Engineering cost audit analysis method and system based on big data

Abstract

The present application provides a method and system for auditing and analyzing engineering cost based on big data. By debugging the auditing and analyzing neural network to obtain abnormal reasoning information, associated parameter reasoning information, and noise data reasoning information, the auditing and analyzing neural network can be enabled to have the ability to identify and determine noise data in engineering cost sample data, identify and determine abnormal data classification of engineering cost data, and identify and determine associated parameters corresponding to engineering cost data involved in the expression of abnormal data classification, so as to accurately analyze engineering cost data containing noise data and improve the network performance of the auditing and analyzing neural network. Furthermore, by coordinating the debugging of the network through each branch task, it helps to learn the potential involvement relationship of the results of different branch tasks, so as to increase the performance of the auditing and analyzing neural network, so that the obtained target auditing and analyzing neural network can accurately analyze the engineering cost data containing noise data.

Description

Technical Field

[0001] The present application relates to but is not limited to the field of data processing technology, and in particular to a method and system for engineering cost audit analysis based on big data. Background Art

[0002] Project cost audit is the process of reviewing and evaluating the cost, compliance and efficiency of engineering projects. Project cost audit involves a large amount of data, including contracts, payment lists, cost data, project progress, etc. These data usually have various formats and structures. Traditional manual processing is inefficient and prone to errors. In addition, project cost audit needs to quickly and accurately detect cost anomalies, compliance risks and other problems so that timely measures can be taken to reduce risks. Artificial intelligence technology can help auditors quickly identify common anomalies in various engineering costs through data mining, machine learning and other means, and provide early warnings and suggestions. However, the noise data generated in the engineering cost data due to various reasons may cause large deviations in the results of data processing and analysis. How to accurately identify abnormal data in this situation is a technical issue that needs to be considered. Summary of the invention

[0003] In view of this, the embodiments of the present application at least provide a method and system for engineering cost audit analysis based on big data.

[0004] The technical solution of the embodiment of the present application is implemented as follows:

[0005] On the one hand, the embodiment of the present application provides a project cost audit analysis method based on big data, which is characterized by comprising:

[0006] Obtain the engineering cost data to be processed;

[0007] By using the previously debugged target audit analysis neural network, abnormal reasoning information, associated parameter reasoning information, and noise data distribution reasoning information are obtained based on the engineering cost data analysis;

[0008] Determine the abnormal data classification corresponding to the engineering cost data according to the abnormal reasoning information, and determine the engineering cost data related to the abnormal data classification in the engineering cost data when the noise data disturbance is removed through the noise data distribution reasoning information and the associated parameter reasoning information;

[0009] The debugging process of the target audit analysis neural network includes:

[0010] Based on a preset learning sample data set, the audit analysis neural network to be debugged is repeatedly updated and debugged to obtain a debugged target audit analysis neural network; wherein part of the learning sample data in the learning sample data set includes engineering cost sample data containing noise data and corresponding supervision indication information; and, in each update and debugging, including:

[0011] Perform embedding mapping operation on the acquired engineering cost sample data to obtain the corresponding data embedding representation vector;

[0012] Based on the data embedding representation vector, anomaly identification is performed on the engineering cost sample data to obtain anomaly reasoning information indicating anomaly data classification of the engineering cost sample data;

[0013] Based on the data embedding characterization vector, the engineering cost sample data is subjected to association parameter identification to obtain association parameter inference information indicating association parameters corresponding to respective data items in the engineering cost sample data;

[0014] Based on the data embedding representation vector, performing noise data recognition on the data embedding representation vector to obtain noise data inference information for annotating the noise data from the engineering cost sample data;

[0015] Based on the abnormal reasoning information, the associated parameter reasoning information and the noise data reasoning information, as well as the identification errors between them and the corresponding supervisory indication information, the audit analysis neural network is optimized for network learnable variables.

[0016] In some embodiments, the step of performing noise data identification on the data embedding representation vector based on the data embedding representation vector to obtain noise data inference information for annotating the noise data from the engineering cost sample data includes:

[0017] Through the noise data recognition layer in the audit analysis neural network, based on the data embedding representation vector and the data item embedding representation vector of each data item in the engineering cost sample data, the sub-inference information corresponding to each data item is respectively identified and determined;

[0018] Based on the sub-inference information corresponding to each of the data items, noise data inference information for annotating the noise data in the engineering cost sample data is obtained.

[0019] In some embodiments, the embedding mapping operation is performed on the acquired engineering cost sample data to obtain a corresponding data embedding representation vector, including:

[0020] Through the data embedding mapping layer pre-trained in the audit analysis neural network, embedding mapping operations are performed on each data item included in the engineering cost sample data to obtain a data embedding representation vector composed of the embedding representation vectors of each data item, wherein the data embedding mapping layer is obtained through multiple rounds of pre-debugging of various types of engineering cost data.

[0021] In some embodiments, each supervisory indication information corresponding to the engineering cost sample data includes abnormal identification indication information, associated parameter identification indication information and noise data identification indication information; the audit analysis neural network is optimized for network learnable variables based on the abnormal reasoning information, the associated parameter reasoning information and the noise data identification reasoning information, and the recognition errors between them and the corresponding supervisory indication information, including:

[0022] Obtaining an abnormality identification cost based on an identification error between the abnormality reasoning information and the corresponding abnormality identification indication information;

[0023] Obtaining an associated parameter identification cost based on an identification error between the associated parameter inference information and the corresponding associated parameter identification indication information;

[0024] Obtaining a noise data recognition cost based on a recognition error between the noise data inference information and the corresponding noise data recognition indication information;

[0025] Based on the anomaly identification cost, the associated parameter identification cost, and the noise data identification cost, the audit analysis neural network is optimized for network learnable variables.

[0026] In some embodiments, the process of generating the learning sample data set includes:

[0027] Acquire an initial engineering cost data set, and acquire each initial engineering cost data that needs to be integrated with noise data from the initial engineering cost data set according to a preset selection ratio, wherein the initial engineering cost data does not include noise data, and each initial engineering cost data is associated with corresponding initial abnormality identification indication information and initial associated parameter indication information;

[0028] According to the preset noise data integration strategy, noise data is integrated into each acquired initial engineering cost data to obtain engineering cost sample data, and the abnormality identification indication information, associated parameter identification indication information and noise data identification indication information corresponding to each engineering cost sample data are annotated respectively.

