Engineering budget deviation correction method based on machine learning model

Abstract

The invention relates to the technical field of engineering budget prediction, in particular to an engineering budget deviation correction method based on a machine learning model. The method comprises the steps of calculating a deviation index of an engineering budget prediction model in a historical period based on a lag relative deviation, a lag numerical deviation and a lag trend deviation, inputting engineering budget data of a current time period into the engineering budget prediction model based on the deviation index, and outputting an initial engineering budget prediction result; calculating the importance weight of the point data corresponding to each sub-moment, sequentially embedding the weighted point data into the initial engineering budget prediction result according to the magnitude of the importance weight value, and generating a corresponding point data related engineering budget sequence in the current time period; according to the method, the trend compensation information of the project budget sequence related to the point data is calculated, the trend compensation value is determined based on the trend compensation information, and the prediction result is corrected based on the trend compensation value, so that a high-precision machine learning model can be constructed, and the project budget deviation is accurately corrected.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of engineering budget prediction, and particularly relates to an engineering budget deviation correction method based on a machine learning model. BACKGROUND

[0002] In engineering practice, budget deviation is a common problem. These deviations may be caused by material price fluctuations, labor cost changes, construction period extension, design changes, unforeseen geological conditions or environmental factors, etc. Budget deviation may not only lead to project cost overruns, but also affect project progress and quality, and even trigger project risks.

[0003] Traditional engineering budget deviation correction methods mainly rely on manual experience and simple statistical analysis, such as proportional adjustment method, experience estimation method, etc. These methods, although to some extent, can correct budget deviation, but have obvious limitations. For example, manual experience method is greatly influenced by personal experience and subjective judgment, and it is difficult to ensure the accuracy and consistency of the correction; simple statistical analysis method cannot fully consider the influence of various complex factors on budget deviation, and the correction result is often not accurate enough.

[0004] Therefore, it is urgent to develop an engineering budget deviation correction method based on a machine learning model to solve the above problems. SUMMARY

[0005] The present application aims to provide an engineering budget deviation correction method based on a machine learning model, which solves the technical problem of large error and inaccurate engineering budget in engineering budget prediction due to reliance on manual experience and simple statistical analysis.

[0006] The object of the present application can be achieved by the following technical solution: An engineering budget deviation correction method based on a machine learning model, the method comprising: Obtaining engineering budget data of several historical periods, analyzing the engineering budget data of the several historical periods to obtain lag relative deviation, lag numerical deviation and lag trend deviation of an engineering budget prediction model, and calculating a deviation index of the engineering budget prediction model in the historical period based on the lag relative deviation, the lag numerical deviation and the lag trend deviation, wherein the engineering budget prediction model is an LSTM network; Judging whether the engineering budget prediction model meets the current period engineering budget prediction requirement based on the deviation index, if yes, inputting the current period engineering budget data into the engineering budget prediction model to output an initial engineering budget prediction result; Obtaining point data corresponding to each sub-moment in the current period, calculating the importance weight of the point data corresponding to each sub-moment, embedding the weighted point data into the initial engineering budget prediction result in turn according to the importance weight value, and generating the corresponding point data related engineering budget sequence in the current period; Calculating the trend compensation information of the point data related engineering budget sequence, determining the trend compensation value based on the trend compensation information, and correcting the prediction result based on the trend compensation value.

[0007] Further, the lag relative deviation, lag value deviation and lag trend deviation of the engineering budget prediction model are obtained by analyzing the engineering budget data of a plurality of historical periods, and the specific process includes the following steps: Calculate the lag relative deviation LRE: ; Calculate the lag value deviation LVE: ; Calculate the lag trend deviation LTE: ; Wherein, represents the true engineering budget value of the i-th historical period, represents the engineering budget prediction value of the i-th historical period obtained by the engineering budget prediction model, represents the true engineering budget value input into the engineering budget prediction model of the i-th historical period, and N represents the number of historical periods.

