and through clustering processing and training loss function optimization, a target predicted cost is generated, reflecting the characteristics of individual asset status and the commonalities of costs for similar assets, thus achieving dynamic adaptation.
It improves the accuracy and flexibility of cost analysis, solves the problem of allocation distortion caused by ignoring asset heterogeneity in traditional methods, and realizes dynamic adaptation of cost and asset status information.
Smart Images
Figure CN120725285B_ABST
Abstract
Description
TECHNICAL FIELD
[0001] The present application relates to the technical field of resource management, in particular to a single asset cost allocation and collection analysis method, device and medium. BACKGROUND
[0002] In the fields of enterprise asset management, engineering construction project accounting, and equipment operation and maintenance, single asset cost allocation and collection is a core link to realize fine cost control. Traditional cost allocation methods mainly rely on manual rules, such as equal distribution according to asset quantity, proportional distribution according to asset original value, or simple linear model mapping distribution. These traditional methods ignore asset state differences and treat similar assets as homogeneous individuals, without considering the actual state of assets, such as service life, degree of wear and tear, and operating efficiency, which affects cost consumption, resulting in distorted cost accounting and poor accuracy of cost analysis.
[0003] Some improved methods introduce clustering technology for asset grouping, but the clustering results and cost prediction models are independent of each other, and the clustering information cannot optimize the accuracy of cost prediction, resulting in allocation results that do not conform to the actual state of assets and are difficult to reflect the cost commonality of similar assets.
[0004] In the face of complex scenarios such as multiple cost categories and dynamically changing asset states, traditional methods need to frequently adjust manual rules or model parameters, have weak generalization ability, and are difficult to meet the needs of large-scale and high-dynamic asset cost management, resulting in poor flexibility of cost analysis.
[0005] Therefore, how to improve the accuracy and flexibility of single asset cost allocation and collection analysis has become a problem to be solved. SUMMARY
[0006] To solve the above technical problems, the present application adopts a single asset cost allocation and collection analysis method, which includes the following steps:
[0007] S101, obtaining a cost to be analyzed, wherein the cost to be analyzed includes cost information, a cost category, and M associated assets, M being a positive integer;
[0008] S102, if M = 1, then the cost information is collected into the cost category of a single associated asset, and if M ≠ 1, then asset state information corresponding to the M associated assets is obtained;
[0009] S103, obtaining dimension weight information set by a user, and calculating a first distance between asset state information corresponding to each two associated assets according to the dimension weight information and the asset state information corresponding to the M associated assets;
[0010] S104, clustering the M associated assets according to all the first distances, to obtain N clustering sets, wherein N is a positive integer;
[0011] S105, for any associated asset, inputting the cost category and asset state information corresponding to the associated asset into a pre-trained cost prediction model, to obtain an initial predicted cost corresponding to the associated asset;
[0012] S106, according to the cost information, the initial predicted cost corresponding to the M associated assets respectively, the asset state information corresponding to the M associated assets respectively and the clustering set to which the M associated assets respectively belong, calculating to obtain a training loss;
[0013] S107, according to the training loss, fine-tuning training the pre-trained cost prediction model until the training loss converges, to obtain a trained cost prediction model;
[0014] S108, for any associated asset, inputting the cost category and asset state information corresponding to the associated asset into the trained cost prediction model, to obtain a target predicted cost corresponding to the associated asset, taking the target predicted cost corresponding to the associated asset as the cost allocation cost of the cost category allocated to the associated asset.
[0015] The application also provides a single asset cost allocation and collection analysis device, which comprises:
[0016] A cost acquisition module is configured to acquire a to-be-analyzed cost, wherein the to-be-analyzed cost comprises cost information, a cost category and M associated assets, and M is a positive integer.
[0017] A cost collection module is configured to, if M=1, collect the cost information into the cost category of a single associated asset, and if M≠1, acquire asset state information corresponding to the M associated assets respectively.