[0029] In some embodiments, the associated parameter identification indication information corresponding to each engineering cost sample data is annotated, including:

[0030] For each project cost sample data, including:

[0031] Based on the initial correlation parameter indication information corresponding to the engineering cost sample data, determining the correlation parameters respectively corresponding to the adjacent data items before and after the integrated noise data in the engineering cost sample data;

[0032] According to the associated parameters corresponding to each adjacent data item, the associated parameter identification result corresponding to the noise data is configured, and the corresponding associated parameter identification indication information is obtained through the associated parameter identification results corresponding to each data item in the engineering cost sample data.

[0033] In some embodiments, the initial associated parameter indication information indicates one or more associated parameter ranges, and the associated parameter ranges are used to annotate a classification of engineering cost data involved in describing a classification of abnormal data in the initial engineering cost data;

[0034] The configuring the association parameter identification result corresponding to the noise data according to the association parameters respectively corresponding to the adjacent data items includes:

[0035] If it is determined that the association parameters of the adjacent data items correspond to the same association parameter range, setting the association parameter identification result corresponding to the noise data to represent the association parameter located in the association parameter range;

[0036] If it is determined that the association parameters of the adjacent data items do not correspond to the same association parameter range, the association parameter identification result corresponding to the noise data is set to represent the association parameter that is not within the association parameter range.

[0037] In some embodiments, the process of generating the noise data identification indication information corresponding to each engineering cost sample data includes:

[0038] For each engineering cost sample data, determining the data item distribution positions corresponding to the engineering cost data of the initial engineering cost data and the incorporated noise data in the engineering cost sample data;

[0039] Through different distribution position annotation strategies, the distribution positions of data items corresponding to the initial engineering cost data are annotated in the engineering cost sample data, and the distribution positions of data items corresponding to the noise data are annotated to obtain the identification indication information of the noise data corresponding to the engineering cost sample data.

[0040] In some embodiments, according to a preset noise data integration strategy, noise data is integrated into each acquired initial engineering cost data to obtain engineering cost sample data, including:

[0041] For each initial project cost data, including:

[0042] Selecting one or more noise data items from a preset noise database to determine as noise data to be integrated;

[0043] According to a distribution position of a target data item determined arbitrarily in the initial construction cost data, determining a distribution position mark of the noise data in the initial construction cost data relative to the distribution position of the target data item;

[0044] Based on the integration distribution position mark, the noise data is integrated into the initial engineering cost data to obtain corresponding engineering cost sample data.

[0045] In a second aspect, the present application provides an audit analysis system, comprising a memory and a processor, wherein the memory stores a computer program executable on the processor, and the processor implements the steps in the above-described method when executing the program.

[0046] The beneficial effects of this application include at least:

[0047] The engineering cost audit analysis method and system based on big data provided in the embodiment of the present application can obtain abnormal reasoning information, associated parameter reasoning information, and noise data reasoning information by debugging the audit analysis neural network, so that the audit analysis neural network can have the ability to identify and determine noise data in engineering cost sample data, identify and determine abnormal data classification of engineering cost data, and identify and determine the associated parameters corresponding to the engineering cost data involved in the expression of abnormal data classification, so as to accurately analyze the engineering cost data containing noise data and improve the network performance of the audit analysis neural network; further, by debugging the audit analysis neural network, the abnormal identification branch task, the associated parameter identification branch task, and the noise data identification branch task are completed simultaneously, and the network is debugged collaboratively. For different branch tasks completed based on the same data embedding representation vector, collaborative debugging can help learn the potential involvement relationship of different branch task results, so as to increase the performance of the audit analysis neural network, so that the obtained target audit analysis neural network can accurately analyze the engineering cost data containing noise data.

[0048] It should be understood that the above general description and the following detailed description are merely exemplary and explanatory, and are not intended to limit the technical solutions of the present application. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] The drawings herein are incorporated into the specification and constitute a part of the specification. These drawings illustrate embodiments consistent with the present application and are used together with the specification to illustrate the technical solution of the present application.

[0050] Figure 1 and Figure 2A schematic diagram of the implementation flow of the debugging process of an audit analysis neural network provided in an embodiment of the present application.

[0051] Figure 3 A schematic diagram of the composition structure of an audit analysis device provided in an embodiment of the present application.

[0052] Figure 4 A hardware entity diagram of an audit analysis system provided in an embodiment of the present application. DETAILED DESCRIPTION

[0053] In order to make the purpose, technical solutions and advantages of the present application clearer, the technical solutions of the present application are further elaborated in detail below in conjunction with the drawings and embodiments. The described embodiments should not be regarded as limiting the present application. All other embodiments obtained by ordinary technicians in the field without making creative work are within the scope of protection of the present application.

[0054] In the following description, reference is made to "some embodiments", which describe a subset of all possible embodiments, but it is understood that "some embodiments" may be the same subset or different subsets of all possible embodiments, and may be combined with each other without conflict. The terms "first / second / third" involved are merely used to distinguish similar objects and do not represent a specific order for the objects. It is understood that "first / second / third" may be interchanged in a specific order or sequence where permitted, so that the embodiments of the present application described herein can be implemented in an order other than that illustrated or described herein.

[0055] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those commonly understood by those skilled in the art to which this application belongs. The terms used herein are only for the purpose of describing this application and are not intended to limit this application.

[0056] The embodiment of the present application provides a method for audit analysis of construction cost based on big data, which can be executed by a processor of an audit analysis system. The audit analysis system can refer to a server, a laptop, a tablet computer, a desktop computer, a smart TV, a mobile device (such as a mobile phone, a portable video player, a personal digital assistant, a dedicated messaging device, a portable gaming device) and other devices with data processing capabilities.

[0057] Figure 1 A schematic diagram of the implementation flow of the debugging process of the audit analysis neural network provided in the embodiment of the present application is as follows: Figure 1 As shown, the method comprises the following steps:

[0058] Step S110, generating a preset learning sample data set.

[0059] It can be understood that before debugging the audit analysis neural network to be debugged, generating a learning sample data set for debugging is the basis for network training. Specifically, a large amount of historical engineering cost data can be used as learning sample data for debugging, and each initial engineering cost data in the generated initial engineering cost data set does not contain noise data. At the same time, each initial engineering cost data is associated with corresponding initial abnormal identification indication information and initial associated parameter indication information. The initial engineering cost data in the initial engineering cost data set is processed to generate learning sample data. Among them, the types of noise data include, for example, data with irregular values ​​or formats, missing data, garbled data, repeated data, and other noise data types caused by human negligence.