[0008] Further, the deviation index of the engineering budget prediction model in the historical period is calculated based on the lag relative deviation, the lag value deviation and the lag trend deviation, and the specific process includes the following steps: Calculate the average value of the lag relative deviation, the lag value deviation and the lag trend deviation in each historical period; Take the execution time of the historical period as the X-axis and the average value as the Y-axis to construct a rectangular coordinate system, mark all the average values in the form of points in the rectangular coordinate system, and connect the adjacent points of the average values in the rectangular coordinate system to generate a deviation ratio curve; Draw a vertical line from both ends of the deviation ratio curve to the X-axis to obtain four starting and ending line segments, and form a closed figure by the deviation ratio curve, the two starting and ending line segments and the X-axis. Calculate the total area of the closed figure, and record the total area as the deviation index of the engineering budget prediction model in the historical period.

[0009] Further, whether the engineering budget prediction model meets the current period engineering budget prediction requirement is judged based on the deviation index, and the specific process includes the following steps: Load the deviation index threshold, and determine whether the deviation index exceeds the deviation index threshold. If it does, determine that the engineering budget prediction model does not meet the engineering budget prediction requirements for the current period. If not, determine that the engineering budget prediction model meets the engineering budget prediction requirements for the current period.

[0010] Furthermore, inputting the current period's project budget data into the project budget prediction model and outputting the initial project budget prediction results specifically includes the following processes: Extract the feature vector of the project budget data for the current period. , to feature vector The input is fed into the LSTM network to obtain the last state of the forward LSTM network. and the last state of the backward LSTM network : ; ; and These represent forward and backward state operations, respectively; Will , Concatenate them into a single vector : ;in, This indicates a concatenation operation; Will The input is fed into a fully connected network, and the predicted label for the current time period is output through a softmax function: ; This represents the probability distribution of the predicted value labels for the current time period. , Indicates training parameters; The objective function is: ; in, Indicates learnable parameters, Indicates the first The probability that the project budget for each time period belongs to the label of the k-th predicted value. The number of time periods. This represents the number of intervals within the predicted value range.

[0011] Furthermore, calculating the importance weight of the point data corresponding to each sub-time point specifically includes the following process: Point data includes information on increases in construction costs, installation costs, equipment and tool purchase costs, and other costs. Establish a hierarchical model with progressively higher levels, where the levels include the target layer, the criterion layer, and the measure layer; Constructing a judgment matrix: Determining the number of relevant influencing factors at each level. Construct a set of factors influencing importance weights , ,in, For the subset of construction project cost growth rate, For the subset of installation project cost growth rate, The subset representing the increase in equipment and tooling purchase costs. For a subset of other cost growth rates, from the set Two subsets at the same level are extracted and compared, using... This represents the ratio of importance, and assigns the corresponding importance values ​​according to a preset ratio, combining the importance of each layer to form a judgment matrix; Calculate the largest eigenvalue of the judgment matrix : ; in, It is a matrix obtained by normalizing each column vector of the judgment matrix. The value is 1, 2...b. For matrix The elements are added row by row to obtain a vector, which is then normalized into a matrix. For matrix The elements are added column by column to obtain a vector, which is then normalized into a matrix. Calculate the importance weight of the point data corresponding to each sub-time point. : ; in, This indicates the order of the judgment matrix.

[0012] Furthermore, the process of calculating trend compensation information for the engineering budget sequence related to the point data specifically includes the following steps: Obtain the relevant engineering budget sequence of point data ; By taking turns using the corresponding node at each sub-time point as the parent node, j dependency classifiers are constructed. For each dependent subclassifier, calculate the posterior probability of the engineering budget sample. : ;in, Indicates project budget trend categories Engineering budget sequence related to point data The maximum conditional probability. representing a project budget trend category in a project budget trend sequence of point data Probability of occurrence After obtaining each dependent sub-classifier posterior probability, calculate the posterior probability average value : J is the number of dependent classifiers Select the posterior probability average value The largest project budget trend category is taken as the trend compensation information, wherein the project budget trend category includes a long-term continuous growth trend category, a periodic fluctuation trend category and a project-specific fluctuation trend category.

[0013] Further, based on the trend compensation information, a trend compensation value is determined, and the prediction result is corrected based on the trend compensation value, which specifically includes the following processes: Based on the trend compensation information, a trend compensation value is queried, and each project budget trend category is preset with a trend compensation value, and the prediction result is corrected based on the trend compensation value.