[0018] A distance calculation module is configured to acquire dimension weight information set by a user, and calculate a first distance between asset state information corresponding to each two associated assets according to the dimension weight information and the asset state information corresponding to the M associated assets respectively.
[0019] An asset clustering module is configured to cluster the M associated assets according to all the first distances, to obtain N clustering sets, wherein N is a positive integer.
[0020] A cost prediction module is configured to, for any associated asset, input the cost category and asset state information corresponding to the associated asset into a pre-trained cost prediction model, to obtain an initial predicted cost corresponding to the associated asset.
[0021] a loss calculation module configured to calculate a training loss according to the cost information, initial predicted costs of the M associated assets respectively, asset state information of the M associated assets respectively, and the clustering set to which the M associated assets respectively belong;
[0022] a model fine-tuning module configured to fine-tune train the pre-trained cost prediction model according to the training loss until the training loss converges, to obtain a trained cost prediction model;
[0023] a cost allocation module configured to, for any associated asset, input the cost category and asset state information corresponding to the associated asset into the trained cost prediction model, to obtain a target predicted cost corresponding to the associated asset, and take the target predicted cost corresponding to the associated asset as an allocation cost of the cost category to which the associated asset is allocated.
[0024] The application further provides a computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the single-asset cost allocation and aggregation analysis method when executing the computer program.
[0025] The application further provides a computer-readable storage medium storing a computer program, wherein the computer program is executable on a processor to implement the single-asset cost allocation and aggregation analysis method.
[0026] The application has at least the following beneficial effects: the pre-trained cost prediction model is combined with asset state information to generate initial predicted costs, and the model training is optimized based on the distance of clustering results and asset state information, so that the final target predicted costs can reflect the state characteristics of single assets and the cost commonality of similar assets, the allocation distortion problem caused by ignoring asset heterogeneity in traditional methods is solved, the accuracy of cost analysis is improved, assets are grouped according to the distance of asset state information by means of clustering processing, the cost information, initial predicted costs, and asset state distance are associated by means of a training loss function, the cost allocation result and the asset state information difference between assets form a quantitative mapping, the dynamic adaptation of costs and asset state information is realized, the static rule limitation of traditional methods is overcome, and the flexibility of cost analysis is improved. BRIEF DESCRIPTION OF DRAWINGS
[0027] In order to more clearly illustrate the technical solutions in the embodiments of the application, the drawings needed in the embodiment description will be briefly introduced. Obviously, the drawings in the following description are only some embodiments of the application, and other drawings can be obtained by those skilled in the art without creative effort.
[0028] Figure 1 A flowchart of a single asset cost allocation and collection analysis method provided for the first embodiment of the present application is shown in the figure.
[0029] Figure 2 A structural schematic diagram of a single asset cost allocation and collection analysis device provided for the second embodiment of the present application is shown in the figure. DETAILED DESCRIPTION
[0030] The technical solutions in the embodiments of the present application will be clearly and completely described in combination with the accompanying drawings of the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments of the present application. Based on the embodiments of the present application, all other embodiments obtained by those skilled in the art without creative labor fall within the scope of protection of the present application.
[0031] It should be noted that the terms "first", "second", and the like in the specification and claims of the present application and the above-mentioned drawings are used to distinguish similar objects, and do not necessarily indicate a specific order or sequence. It can be understood that the above-mentioned terms for distinguishing similar objects can be interchanged under appropriate circumstances, so that the present application can also be implemented in other embodiments in addition to the above-illustrated embodiments or described embodiments. In addition, the terms "include" and "have" and any variations thereof are intended to cover non-exclusive inclusion, for example, a process, method, system, product, or server including a series of steps or units does not necessarily have to be limited to those steps or units clearly listed, but can include other steps or units not clearly listed or inherent to the process, method, product, or device.