[0060] When auditing and analyzing construction cost data, it is necessary to focus on abnormal information (also known as abnormal events), such as project cost anomalies, change event anomalies (such as excessive frequency, too many claims), cost structure anomalies, etc. During the audit analysis, if anomalies are identified, it is necessary to determine the distribution of specific data items associated with the identified anomalies, wherein the distribution of specific data items associated with the anomalies is determined to be the associated parameters in the embodiment of the present application, and the initial anomaly identification indication information and initial associated parameter indication information associated with the initial construction cost data are information indicating the anomalies of the initial construction cost data and the associated parameters corresponding to the anomalies, such as information that can be indicated by labels.

[0061] In the embodiment of the present application, generating a learning sample data set based on each initial engineering cost data may specifically include the following steps:

[0062] Step S111, obtaining an initial engineering cost data set, and obtaining each initial engineering cost data that needs to be integrated with noise data in the initial engineering cost data set according to a preset selection ratio.

[0063] As an implementation method, the historical engineering cost data of the enterprise can be determined as an initial engineering cost data set, and then according to a preset selection ratio, the initial engineering cost data corresponding to the selection ratio for constructing various types of engineering cost sample data can be extracted from the initial engineering cost data set, wherein the initial engineering cost data does not contain noise data, and each initial engineering cost data is associated with corresponding initial abnormality identification indication information and initial associated parameter indication information, and the specific value of the preset selection ratio is configured according to actual needs.

[0064] In this way, in the learning sample data set generated by the present application, part of the learning sample data in the learning sample data set includes engineering cost sample data containing noise data and corresponding supervisory indication information (i.e., various labels or marks used for supervised learning), and part of the learning sample data includes engineering cost sample data not containing noise data and corresponding supervisory indication information, wherein each supervisory indication information can specifically include one or more of abnormality identification indication information, associated parameter identification indication information and noise data identification indication information.

[0065] For example, it is assumed that the initial engineering cost data set includes 20,000 items of initial engineering cost data, that there is no noise data in the initial engineering cost data, and that the selection ratio is set to 0.5. Then, 50% of the initial engineering cost data are used to generate engineering cost sample data for debugging. Among the initial engineering cost data used to generate the engineering cost sample data, 80% of the initial engineering cost data are determined to generate engineering cost sample data that incorporate noise data, and 20% of the initial engineering cost data are used to generate engineering cost sample data that does not incorporate noise data. That is, 10,000 items of initial engineering cost data are arbitrarily determined in the initial engineering cost data set to generate engineering cost sample data, and among the 10,000 items of initial engineering cost data, 8,000 items of initial engineering cost data are determined to incorporate noise data, and the remaining 2,000 items of initial engineering cost data are determined to be engineering cost sample data.

[0066] Step S112, according to the preset noise data integration strategy, the noise data is integrated into each acquired initial engineering cost data to obtain engineering cost sample data, and the abnormality identification indication information, associated parameter identification indication information and noise data identification indication information corresponding to each engineering cost sample data are respectively annotated.

[0067] In an embodiment of the present application, when noise data is integrated into each initial engineering cost data according to a preset noise data integration strategy (i.e., how to integrate it) to obtain engineering cost sample data, considering that each engineering cost sample data may include the same type of noise data, or each engineering cost sample data may include different types of noise data, then when the initial engineering cost data is denoised to obtain the engineering cost sample data, the noise data item is integrated into the initial engineering cost data to obtain the engineering cost sample data.

[0068] When generating construction cost sample data by integrating noise data items, for each initial construction cost data, one or more noise data items are selected from a preset noise database and determined as noise data to be integrated. Then, based on the distribution position of a target data item arbitrarily determined in the initial construction cost data (the data item is the basic component unit of the initial construction cost data), the integration distribution position mark (i.e., the information on the position of the mark) of the noise data in the initial construction cost data relative to the distribution position of the target data item is determined. Based on the integration distribution position mark, the noise data is integrated into the initial construction cost data to obtain the corresponding construction cost sample data.

[0069] In the above method of generating engineering cost sample data by arbitrarily integrating noise data items into the initial engineering cost data, a preset noise database is generated in advance according to the available noise data items, and then the noise data items are integrated by arbitrarily selecting the integration position of the noise data items in the initial engineering cost data to generate the corresponding engineering cost sample data. Wherein, when determining the distribution position of the target data item, optionally, based on actual needs, the data item distribution position can be numbered according to the distribution order of different data items in the initial engineering cost data, and the target data item distribution position can be arbitrarily selected based on the arbitrarily selected data item distribution position number. In addition, the target data item distribution position selected based on actual needs can be one or more, so the number of noise data items arbitrarily determined in the noise database can be one or more.

[0070] In this way, it is possible to generate engineering cost sample data containing noisy data by arbitrarily integrating noise data items into the complete and noise-free initial engineering cost data, thereby completing the generation of engineering cost sample data containing noisy data and helping to debug based on learning sample data containing noise data.

[0071] The noise data incorporated may be common noise type data such as repeated noise data, missing noise data (eg, data items at corresponding positions are deleted), and erroneous noise data, without specific limitation.

[0072] Furthermore, the embodiment of the present application annotates the abnormality identification indication information, associated parameter identification indication information and noise data identification indication information corresponding to each engineering cost sample data, respectively. The engineering cost sample data is constructed based on the initial engineering cost data associated with the initial abnormality identification indication information and the initial associated parameter indication information.

[0073] The abnormality identification indication information corresponding to the engineering cost sample data may be set to be the same as the initial abnormality identification indication information of the corresponding initial engineering cost data.

[0074] For the associated parameter identification indication information corresponding to the engineering cost sample data, when annotating the associated parameter identification indication information corresponding to each engineering cost sample data, for each engineering cost sample data, it includes: based on the initial associated parameter indication information corresponding to the engineering cost sample data, determining the associated parameters corresponding to the adjacent data items before and after the integrated noise data in the engineering cost sample data, and then setting the associated parameter identification results corresponding to the noise data according to the associated parameters corresponding to each adjacent data item, and obtaining the corresponding associated parameter identification indication information through the associated parameter identification results corresponding to each data item in the engineering cost sample data. Specifically, when setting the associated parameter identification indication information corresponding to the engineering cost sample data, adjustments are made based on the initial associated parameter indication information of the initial engineering cost data corresponding to the engineering cost sample data, and the newly integrated noise data is filled with corresponding associated targets, wherein, for the initial associated parameter indication information, the initial associated parameter indication information indicates different types of engineering cost data that generate abnormal events, and the initial associated parameter indication information indicates one or more associated parameter ranges, that is, indicating which data are distributed in which areas of the initial engineering cost data. In other words, an associated parameter range is used to annotate the initial engineering cost data to construct a classification of abnormal data classification, wherein a classification of engineering cost data is composed of one or more data items, and different types of engineering cost data are labeled with different associated parameters.