[0014] Compared with the prior art, the present application has the following beneficial effects: The present application analyzes the project budget data of several historical periods, obtains the lag relative deviation, lag numerical deviation and lag trend deviation of the project budget prediction model, calculates the deviation index of the project budget prediction model in the historical period, inputs the project budget data of the current period into the project budget prediction model based on the deviation index, and outputs the initial project budget prediction result; the importance weight of the point data corresponding to each sub-time is calculated, the weighted point data is sequentially embedded into the initial project budget prediction result according to the high and low values of the importance weight, and the corresponding point data related project budget sequence in the current period is generated; the trend compensation information of the point data related project budget sequence is calculated, the trend compensation value is determined based on the trend compensation information, and the prediction result is corrected based on the trend compensation value, so that a high-precision machine learning model can be constructed, the project budget deviation can be accurately corrected, the precision of the project budget can be improved, and the project progress and quality can be ensured. BRIEF DESCRIPTION OF DRAWINGS

[0015] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the drawings needed in the embodiments will be briefly introduced below. Obviously, the drawings in the following description are only some embodiments described in the present application, and other drawings can also be obtained by those skilled in the art based on these drawings.

[0016] Figure 1 is the workflow diagram of the first project budget deviation correction method based on the machine learning model in the embodiment of the present application; Figure 2is a workflow diagram of a second engineering budget deviation correction method based on a machine learning model according to an embodiment of the present application. DETAILED DESCRIPTION

[0017] The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments of the present application. Based on the embodiments in the present application, all other embodiments obtained by a person of ordinary skill in the art without creative work fall within the protection scope of the present application.

[0018] In addition, the described features, structures, or characteristics can be combined in any suitable manner in one or more example embodiments. In the following description, numerous specific details are provided to give a thorough understanding of example embodiments of the present disclosure. One skilled in the relevant art will recognize, however, that the techniques of the present disclosure can be practiced without one or more of the specific details, or with other methods, components, materials, and so forth. In other instances, well-known structures, materials, or operations are not shown or described in detail in order to avoid obscuring aspects of the present disclosure.

[0019] The embodiment provides a method for correcting deviation of engineering budget based on a machine learning model, Figure 1 is a workflow diagram of a first engineering budget deviation correction method based on a machine learning model according to an embodiment of the present application, as Figure 1 shown, the method comprises the following steps: Step S101: Obtain engineering budget data of a plurality of historical periods, analyze the engineering budget data of the plurality of historical periods, and obtain lag relative deviation, lag numerical deviation and lag trend deviation of an engineering budget prediction model; It is worth noting that the engineering budget prediction model is a bidirectional LSTM network. Step S102: Calculate the deviation index of the engineering budget prediction model in the historical period based on the lag relative deviation, the lag numerical deviation and the lag trend deviation; Step S103: Determine whether the engineering budget prediction model meets the current period engineering budget prediction requirement based on the deviation index, if yes, input the engineering budget data of the current period into the engineering budget prediction model, and output an initial engineering budget prediction result; It is worth noting that if the engineering budget prediction model does not meet the current period engineering budget prediction requirement, the engineering budget data of the current period is rejected to be input into the engineering budget prediction model. Step S104: obtaining point data corresponding to each sub-moment in the current period, calculating the importance weight of the point data corresponding to each sub-moment, embedding the weighted point data into the initial engineering budget prediction result in turn according to the high and low of the importance weight value, and generating the point data related engineering budget sequence corresponding to the current period; Step S105: calculating the trend compensation information of the point data related engineering budget sequence, determining the trend compensation value based on the trend compensation information, and correcting the prediction result based on the trend compensation value.

[0020] In summary, the engineering budget data of several historical periods is analyzed to obtain the lag relative deviation, lag value deviation and lag trend deviation of the engineering budget prediction model, the deviation index of the engineering budget prediction model in the historical period is calculated, the engineering budget data of the current period is input into the engineering budget prediction model based on the deviation index, and the initial engineering budget prediction result is output; the importance weight of the point data corresponding to each sub-moment is calculated, the weighted point data is embedded into the initial engineering budget prediction result in turn according to the high and low of the importance weight value, and the point data related engineering budget sequence corresponding to the current period is generated; the trend compensation information of the point data related engineering budget sequence is calculated, the trend compensation value is determined based on the trend compensation information, and the prediction result is corrected based on the trend compensation value, so that a high-precision machine learning model can be constructed, the engineering budget deviation can be accurately corrected, the precision of the engineering budget can be improved, and the project progress and quality can be ensured.