[0032] Embodiment one
[0033] The first embodiment provides a single asset cost allocation and collection analysis method, as shown in the figure. Figure 1 A flowchart of a single asset cost allocation and collection analysis method provided for the first embodiment of the present application is shown in the figure. The single asset cost allocation and collection analysis method includes the following steps:
[0034] S101, obtaining a cost to be analyzed, wherein the cost to be analyzed includes cost information, a cost category, and M associated assets, M being a positive integer;
[0035] S102, if M = 1, then collecting the cost information into the cost category of a single associated asset, and if M ≠ 1, then obtaining asset state information corresponding to the M associated assets respectively;
[0036] S103, obtain dimension weight information set by a user, and calculate a first distance between asset state information respectively corresponding to each two associated assets according to the dimension weight information and the asset state information respectively corresponding to the M associated assets;
[0037] S104, perform clustering processing on the M associated assets according to all the first distances to obtain N clustering sets, wherein N is a positive integer;
[0038] S105, for any associated asset, input the cost category and asset state information corresponding to the associated asset into a pre-trained cost prediction model to obtain an initial predicted cost corresponding to the associated asset;
[0039] S106, calculate a training loss according to the cost information, the initial predicted cost respectively corresponding to the M associated assets, the asset state information respectively corresponding to the M associated assets, and the clustering set to which the M associated assets respectively belong;
[0040] S107, fine-tune the pre-trained cost prediction model according to the training loss until the training loss converges to obtain a trained cost prediction model;
[0041] S108, for any associated asset, input the cost category and asset state information corresponding to the associated asset into the trained cost prediction model to obtain a target predicted cost corresponding to the associated asset, and take the target predicted cost corresponding to the associated asset as a cost allocation cost of the cost category allocated to the associated asset.
[0042] The to-be-analyzed cost can refer to a part of the asset life cycle cost, the cost information can refer to a specific cost value, and the cost category can include maintenance cost, patrol cost, detection cost, routine maintenance cost, test cost, defect elimination cost, fault repair cost, and fault loss cost.
[0043] The to-be-analyzed cost can be represented in the form of a work order. According to the content of the work order, the associated asset corresponding to the to-be-analyzed cost can be determined. In this embodiment, a power grid scene is taken as an example for description. The associated asset can include a main transformer, a circuit breaker, a combined electric appliance, a disconnector, a current transformer, a voltage transformer, a reactor, a power capacitor, a coupling capacitor, a grounding transformer, a used transformer, a switch cabinet, a lightning arrester, and an arc extinction device.
[0044] When M = 1, it means that the cost information of the to-be-analyzed cost only has a single associated asset. In this case, the cost information can be directly collected into the cost category of the associated asset, which can be single maintenance or single detection of the associated asset.
[0045] When M≠1, it indicates that there are multiple associated assets of the cost information to be analyzed, and such a case can be common maintenance, common inspection, common detection, etc. of all associated assets, and the cost information needs to be allocated to the cost categories of each associated asset.
[0046] The pre-trained cost prediction model can adopt a regression prediction model, the architecture of the cost prediction model can adopt a fully connected layer, the asset state information can be represented by a 1×K size vector, the 1×1 size cost category and the asset state information are spliced into a 1×(K+1) size vector, input into the fully connected layer, and an initial predicted cost of 1×1 size is obtained by mapping.
[0047] Specifically, there is a state dimension representing the asset category in the asset state information, and the training data used for pre-training of the cost prediction model can be the asset state information of an asset at a historical time point and the allocated cost of the asset at the historical time point in a certain cost category. The asset state information of the asset at the historical time point and the cost category form a training sample, and the allocated cost of the asset at the historical time point in the cost category is used as a training label. According to a plurality of training samples and training labels, the cost prediction model is pre-trained by using a mean square error loss function until the mean square error loss function converges, and a pre-trained cost prediction model is obtained.
[0048] In a specific embodiment, the dimension weight information includes reference weights corresponding to K state dimensions respectively, and the asset state information includes sub-state information corresponding to the corresponding associated assets in K state dimensions respectively, K being a positive integer.