[0075] In the embodiment of the present application, the engineering cost sample data includes different types of engineering cost data, the engineering cost data is composed of data items, the number of associated targets corresponding to the engineering cost data is equal to the number of data items in the engineering cost data, a data item in the engineering cost data corresponds to an associated target, each associated target has a corresponding associated parameter, and the associated parameter can identify whether the corresponding data item is the content of the abnormal information. For example, the engineering cost data can be marked by BIO (Begin, Inside, Outside), specifically, the associated parameter O represents the data item of the non-associated parameter range, the associated parameter Bm represents the starting position in the associated parameter range m, and the associated parameter Im represents the remaining positions in the associated parameter range m. When setting the associated parameter identification indication information corresponding to the engineering cost sample data, the associated parameters corresponding to the adjacent data items before and after the noise data are respectively determined in the corresponding initial associated parameter indication information, and then the associated parameter identification result corresponding to the noise data is set based on the associated parameters corresponding to each adjacent data item. For example, when setting the associated parameter identification result corresponding to the noise data, if it is determined that the associated parameters of each adjacent data item correspond to the same associated parameter range, the associated parameter identification result corresponding to the noise data is set to represent the associated parameters within the associated parameter range; if it is determined that the associated parameters of each adjacent data item do not correspond to the same associated parameter range, the associated parameter identification result corresponding to the noise data is set to represent the associated parameters that are not within the associated parameter range.

[0076] In a specific example, suppose the initial project cost data is "unit price D, total price E, and tax rate F of project A's construction site B and project type C". If it is set through the associated parameters of the BIO strategy, in the initial associated parameter indication information, the associated parameter corresponding to the data item "project A" is O, the associated parameter corresponding to the data item "construction site B" is O, the associated parameter corresponding to the data item "project type C" is B_tpye, the associated parameter corresponding to the data item "unit price D" is B_price, the associated parameter corresponding to the data item "total price E" is I_price, and the associated parameter corresponding to the data item "tax rate F" is I_price, among which "unit price D, total price E, and tax rate F" are specific costs, which are within an associated parameter range, and "project type C" is within an associated parameter range, representing the project type.

[0077] Based on the above-mentioned assumed initial engineering cost data, the engineering cost sample data can be generated as "unit price D, error, total price E, tax rate F for construction site B and engineering type C of project A". For the adjacent data items before and after the noise data "error", the data items are "unit price D" and "total price E". At the same time, each adjacent data item corresponds to the same associated parameter range, then the associated parameter identification result corresponding to "error" is considered to be within the associated parameter range, that is, I_price. In this way, when the associated parameters corresponding to the noise data in the engineering cost sample data are set and the associated parameter identification indication information corresponding to the engineering cost sample data is generated, the associated parameters corresponding to the adjacent data items before and after the noise data in the initial associated parameter indication information corresponding to the initial engineering cost data are combined to determine whether the noise data is within the associated parameter range, and the associated parameter configuration of the noise data is accurately completed. Further, based on the associated parameter identification results corresponding to each data item in the engineering cost sample data, the associated parameter identification indication information corresponding to the engineering cost sample data is obtained.

[0078] Based on this, the initial associated parameter indication information associated with the initial engineering cost data can be used to effectively set the associated parameter identification indication information for the corresponding generated engineering cost sample data, which not only ensures the generation quality of the learning sample data, but also makes the generation of the learning sample data easier.

[0079] For the noise data identification indication information corresponding to the engineering cost sample data, for each engineering cost sample data, the data item distribution positions corresponding to the engineering cost data of the initial engineering cost data and the integrated noise data are determined in the engineering cost sample data, and then through different distribution position annotation strategies, the data item distribution positions corresponding to the initial engineering cost data are annotated in the engineering cost sample data, and the data item distribution positions corresponding to the noise data are annotated to obtain the noise data identification indication information corresponding to the engineering cost sample data.

[0080] In actual application, the engineering cost data and the integrated noise data in the engineering cost sample data belonging to the corresponding initial engineering cost data can be distinguishably marked (for example, the noise data item is assigned mark Y, and other data items are assigned mark N), so that the noise data identification indication information obtained based on the mark can be determined to determine the data item distribution position of the noise data in the engineering cost sample data, and to determine the data item distribution position of the non-noise data.

[0081] Based on this, the noise data and non-noise data in the engineering cost sample data are annotated in the set noise data identification indication information to provide a reference for the classification of noise data in the debugging process of the audit analysis neural network.

[0082] Step S120, based on a preset learning sample data set, repeatedly updating and debugging the audit analysis neural network to be debugged to obtain a debugged target audit analysis neural network.

[0083] In actual implementation, after obtaining the preset learning sample data set, the pre-built audit analysis neural network to be debugged is repeatedly updated and debugged through the preset learning sample data set to obtain the debugged target audit analysis neural network, wherein part of the learning sample data set includes engineering cost sample data containing noise data and corresponding supervision indication information. The built audit analysis neural network can include a data embedding mapping layer, an anomaly recognition layer, an associated parameter recognition layer, and a noise data recognition layer.

[0084] Among them, the data embedding mapping layer is used to obtain the input engineering cost sample data in the debugging phase, and obtain the corresponding data item embedding representation vector (that is, the vector representing the embedding feature of the data item) for each data item in the engineering cost sample data, that is, to complete the encoding process of the corresponding data item and map the high-dimensional data into a low-dimensional vector. Among them, the data embedding mapping layer can be constructed based on the Encoder component in the Transformer.

[0085] In an embodiment of the present application, the data embedding mapping layer can be a network that has been pre-trained (i.e., Pre-training, the specific method can refer to the prior art). When the data embedding mapping layer is constructed through the Transformer structure, self-supervisory debugging is performed through massive engineering cost data without supervisory indication information.

[0086] The anomaly identification layer is used to complete the anomaly identification branch task in the audit analysis neural network. The anomaly identification layer can specifically include an average downsampling module and an affine module (i.e., fully connected structure - FC). When the affine module performs classification, it combines the activation function and the classifier function (such as softmax) based on actual needs to obtain the classification result. In actual implementation, the data embedding mapping layer encodes each data item in the engineering cost sample data separately to obtain the data item embedding representation vector, outputs the data embedding representation vector composed of each data item embedding representation vector, inputs the data embedding representation vector into the anomaly identification layer, and averages the elements corresponding to the same dimensional position in each data item embedding representation vector in the average downsampling module to obtain the average representation vector of the engineering cost sample data. The affine module in the anomaly identification layer completes the anomaly identification prediction based on the average representation vector, and according to the preset anomaly classification, inference information of the engineering cost sample data on each anomaly classification is obtained, such as a probability distribution.