[0021] In some embodiments, analyzing the engineering budget data of several historical periods to obtain the lag relative deviation, lag value deviation and lag trend deviation of the engineering budget prediction model specifically includes the following processes:

[0022] Calculate the lag relative deviation LRE:

[0023] ; Calculate the lag value deviation LVE: ; Calculate the lag trend deviation LTE: ; Wherein, represents the true engineering budget value of the i th historical period, represents the engineering budget prediction value of the i th historical period obtained by the engineering budget prediction model, represents the true engineering budget value input into the engineering budget prediction model of the i th historical period, and N represents the number of historical periods.

[0024] In some embodiments, Figure 2is a workflow diagram of a second engineering budget deviation correction method based on a machine learning model according to an embodiment of the present application, as Figure 2 The deviation index of the engineering budget prediction model in the historical period is calculated based on the lag relative deviation, the lag numerical deviation, and the lag trend deviation, and specifically includes the following steps: Step S201: average the lag relative deviation, the lag numerical deviation, and the lag trend deviation in each historical period; Step S202: construct a rectangular coordinate system with the execution time of the historical period as the X-axis and the average value as the Y-axis, mark all the average values in the form of points in the rectangular coordinate system, and connect the adjacent points of the average values in the rectangular coordinate system to generate a deviation multiple curve; Step S203: draw a perpendicular line from both ends of the deviation multiple curve to the X-axis to obtain four start-stop line segments, and form a closed figure with the deviation multiple curve, the two start-stop line segments, and the X-axis. Calculate the total area of the closed figure, and record the total area as the deviation index of the engineering budget prediction model in the historical period.

[0025] In some embodiments, determining whether the engineering budget prediction model meets the current period engineering budget prediction requirement based on the deviation index specifically includes the following process: Load the deviation index threshold, determine whether the deviation index exceeds the deviation index threshold, if yes, determine that the engineering budget prediction model does not meet the current period engineering budget prediction requirement, and if no, determine that the engineering budget prediction model meets the current period engineering budget prediction requirement.

[0026] In some embodiments, inputting the engineering budget data of the current period into the engineering budget prediction model to output an initial engineering budget prediction result specifically includes the following process: extracting the feature vector of the engineering budget data of the current period , inputting the feature vector into the LSTM network to obtain the last state of the forward LSTM network and the last state of the backward LSTM network : ; ; and represent the forward state and the backward state operation, respectively; concatenate , into a vector : ; wherein represents the concatenation operation; inputting The input is fed into a fully connected network, and the predicted value label of the current period is output through a softmax function: ; The probability distribution of the predicted value label of the current period is represented by , The training parameters are represented by The objective function is: ; wherein The learnable parameters are represented by The probability that the engineering budget of the i-th period belongs to the k-th predicted value label is represented by The number of periods is represented by The number of prediction value range intervals is represented by .

[0027] In some embodiments, calculating the importance weight of the point data corresponding to each sub-time point specifically includes the following process: The point data includes building engineering cost growth information, installation engineering cost growth information, equipment and tool purchase cost growth information, and other cost growth information; specifically, the building engineering cost growth information includes: Labor cost: various fees paid to production workers directly engaged in building and installation engineering construction, such as basic wages, wage subsidies, production worker auxiliary wages, employee welfare fees, and production worker labor protection fees.

[0028] Material cost: the cost of raw materials, auxiliary materials, components, parts, and semi-finished products that constitute the engineering entity during the construction process, as well as the amortization (or rental) cost of turnover materials.

[0029] Construction machinery usage fee: mechanical usage fee and mechanical installation and disassembly fee and off-site transportation fee incurred by construction machinery operation, including depreciation fee, major repair fee, regular repair fee, installation and disassembly fee and off-site transportation fee, labor cost, fuel and power cost, road maintenance fee, and vehicle and ship usage tax, etc.

[0030] Measures: fees incurred before and during the construction of the project for non-engineering entity projects, such as environmental protection fees, civilized construction fees, safe construction fees, temporary facility fees, night construction fees, secondary handling fees, large-scale mechanical equipment access and installation and disassembly fees, concrete formwork and support fees, scaffolding fees, completed engineering and equipment protection fees, construction drainage and dewatering fees, etc.