[0049] The first distance between the asset state information corresponding to each of the two associated assets is calculated according to the dimension weight information and the asset state information corresponding to the M associated assets, comprising:
[0050] For any two associated assets, the initial distance between the sub-state information corresponding to the two associated assets in each state dimension is calculated.
[0051] The initial distance in each state dimension is normalized respectively to obtain the normalized distance in each state dimension.
[0052] The normalized distance in each state dimension is multiplied by the reference weight corresponding to each state dimension respectively to obtain the weighted distance in each state dimension.
[0053] The weighted distances in each state dimension are added to obtain the first distance between the asset state information corresponding to the two associated assets.
[0054] The state dimensions can include asset category, average load, use time length, defect level, failure times, insulation aging degree, environment temperature, environment humidity, dust concentration, rated parameter, maintenance period, and load fluctuation degree.
[0055] The initial distance can be calculated by using the Euclidean distance, and the dimension weight information can be assigned by the implementer based on experience and actual scene for each state dimension. If the implementer does not set the dimension weight information, the reference weight corresponding to each state dimension is 1 by default.
[0056] The normalization processing can be to compare the initial distance between the two sub-state information used for calculation and the maximum value in the two sub-state information.
[0057] It should be noted that for the asset category state dimension, the normalization processing of the state dimension is obtained by using a mapping function, which can be r=1-exp(-1 / (t 2 )), where t is the initial distance between the two sub-state information corresponding to the asset category state dimension of the two associated assets, r is the normalized distance between the two sub-state information corresponding to the asset category state dimension of the two associated assets, exp() can be an exponential function. In this embodiment, the sub-state information corresponding to the asset category is obtained by encoding, the sub-state information corresponding to different asset categories is different, and the sub-state information corresponding to the same asset category is the same. When the sub-state information corresponding to the asset category state dimension of the two associated assets is the same, that is, the asset categories of the two associated assets are consistent, t=0, r tends to 0. When the sub-state information corresponding to the asset category state dimension of the two associated assets is different, that is, the asset categories of the two associated assets are inconsistent, t≠0, r tends to 1. It can be known that the mapping function approximately realizes the processing of the binary function, and the mapping function is derivable, which is convenient for direct application in the subsequent training process.
[0058] In a specific embodiment, the clustering processing of the M associated assets according to all the first distances to obtain N clustering sets includes:
[0059] According to all the first distances, the M associated assets are clustered by using a density-based spatial clustering algorithm to obtain N clustering sets;
[0060] For any associated asset, according to the clustering set to which the associated asset belongs, the clustering set identifier corresponding to the associated asset is determined;
[0061] Correspondingly, the training loss is calculated according to the cost information, initial predicted costs of the M associated assets respectively, asset state information of the M associated assets respectively, and cluster set identifiers of the M associated assets respectively.
[0062] The training loss is calculated according to the cost information, initial predicted costs of the M associated assets respectively, asset state information of the M associated assets respectively, and cluster set identifiers of the M associated assets respectively.
[0063] The density-based spatial clustering algorithm can adopt a density-based spatial clustering of applications with noise (DBSCAN).
[0064] The cluster set identifiers corresponding to different cluster sets are different.
[0065] In a specific embodiment, the training loss is calculated according to the cost information, initial predicted costs of the M associated assets respectively, asset state information of the M associated assets respectively, and cluster set identifiers of the M associated assets respectively, including:
[0066] A first sub-loss is calculated according to the cost information and initial predicted costs of the M associated assets respectively.
[0067] A second sub-loss is calculated according to initial predicted costs of the M associated assets respectively, asset state information of the M associated assets respectively, and cluster set identifiers of the M associated assets respectively.
[0068] The first sub-loss and the second sub-loss are weighted and summed, and the weighted and summed result is taken as the training loss.
[0069] The weight corresponding to the first sub-loss can be a large value, so that the first sub-loss is converged to 0 in the training process, thereby ensuring that the sum of the allocated costs does not exceed the cost information, nor is less than the cost information.