[0087] The associated parameter identification layer is used to complete the associated parameter identification branch task in the audit analysis neural network. The associated parameter identification layer includes an affine module. When the affine module is classified, it combines the activation function and the classifier function based on actual needs to obtain the classification result. In actual implementation, the data embedding mapping layer encodes each data item in the engineering cost sample data to obtain the data item embedding representation vector, outputs the data embedding representation vector composed of the embedding representation vectors of each data item, and classifies the embedding representation vectors of each data item through the associated parameter identification layer, estimates the associated parameters corresponding to each data item, and obtains the associated parameter reasoning information corresponding to the engineering cost sample data.

[0088] The noise data recognition layer is used to complete the noise data recognition branch task in the audit analysis neural network. The noise data recognition layer can include an affine module. When the affine module performs classification, it combines the activation function and the classifier function based on actual needs to obtain the classification result. In actual implementation, the data embedding mapping layer encodes each data item in the engineering cost sample data to obtain a data item embedding representation vector, outputs a data embedding representation vector composed of each data item embedding representation vector, and classifies each data item embedding representation vector through the noise data recognition layer to obtain the noise data inference information corresponding to each data item, and judges whether each data item is noise data.

[0089] Please refer to Figure 2 In the embodiment of the present application, each update and debugging may specifically include the following steps:

[0090] Step S121, performing an embedding mapping operation on the acquired engineering cost sample data to obtain a corresponding data embedding representation vector.

[0091] For example, by performing embedding and mapping operations on each data item included in the engineering cost sample data through the data embedding mapping layer that has been pre-trained in the audit analysis neural network, a data embedding representation vector composed of the embedding representation vectors of each data item is obtained, wherein the data embedding mapping layer is obtained through multiple pre-debugging of various types of engineering cost data.

[0092] Based on this, the debugging complexity of the audit analysis neural network can be reduced through the pre-trained data embedding mapping layer, so that in the debugging phase of the audit analysis neural network, only minor adjustments are needed for the data embedding mapping layer to meet the network quality requirements.

[0093] Step S122, performing anomaly identification on the engineering cost sample data based on the data embedding representation vector, and obtaining anomaly reasoning information indicating anomaly data classification of the engineering cost sample data.

[0094] For example, by auditing and analyzing the anomaly recognition layer in the neural network, processing is performed based on the data embedding representation vector, and the support of each preset anomaly recognition classification, such as probability, is inferred to obtain anomaly reasoning information.

[0095] Step S123 , based on the data embedding characterization vector, the associated parameters of the engineering cost sample data are identified to obtain associated parameter inference information indicating the associated parameters corresponding to each data item in the engineering cost sample data.

[0096] In specific implementation, the associated parameter identification layer in the audit analysis neural network is processed based on the data embedding representation vector, and the probability of each data item in the engineering cost data corresponding to each preset associated parameter identification category is determined respectively, and the corresponding associated parameter reasoning information is obtained based on the associated parameters corresponding to each data item in the engineering cost sample data. In the embodiment of the present application, the content expressing the abnormality corresponding to the engineering cost sample data can be determined in the engineering cost sample data based on the associated parameter reasoning information, that is, the engineering cost data constituting the classification expression of abnormal data in the engineering cost sample data can be annotated based on the associated parameter reasoning information.

[0097] Step S124, performing noise data identification on the data embedding representation vector based on the data embedding representation vector, and obtaining noise data inference information for annotating noise data from the engineering cost sample data.

[0098] In specific implementation, the noise data recognition layer in the audit analysis neural network can be used to identify and determine the sub-inference information corresponding to each data item based on the data embedding representation vector and the data item embedding representation vector of each data item in the engineering cost sample data. Based on the sub-inference information corresponding to each data item, the noise data reasoning information used to annotate the noise data in the engineering cost sample data can be obtained.

[0099] In an embodiment of the present application, based on actual needs, it is implemented to set various types of noise data identification categories for identifying and determining noise data, and respectively determine the support, such as probability, of each data item in the engineering cost data in each preset noise data identification category. After determining the sub-inference information corresponding to each data item, the noise data inference information is obtained based on the sub-inference information corresponding to each data item.

[0100] Based on this, by setting up a noise data recognition layer in the audit analysis neural network to complete the noise data recognition branch task, we can learn to identify noise data and non-noise data in the engineering cost sample data, providing a reference for abnormal analysis of engineering cost data with noise data.

[0101] Step S125, based on the abnormal reasoning information, the associated parameter reasoning information and the noise data reasoning information, as well as the recognition errors between them and the corresponding supervisory indication information, the audit analysis neural network is optimized for network learnable variables.

[0102] In an embodiment of the present application, after obtaining the abnormal reasoning information, associated parameter reasoning information and noise data reasoning information output by the audit analysis neural network, a cost value is obtained based on the recognition error between each reasoning information and the corresponding supervision indication information, and the learnable variables of the network, such as the network's internal weights, biases and other parameters, are corrected using the obtained cost value.

[0103] In the specific implementation, based on the recognition error between the abnormal reasoning information and the corresponding abnormal recognition indication information, the abnormal recognition cost is obtained; based on the recognition error between the associated parameter reasoning information and the corresponding associated parameter recognition indication information, the associated parameter recognition cost is obtained; based on the recognition error between the noise data reasoning information and the corresponding noise data recognition indication information, the noise data recognition cost is obtained; and then based on the abnormal recognition cost, the associated parameter recognition cost, and the noise data recognition cost, the network learnable variables of the audit analysis neural network are optimized.

[0104] The above steps S122 to S124 can be executed simultaneously. The present application debugs the audit analysis neural network through the generated engineering cost data. The three layers in the audit analysis neural network include three debugging tasks, and three costs are obtained accordingly, namely, the anomaly identification cost, the associated parameter identification cost and the noise data identification cost; the total cost of the audit analysis neural network is determined based on the three costs obtained (for example, the total cost is obtained by weighted summation), and then the network learnable variables of the audit analysis neural network are optimized according to the determined total cost. Optionally, the cost can be obtained based on a general cross entropy function.

[0105] In this way, the network learnable variables of each layer in the audit analysis neural network are modified according to the obtained total cost.

[0106] In this way, based on the cross-entropy cost between the reasoning information of each layer and the corresponding supervisory indication information, the total cost is obtained by weighted summation to optimize the network learnable variables, that is, to complete the coordinated optimization of the three branch tasks of anomaly identification, associated parameter identification, and noise data identification, which makes it easier to perceive the potential involvement relationship between the results of different branch tasks and helps improve the performance of the audit analysis neural network.