[0070] In a specific embodiment, the first sub-loss is calculated according to the cost information and initial predicted costs of the M associated assets respectively, including:
[0071] The initial predicted costs of the M associated assets respectively are added to obtain a first addition result.
[0072] An absolute value of a difference between the cost information and the first addition result is calculated to obtain a result as the first sub-loss.
[0073] The absolute value of the difference between the cost information and the first addition result can be a square of a difference between the cost information and the first addition result, and then a square root of the square is obtained, so as to ensure the derivability of the first sub-loss.
[0074] The first sub-loss is used to supervise a difference between the sum of the allocated costs and the cost information, and the greater the difference between the sum of the allocated costs and the cost information, the greater the first sub-loss.
[0075] In a specific implementation, the second sub-loss is calculated according to the initial predicted cost corresponding to each of the M associated assets, the asset state information corresponding to each of the M associated assets, and the cluster set identifier corresponding to each of the M associated assets.
[0076] For any two associated assets, an identification indication value is determined according to the cluster set identifier corresponding to each of the two associated assets.
[0077] A second distance is calculated according to the initial predicted cost corresponding to each of the two associated assets.
[0078] A temporary loss is calculated according to a first distance between the asset state information corresponding to each of the two associated assets, a second distance between the initial predicted cost corresponding to each of the two associated assets, and the identification indication value.
[0079] Each two associated assets are traversed to obtain a temporary loss corresponding to each two associated assets.
[0080] All temporary losses are added to obtain the second sub-loss.
[0081] The identification indication value can also be obtained by using the mapping function, that is, when the cluster set identifiers corresponding to each of the two associated assets are the same, the identification indication value is 0, and when the cluster set identifiers corresponding to each of the two associated assets are different, the identification indication value is 1.
[0082] The second distance can be an absolute value of a difference between the initial predicted cost corresponding to each of the two associated assets. Similarly, the absolute value of the difference between the initial predicted cost can be a square of a difference between the initial predicted cost, and then a square root of the square is obtained, so as to ensure the derivability of the second distance.
[0083] In a specific implementation, a calculation formula of the temporary loss l is as follows:
[0084] l=c×(dis1-a×f(dis2)) 2+(1-c) x (dis1-b x g(dis2)) 2 wherein dis1 is a first distance between asset state information respectively corresponding to two associated assets, dis2 is a second distance between initial predicted costs respectively corresponding to the two associated assets, c is the identification indication value, a is a first learnable parameter, b is a second learnable parameter, f() is a first nonlinear mapping function, g() is a second nonlinear mapping function, and l is a temporary loss corresponding to the two associated assets.
[0085] The first nonlinear mapping function can adopt an open 1 / 2 root function to provide a nonlinear mapping within the cluster set class, and the second nonlinear mapping function can adopt an open square root function to provide a stronger growth nonlinear mapping between the cluster set classes. The parameters a and b are also updated by using the gradient descent method, and can be initialized as 1.
[0086] The temporary loss can supervise that the target predicted cost difference respectively corresponding to the associated assets within the cluster set class is small, the target predicted cost difference respectively corresponding to the associated assets within the cluster set class is large, and the difference between the target predicted cost difference and the asset state information is positively correlated within or between the cluster set classes, so that the asset state information can still be learned by the cost prediction model to affect the target predicted cost without labeling, and compared with the prior art, the calculation of the training loss in the embodiment needs to combine the connection and difference between multiple associated assets, so that the prediction of the cost prediction model is more accurate and reliable, and each cost to be analyzed is fine-tuned based on the information of the associated assets, rather than using a general cost prediction model, so that the prediction can be more combined with the actual needs of the associated assets for cost allocation, and the accuracy of cost allocation prediction is improved.
[0087] In the first embodiment, the initial predicted cost is generated by the pre-trained cost prediction model combined with the asset state information, and the model training is optimized based on the distance of the clustering result and the asset state information, so that the final target predicted cost can reflect the state characteristics of a single asset and the cost commonality of similar assets, solve the distortion problem of cost allocation caused by ignoring asset heterogeneity in the traditional method, improve the accuracy of cost analysis, group assets according to the distance of asset state information by means of clustering processing, and associate the cost information, the initial predicted cost and the asset state distance by means of the training loss function, so that the cost allocation result and the asset state information difference between assets form a quantitative mapping, realize the dynamic adaptation of the cost and the asset state information, overcome the static rule limitation of the traditional method, and improve the flexibility of cost analysis.
[0088] Embodiment two
[0089] The second embodiment provides a single asset cost allocation and collection analysis device, as shown in Figure 2 The second embodiment provides a single asset cost allocation and collection analysis device, as shown in
[0090] The cost acquisition module 201 is configured to acquire a cost to be analyzed, wherein the cost to be analyzed includes cost information, a cost category, and M associated assets, and M is a positive integer.
[0091] The cost collection module 202 is configured to, if M = 1, collect the cost information into a cost category of a single associated asset, and if M ≠ 1, acquire asset state information corresponding to the M associated assets respectively.
[0092] The distance calculation module 203 is configured to acquire dimension weight information set by a user, and calculate a first distance between asset state information corresponding to each two associated assets according to the dimension weight information and the asset state information corresponding to the M associated assets respectively.
[0093] The asset clustering module 204 is configured to perform clustering processing on the M associated assets according to all the first distances, to obtain N clustering sets, wherein N is a positive integer.
[0094] The cost prediction module 205 is configured to, for any associated asset, input the cost category and asset state information corresponding to the associated asset into a pre-trained cost prediction model, to obtain an initial predicted cost corresponding to the associated asset.
[0095] The loss calculation module 206 is configured to calculate a training loss according to the cost information, initial predicted costs corresponding to the M associated assets respectively, asset state information corresponding to the M associated assets respectively, and clustering sets to which the M associated assets respectively belong.
[0096] The model fine-tuning module 207 is configured to fine-tune train the pre-trained cost prediction model according to the training loss until the training loss converges, to obtain a trained cost prediction model.
[0097] The cost allocation module 208 is configured to, for any associated asset, input the cost category and asset state information corresponding to the associated asset into the trained cost prediction model, to obtain a target predicted cost corresponding to the associated asset, and take the target predicted cost corresponding to the associated asset as an allocation cost of the cost to be analyzed allocated to the cost category of the associated asset.
[0098] It should be noted that the specific definition of the single asset cost allocation and collection analysis device can refer to the definition of the single asset cost allocation and collection analysis method in the above, and will not be repeated here. The information interaction, execution process and the like between the above modules, since the same concept as the method embodiment of the present application, the specific functions and the technical effects brought by it, can be referred to the method embodiment part, and will not be repeated here.
[0099] Embodiment three
[0100] The embodiment three provides a computer device, which can be a server. The computer device can include a processor, a memory, a network interface and a database connected through a system bus. Wherein, the processor of the computer device is used to provide computing and control capability. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operating system and the computer program in the non-volatile storage medium. The network interface of the computer device is used to communicate with the external terminal through the network connection. The computer program is executed by the processor to implement a single asset cost allocation and collection analysis method.
[0101] Embodiment four
[0102] The embodiment four provides a computer readable storage medium, and the computer readable storage medium stores a computer program. The computer program is executed by the processor to implement the single asset cost allocation and collection analysis method in the above embodiment. To avoid repetition, it will not be repeated here. Alternatively, the computer program is executed by the processor to implement the functions of each module / unit in the embodiment of the single asset cost allocation and collection analysis device. To avoid repetition, it will not be repeated here.
[0103] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program. The computer program can be stored in a non-volatile computer readable storage medium, and when executed, can include the processes of the above-mentioned embodiment methods. Any reference to memory, storage, database or other medium used in the embodiments provided in the present application can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. As an illustration but not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDR SDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), etc.
[0104] Those skilled in the art can clearly understand that, for the convenience and brevity of description, only the above-mentioned division of functional units and modules is exemplified, and in actual application, the above-mentioned functions can be completed by different functional units and modules according to needs, that is, the internal structure of the device is divided into different functional units or modules to complete all or part of the above-described functions.