[0107] When it is determined that the debugging of the audit analysis neural network meets the preset debugging cutoff requirements, the audit analysis neural network obtained when the debugging cutoff requirements are met is determined as the debugged target audit analysis neural network, wherein the debugging cutoff requirements can be set based on actual needs, such as the number of debugging times meets the maximum set number of debugging times or the total cost does not change continuously. In this way, the debugged target audit analysis neural network that can complete the branch tasks of abnormality identification, associated parameter identification, and noise data identification is obtained.

[0108] After the audit analysis neural network is debugged, the embodiment of the present application further provides a specific implementation process of performing audit analysis of engineering cost data based on the audit analysis neural network, which may specifically include:

[0109] Step S210, obtaining the project cost data to be processed.

[0110] Step S220, through the previously debugged target audit analysis neural network, abnormal reasoning information, associated parameter reasoning information and noise data distribution reasoning information are obtained based on the engineering cost data analysis.

[0111] In specific implementation, after obtaining the engineering cost data to be processed, the target audit analysis neural network that has been debugged in advance is used to analyze the engineering cost data. Three branch tasks can be completed based on the engineering cost data to obtain three inference information, namely, abnormal inference information, associated parameter inference information, and noise data distribution inference information. Abnormal information (also known as abnormal events), such as project cost abnormalities, change event abnormalities (such as excessive frequency, too many claims), cost structure abnormalities, etc., during audit analysis, if an abnormality is identified, it is necessary to determine the distribution of specific data items associated with the identified abnormality, where the distribution of the specific data items associated with the abnormality is the associated parameter, and the noise data distribution is the location information of the corresponding noise information.

[0112] Step S230, determining the abnormal data classification corresponding to the engineering cost data based on the abnormal reasoning information, and determining the engineering cost data related to the abnormal data classification in the engineering cost data when the noise data disturbance is removed through the noise data distribution reasoning information and the associated parameter reasoning information.

[0113] After the target audit analysis neural network obtains abnormal reasoning information, associated parameter reasoning information, and noise data distribution reasoning information based on the engineering cost data to be processed, the abnormal data classification contained in the engineering cost data to be processed can be determined based on the abnormal reasoning information. At the same time, the distribution position of the noise data in the engineering cost data can be determined by combining the noise data distribution reasoning information and the associated parameter reasoning information. The data content associated with the abnormal data classification can be determined and its distribution position in the engineering cost data. Therefore, when the noise data disturbance is removed, the engineering cost data related to the abnormal data classification in the engineering cost data can be determined.

[0114] Based on the foregoing embodiments, an embodiment of the present application provides an audit analysis device, and each unit included in the device, and each module included in each unit, can be implemented by a processor in a computer device; of course, it can also be implemented by a specific logic circuit; in the implementation process, the processor can be a central processing unit (CPU), a microprocessor (MPU), a digital signal processor (DSP) or a field programmable gate array (FPGA), etc.

[0115] Figure 3 A schematic diagram of the structure of an audit analysis device provided in an embodiment of the present application is shown in FIG. Figure 3 As shown, the audit analysis device 300 includes:

[0116] The data acquisition module 310 is used to acquire the engineering cost data to be processed;

[0117] The data reasoning module 320 is used to obtain abnormal reasoning information, associated parameter reasoning information, and noise data distribution reasoning information based on the engineering cost data analysis through the target audit analysis neural network that has been debugged in advance;

[0118] A data locking module 330 is used to determine the abnormal data classification corresponding to the engineering cost data according to the abnormal reasoning information, and determine the engineering cost data related to the abnormal data classification in the engineering cost data when the noise data disturbance is removed through the noise data distribution reasoning information and the associated parameter reasoning information;

[0119] The network debugging module 340 is used to debug the target audit analysis neural network, and the specific process includes:

[0120] Based on a preset learning sample data set, the audit analysis neural network to be debugged is repeatedly updated and debugged to obtain a debugged target audit analysis neural network; wherein part of the learning sample data in the learning sample data set includes engineering cost sample data containing noise data and corresponding supervision indication information; and, in each update and debugging, including:

[0121] Perform embedding mapping operation on the acquired engineering cost sample data to obtain the corresponding data embedding representation vector;

[0122] Based on the data embedding representation vector, anomaly identification is performed on the engineering cost sample data to obtain anomaly reasoning information indicating anomaly data classification of the engineering cost sample data;

[0123] Based on the data embedding characterization vector, the engineering cost sample data is subjected to association parameter identification to obtain association parameter inference information indicating association parameters corresponding to respective data items in the engineering cost sample data;

[0124] Based on the data embedding representation vector, performing noise data recognition on the data embedding representation vector to obtain noise data inference information for annotating the noise data from the engineering cost sample data;

[0125] Based on the abnormal reasoning information, the associated parameter reasoning information and the noise data reasoning information, and the identification errors between them and the corresponding supervisory indication information, the audit analysis neural network is optimized for network learnable variables.

[0126] The description of the above device embodiment is similar to the description of the above method embodiment, and has similar beneficial effects as the method embodiment. In some embodiments, the functions or modules included in the device provided in the embodiment of the present application can be used to execute the method described in the above method embodiment. For technical details not disclosed in the device embodiment of the present application, please refer to the description of the method embodiment of the present application for understanding.

[0127] It should be noted that in the embodiment of the present application, if the above-mentioned engineering cost audit analysis method based on big data is implemented in the form of a software function module and sold or used as an independent product, it can also be stored in a computer-readable storage medium. Based on such an understanding, the technical solution of the embodiment of the present application is essentially or the part that contributes to the relevant technology can be embodied in the form of a software product, which is stored in a storage medium, including several instructions to enable a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the methods described in each embodiment of the present application. The aforementioned storage medium includes: various media that can store program codes, such as a U disk, a mobile hard disk, a read-only memory (ROM), a disk or an optical disk. In this way, the embodiment of the present application is not limited to any specific hardware, software or firmware, or any combination of hardware, software, and firmware.

[0128] An embodiment of the present application provides an audit analysis system, including a memory and a processor, wherein the memory stores a computer program that can be executed on the processor, and when the processor executes the program, some or all of the steps in the above method are implemented.

[0129] The embodiment of the present application provides a computer-readable storage medium on which a computer program is stored, and when the computer program is executed by a processor, some or all of the steps in the above method are implemented. The computer-readable storage medium can be transient or non-transient.

[0130] An embodiment of the present application provides a computer program, including a computer-readable code. When the computer-readable code is run in a computer device, a processor in the computer device executes some or all of the steps for implementing the above method.

[0131] The embodiment of the present application provides a computer program product, which includes a non-transitory computer-readable storage medium storing a computer program, and when the computer program is read and executed by a computer, some or all of the steps in the above method are implemented. The computer program product can be implemented specifically by hardware, software or a combination thereof. In some embodiments, the computer program product is specifically embodied as a computer storage medium, and in other embodiments, the computer program product is specifically embodied as a software product, such as a software development kit (SDK) and the like.

[0132] It should be noted here that the description of the various embodiments above tends to emphasize the differences between the various embodiments, and the same or similar aspects can be referenced to each other. The description of the above device, storage medium, computer program and computer program product embodiments is similar to the description of the above method embodiment, and has similar beneficial effects as the method embodiment. For technical details not disclosed in the embodiments of the device, storage medium, computer program and computer program product of this application, please refer to the description of the method embodiment of this application for understanding.

[0133] Figure 4 A hardware entity diagram of an audit analysis system provided in an embodiment of the present application, such as Figure 4 As shown, the hardware entity of the audit analysis system 1000 includes: a processor 1001 and a memory 1002, wherein the memory 1002 stores a computer program that can be run on the processor 1001, and the processor 1001 implements the steps in the method of any of the above embodiments when executing the program.

[0134] The memory 1002 stores a computer program that can be run on the processor. The memory 1002 is configured to store instructions and applications executable by the processor 1001. It can also cache data to be processed or processed by the processor 1001 and various modules in the audit analysis system 1000 (for example, image data, audio data, voice communication data, and video communication data). This can be achieved through flash memory (FLASH) or random access memory (Random Access Memory, RAM).

[0135] When the processor 1001 executes the program, the steps of any one of the above-mentioned engineering cost audit analysis methods based on big data are implemented. The processor 1001 generally controls the overall operation of the audit analysis system 1000.

[0136] An embodiment of the present application provides a computer storage medium, which stores one or more programs. The one or more programs can be executed by one or more processors to implement the steps of the engineering cost audit analysis method based on big data as in any of the above embodiments.

[0137] It should be noted here that the description of the above storage medium and device embodiments is similar to the description of the above method embodiments, and has similar beneficial effects as the method embodiments. For the technical details not disclosed in the storage medium and device embodiments of the present application, please refer to the description of the method embodiments of the present application for understanding. The above processor can be at least one of a target application integrated circuit (Application Specific Integrated Circuit, ASIC), a digital signal processor (Digital Signal Processor, DSP), a digital signal processing device (Digital Signal Processing Device, DSPD), a programmable logic device (Programmable Logic Device, PLD), a field programmable gate array (Field Programmable Gate Array, FPGA), a central processing unit (Central Processing Unit, CPU), a controller, a microcontroller, and a microprocessor. It can be understood that the electronic device that realizes the above processor function can also be other, and the embodiments of the present application are not specifically limited.

[0138] The above-mentioned computer storage medium / memory can be a read-only memory (ROM), a programmable read-only memory (PROM), an erasable programmable read-only memory (EPROM), an electrically erasable programmable read-only memory (EEPROM), a magnetic random access memory (FRAM), a flash memory (Flash Memory), a magnetic surface memory, an optical disc, or a compact disc read-only memory (CD-ROM) and the like; it can also be various terminals including one or any combination of the above-mentioned memories, such as mobile phones, computers, tablet devices, personal digital assistants, etc.

[0139] It should be understood that the "one embodiment" or "one embodiment" mentioned throughout the specification means that the specific features, structures or characteristics related to the embodiment are included in at least one embodiment of the present application. Therefore, "in one embodiment" or "in one embodiment" appearing in various places throughout the specification may not necessarily refer to the same embodiment. In addition, these specific features, structures or characteristics can be combined in one or more embodiments in any suitable manner. It should be understood that in various embodiments of the present application, the size of the sequence number of each step / process mentioned above does not mean the order of execution, and the execution order of each step / process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiment of the present application. The above-mentioned sequence number of the embodiment of the present application is only for description and does not represent the advantages and disadvantages of the embodiment. It should be noted that, in this article, the term "include", "comprise" or any other variant thereof is intended to cover non-exclusive inclusion, so that the process, method, article or device including a series of elements includes not only those elements, but also includes other elements that are not explicitly listed, or also includes elements inherent to such process, method, article or device. Without more constraints, an element defined by the phrase "comprising a..." does not exclude the existence of other identical elements in the process, method, article or apparatus comprising the element.

[0140] In the several embodiments provided in the present application, it should be understood that the disclosed devices and methods can be implemented in other ways. The device embodiments described above are only schematic. For example, the division of the units is only a logical function division. There may be other division methods in actual implementation, such as: multiple units or components can be combined, or can be integrated into another system, or some features can be ignored or not executed. In addition, the coupling, direct coupling, or communication connection between the components shown or discussed can be through some interfaces, and the indirect coupling or communication connection of the devices or units can be electrical, mechanical or other forms.

[0141] The units described above as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units; they may be located in one place or distributed on multiple network units; some or all of the units may be selected according to actual needs to achieve the purpose of the present embodiment.

[0142] In addition, all functional units in the embodiments of the present application may be integrated into one processing unit, or each unit may be a separate unit, or two or more units may be integrated into one unit; the above-mentioned integrated units may be implemented in the form of hardware or in the form of hardware plus software functional units.

[0143] A person skilled in the art can understand that all or part of the steps of implementing the above method embodiment can be completed by hardware related to program instructions, and the aforementioned program can be stored in a computer-readable storage medium. When the program is executed, it executes the steps of the above method embodiment; and the aforementioned storage medium includes: mobile storage devices, read-only memories (ROM), magnetic disks or optical disks, etc., various media that can store program codes.

[0144] Alternatively, if the above-mentioned integrated unit of the present application is implemented in the form of a software function module and sold or used as an independent product, it can also be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application can essentially or in other words, the part that contributes to the relevant technology can be embodied in the form of a software product, which is stored in a storage medium and includes a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the methods described in each embodiment of the present application. The aforementioned storage medium includes: various media that can store program codes, such as mobile storage devices, ROMs, magnetic disks, or optical disks.

[0145] The above is only an implementation method of the present application, but the protection scope of the present application is not limited thereto. Any technician familiar with the technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application.