[0105] The above is only a preferred embodiment of the present application, and does not limit the present application in any form. Although the present application has been disclosed as above with a preferred embodiment, it is not intended to limit the present application. Any skilled person in the art can make some changes or modifications to the above-mentioned disclosed technical content to obtain equivalent embodiments with equivalent changes, without departing from the technical solution of the present application. Any simple modification, equivalent change and modification of the above-mentioned embodiments according to the technical essence of the present application are still within the scope of the technical solution of the present application.
Claims
1. A method for cost allocation and aggregation analysis of individual assets, characterized in that, The method for cost allocation and aggregation analysis of individual assets includes the following steps: S101, Obtain the cost to be analyzed, wherein the cost to be analyzed includes cost information, cost category and M related assets, where M is a positive integer; S102, if M=1, the cost information is aggregated into the cost category of a single related asset; if M≠1, the asset status information corresponding to each of the M related assets is obtained. S103, obtain the dimension weight information set by the user, and calculate the first distance between the asset status information corresponding to each pair of related assets based on the dimension weight information and the asset status information corresponding to the M related assets respectively. S104, based on all the first distances, perform clustering processing on the M related assets to obtain N cluster sets, where N is a positive integer; S105, For any associated asset, input the cost category and the asset status information corresponding to the associated asset into the pre-trained cost prediction model to obtain the initial predicted cost corresponding to the associated asset. S106, The training loss is calculated based on the cost information, the initial predicted costs corresponding to the M related assets, the asset status information corresponding to the M related assets, and the cluster sets to which the M related assets belong. S107, Fine-tune the pre-trained cost prediction model according to the training loss until the training loss converges to obtain the trained cost prediction model. S108, for any related asset, input the cost category and the asset status information corresponding to the related asset into the trained cost prediction model to obtain the target predicted cost corresponding to the related asset, and use the target predicted cost corresponding to the related asset as the allocation cost of the cost to be analyzed to the cost category of the related asset.
2. The method for cost allocation and aggregation analysis of individual assets according to claim 1, characterized in that, The dimension weight information includes reference weights corresponding to K state dimensions, and the asset state information includes sub-state information corresponding to the associated asset under K state dimensions, where K is a positive integer. The step of calculating the first distance between the asset status information corresponding to each pair of related assets based on the dimension weight information and the asset status information corresponding to the M related assets includes: For any two related assets, calculate the initial distance between the sub-state information corresponding to the two related assets in each state dimension; The initial distance under each state dimension is normalized to obtain the normalized distance under each state dimension. The normalized distance under each state dimension is multiplied by the reference weight corresponding to each state dimension to obtain the weighted distance under each state dimension. The weighted distances under each state dimension are summed to obtain the first distance between the asset state information corresponding to the two related assets.
3. The method for cost allocation and aggregation analysis of individual assets according to claim 1, characterized in that, The M related assets are clustered based on all first distances to obtain N cluster sets, including: Based on all the first distances, a density-based spatial clustering algorithm is used to cluster the M related assets to obtain N cluster sets; For any associated asset, determine the cluster set identifier corresponding to the associated asset based on the cluster set to which the associated asset belongs; Accordingly, the step of calculating the training loss based on the cost information, the initial predicted costs corresponding to the M related assets, the asset status information corresponding to the M related assets, and the cluster sets to which the M related assets belong includes: The training loss is calculated based on the cost information, the initial predicted costs corresponding to the M related assets, the asset status information corresponding to the M related assets, and the cluster set identifiers corresponding to the M related assets.
4. The method for cost allocation and aggregation analysis of individual assets according to claim 3, characterized in that, The training loss is calculated based on the cost information, the initial predicted costs corresponding to the M related assets, the asset status information corresponding to the M related assets, and the cluster set identifiers corresponding to the M related assets, including: Based on the cost information and the initial predicted costs corresponding to the M related assets, the first sub-loss is calculated; The second sub-loss is calculated based on the initial predicted cost corresponding to each of the M related assets, the asset status information corresponding to each of the M related assets, and the cluster set identifier corresponding to each of the M related assets. The first sub-loss and the second sub-loss are weighted and summed, and the weighted sum is used as the training loss.
5. The method for cost allocation and aggregation analysis of individual assets according to claim 4, characterized in that, The calculation of the first sub-loss based on the cost information and the initial predicted costs corresponding to the M related assets includes: The initial predicted costs corresponding to the M related assets are added together to obtain the first summation result; Calculate the absolute value of the difference between the cost information and the first summation result, and use the calculated result as the first sub-loss.
6. The method for cost allocation and aggregation analysis of individual assets according to claim 4, characterized in that, The calculation of the second sub-loss based on the initial predicted costs corresponding to the M related assets, the asset status information corresponding to the M related assets, and the cluster set identifiers corresponding to the M related assets includes: For any two related assets, determine the identifier value based on the cluster set identifiers corresponding to the two related assets respectively; The second distance is calculated based on the initial projected costs corresponding to the two related assets. The temporary loss is calculated based on the first distance between the asset status information corresponding to the two related assets, the second distance between the initial predicted costs corresponding to the two related assets, and the identification indication value. Iterate through every pair of related assets to obtain the temporary loss corresponding to each pair of related assets; Adding all the temporary losses together gives the second sub-loss.
7. The method for cost allocation and aggregation analysis of individual assets according to claim 6, characterized in that, The formula for calculating the temporary loss l is: l = c × (dis1 - a × f(dis2)) 2 +(1-c)×(dis1-b×g(dis2)) 2 Where, dis1 is the first distance between the asset status information corresponding to the two related assets, dis2 is the second distance between the initial predicted costs corresponding to the two related assets, c is the identifier indication value, a is the first learnable parameter, b is the second learnable parameter, f() is the first nonlinear mapping function, g() is the second nonlinear mapping function, and l is the temporary loss corresponding to the two related assets.
8. A device for cost allocation and aggregation analysis of individual assets, characterized in that, The individual asset cost allocation and aggregation analysis device includes: The cost acquisition module is used to acquire the cost to be analyzed, wherein the cost to be analyzed includes cost information, cost category and M related assets, where M is a positive integer; The cost aggregation module is used to aggregate the cost information into the cost category of a single related asset if M=1, and to obtain the asset status information corresponding to M related assets if M≠1. The distance calculation module is used to obtain the dimension weight information set by the user, and calculate the first distance between the asset status information corresponding to each pair of related assets based on the dimension weight information and the asset status information corresponding to the M related assets. The asset clustering module is used to cluster the M related assets based on all first distances to obtain N cluster sets, where N is a positive integer; The cost prediction module is used to input the cost category and the asset status information corresponding to the associated asset into a pre-trained cost prediction model for any associated asset to obtain the initial predicted cost corresponding to the associated asset. The loss calculation module is used to calculate the training loss based on the cost information, the initial predicted costs corresponding to the M related assets, the asset status information corresponding to the M related assets, and the cluster sets to which the M related assets belong. The model fine-tuning module is used to fine-tune the pre-trained cost prediction model based on the training loss until the training loss converges, thereby obtaining a trained cost prediction model. The cost allocation module is used to input the cost category and the asset status information corresponding to the associated asset into the trained cost prediction model for any associated asset, to obtain the target predicted cost corresponding to the associated asset, and to use the target predicted cost corresponding to the associated asset as the allocation cost of the cost to be analyzed to the cost category of the associated asset.
9. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the single asset cost allocation and aggregation analysis method according to any one of claims 1 to 7.
10. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by the processor, it implements the single asset cost allocation and aggregation analysis method according to any one of claims 1 to 7.
Citation Information
Patent Citations
Client risk identification method and device
CN113177840A
Ffeasibility cloud computing analysis method
CN116776240A