Claims

1. A project cost audit analysis method based on big data, characterized in that: include: Obtain the engineering cost data to be processed; By using the previously debugged target audit analysis neural network, abnormal reasoning information, associated parameter reasoning information, and noise data distribution reasoning information are obtained based on the engineering cost data analysis; wherein the associated parameter represents the distribution of specific data items of abnormal associations identified in the audit analysis; Determine the abnormal data classification corresponding to the engineering cost data according to the abnormal reasoning information, and determine the engineering cost data related to the abnormal data classification in the engineering cost data when the noise data disturbance is removed through the noise data distribution reasoning information and the associated parameter reasoning information; The debugging process of the target audit analysis neural network includes: Based on a preset learning sample data set, the audit analysis neural network to be debugged is repeatedly updated and debugged to obtain a debugged target audit analysis neural network; wherein part of the learning sample data in the learning sample data set includes engineering cost sample data containing noise data and corresponding supervision indication information; and, in each update and debugging, including: Perform embedding mapping operation on the acquired engineering cost sample data to obtain the corresponding data embedding representation vector; Based on the data embedding representation vector, anomaly identification is performed on the engineering cost sample data to obtain anomaly reasoning information indicating anomaly data classification of the engineering cost sample data; Based on the data embedding characterization vector, the engineering cost sample data is subjected to association parameter identification to obtain association parameter inference information indicating association parameters corresponding to respective data items in the engineering cost sample data; Based on the data embedding representation vector, performing noise data recognition on the data embedding representation vector to obtain noise data inference information for annotating the noise data from the engineering cost sample data; Based on the abnormal reasoning information, the associated parameter reasoning information and the noise data reasoning information, as well as the identification errors between them and the corresponding supervisory indication information, the audit analysis neural network is optimized for network learnable variables.

2. The method according to claim 1, characterized in that The step of performing noise data identification on the data embedding representation vector based on the data embedding representation vector to obtain noise data inference information for annotating the noise data from the engineering cost sample data includes: Through the noise data recognition layer in the audit analysis neural network, based on the data embedding representation vector and the data item embedding representation vector of each data item in the engineering cost sample data, the sub-inference information corresponding to each data item is respectively identified and determined; Based on the sub-inference information corresponding to each of the data items, noise data inference information for annotating the noise data in the engineering cost sample data is obtained.

3. The method according to claim 1, characterized in that The embedding mapping operation is performed on the acquired engineering cost sample data to obtain the corresponding data embedding representation vector, including: Through the data embedding mapping layer pre-trained in the audit analysis neural network, embedding mapping operations are performed on each data item included in the engineering cost sample data to obtain a data embedding representation vector composed of the embedding representation vectors of each data item, wherein the data embedding mapping layer is obtained through multiple rounds of pre-debugging of various types of engineering cost data.

4. The method according to claim 1, characterized in that: Each supervisory indication information corresponding to the engineering cost sample data includes abnormal identification indication information, associated parameter identification indication information and noise data identification indication information; the audit analysis neural network is optimized for network learnable variables based on the abnormal reasoning information, the associated parameter reasoning information and the noise data identification reasoning information, and the recognition errors between them and the corresponding supervisory indication information, including: Obtaining an abnormality identification cost based on an identification error between the abnormality reasoning information and the corresponding abnormality identification indication information; Obtaining an associated parameter identification cost based on an identification error between the associated parameter inference information and the corresponding associated parameter identification indication information; Obtaining a noise data recognition cost based on a recognition error between the noise data inference information and the corresponding noise data recognition indication information; Based on the anomaly identification cost, the associated parameter identification cost, and the noise data identification cost, the audit analysis neural network is optimized for network learnable variables.

5. The method according to any one of claims 1 to 4, characterized in that: The generation process of the learning sample data set includes: Acquire an initial engineering cost data set, and acquire each initial engineering cost data that needs to be integrated with noise data from the initial engineering cost data set according to a preset selection ratio, wherein the initial engineering cost data does not include noise data, and each initial engineering cost data is associated with corresponding initial abnormality identification indication information and initial associated parameter indication information; According to the preset noise data integration strategy, noise data is integrated into each acquired initial engineering cost data to obtain engineering cost sample data, and the abnormality identification indication information, associated parameter identification indication information and noise data identification indication information corresponding to each engineering cost sample data are annotated respectively.

6. The method according to claim 5, characterized in that Annotate the associated parameter identification information corresponding to each engineering cost sample data, including: For each project cost sample data, including: Based on the initial correlation parameter indication information corresponding to the engineering cost sample data, determining the correlation parameters respectively corresponding to the adjacent data items before and after the integrated noise data in the engineering cost sample data; According to the associated parameters corresponding to each adjacent data item, the associated parameter identification result corresponding to the noise data is configured, and the corresponding associated parameter identification indication information is obtained through the associated parameter identification results corresponding to each data item in the engineering cost sample data.

7. The method according to claim 6, characterized in that The initial associated parameter indication information indicates one or more associated parameter ranges, and the associated parameter ranges are used to annotate a classification of engineering cost data involved in describing a classification of abnormal data in the initial engineering cost data; The configuring the association parameter identification result corresponding to the noise data according to the association parameters respectively corresponding to the adjacent data items includes: If it is determined that the association parameters of the adjacent data items correspond to the same association parameter range, setting the association parameter identification result corresponding to the noise data to represent the association parameter located in the association parameter range; If it is determined that the association parameters of the adjacent data items do not correspond to the same association parameter range, the association parameter identification result corresponding to the noise data is set to represent the association parameter that is not within the association parameter range.

8. The method according to claim 5, characterized in that The process of generating the noise data identification indication information corresponding to each engineering cost sample data includes: For each engineering cost sample data, determining the data item distribution positions corresponding to the engineering cost data of the initial engineering cost data and the incorporated noise data in the engineering cost sample data; Through different distribution position annotation strategies, the distribution positions of data items corresponding to the initial engineering cost data are annotated in the engineering cost sample data, and the distribution positions of data items corresponding to the noise data are annotated to obtain the identification indication information of the noise data corresponding to the engineering cost sample data.

9. The method according to claim 5, characterized in that According to the preset noise data integration strategy, noise data is integrated into each acquired initial engineering cost data to obtain engineering cost sample data, including: For each initial project cost data, including: Selecting one or more noise data items from a preset noise database to determine as noise data to be integrated; According to a distribution position of a target data item determined arbitrarily in the initial construction cost data, determining a distribution position mark of the noise data in the initial construction cost data relative to the distribution position of the target data item; Based on the integration distribution position mark, the noise data is integrated into the initial engineering cost data to obtain corresponding engineering cost sample data.

10. An audit analysis system, comprising a memory and a processor, wherein the memory stores a computer program that can be run on the processor, characterized in that: When the processor executes the program, the steps in the method according to any one of claims 1 to 9 are implemented.

Images