[0031] The installation engineering cost growth information: the installation engineering cost also includes direct cost and measure cost, but the specific project is different, such as the labor cost in the installation engineering may involve the installation worker's salary and additional welfare cost, the material cost includes the cost of various materials, equipment and components used in the installation process.

[0032] The equipment and tool purchase cost growth information includes: equipment purchase cost: the cost of various equipment (such as mechanical equipment, electrical equipment, etc.) purchased for the construction project. Tool purchase cost: the cost of various tools and instruments (such as construction tools, detection instruments, etc.) purchased for the construction project.

[0033] Other cost growth information includes: construction unit management cost, various costs incurred by the construction unit for organizing and managing the construction of the project, survey and design cost, cost incurred for providing survey and design services for the construction project, supervision cost, cost incurred for providing supervision services for the construction project, other related costs: such as bidding agent fee, engineering cost consulting fee, engineering insurance fee and other costs related to engineering construction.

[0034] Establish a hierarchical model with layer-by-layer hierarchy, wherein the hierarchy includes target layer, criterion layer and measure layer; Construct a judgment matrix: determine the number of relevant influencing factors of the hierarchy , construct an importance weight influencing factor set , , wherein, is a subset of building engineering cost growth rate, is a subset of installation engineering cost growth rate, is a subset of equipment and tool purchase cost growth, is a subset of other cost growth rate, and two subsets in the same level are taken out from the set for comparison, the importance ratio is represented by , and the corresponding importance is valued according to the preset proportion, the importance of each layer is combined to form a judgment matrix; Calculate the maximum eigenvalue of the judgment matrix : ; , wherein, is a matrix obtained by normalizing each column vector of the judgment matrix, is 1, 2...b, is a matrix obtained by adding the elements of the matrix row by row and then normalizing the vector, is a matrix obtained by adding the elements of the matrix column by column and then normalizing the vector; Calculate the importance weight of the point data corresponding to each sub-time : ; wherein, represents the order of the judgment matrix.

[0035] In some embodiments, the process of calculating the trend compensation information of the point-data-related engineering budget sequence specifically comprises the following steps: obtaining a point-data-related engineering budget sequence ; making each corresponding node of each sub-time as a parent node in turn to construct j dependent classifiers; calculating the posterior probability of the engineering budget sample for each dependent sub-classifier : ; wherein, represents the engineering budget trend category and the maximum conditional probability of the point-data-related engineering budget sequence , the probability that the engineering budget trend category occurs in the point-data-related engineering budget sequence; after obtaining the posterior probability of each dependent sub-classifier, calculating the average of the posterior probabilities : ; j is the number of dependent classifiers; selecting the engineering budget trend category with the maximum average of the posterior probabilities as the trend compensation information, wherein the engineering budget trend category includes a long-term continuous growth trend category, a periodic fluctuation trend category, and a project-specific fluctuation trend category.

[0036] In some embodiments, the process of determining a trend compensation value based on the trend compensation information and correcting the prediction result based on the trend compensation value specifically comprises the following steps: querying the trend compensation value based on the trend compensation information, each engineering budget trend category is preset with a trend compensation value, and the prediction result is corrected based on the trend compensation value.

[0037] ​The above-described embodiments can be implemented in whole or in part by software, hardware, firmware, or any combination thereof. When implemented by software, the above-described embodiments can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, the processes or functions described in the embodiments of the present application are wholly or partially generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable devices. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another computer-readable storage medium, for example, the computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center through a wired (for example, infrared, wireless, microwave, etc.) manner. The computer-readable storage medium can be any available medium that can be accessed by a computer or a data storage device such as a server, data center, etc. containing one or more available medium collections. The available medium can be a magnetic medium (for example, a floppy disk, a hard disk, a magnetic tape), an optical medium (for example, a DVD), or a semiconductor medium. The semiconductor medium can be a solid-state disk.

[0038] Those skilled in the art can realize that the units and algorithm steps of the examples described in combination with the embodiments disclosed herein can be realized in electronic hardware or a combination of computer software and electronic hardware. Whether the functions are realized in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of the present application.

[0039] Those skilled in the art can clearly understand that, for the convenience and brevity of the description, the specific working processes of the above-described system, device and unit can refer to the corresponding processes in the foregoing method embodiments, which will not be described here.

[0040] In several embodiments provided in the present application, it should be understood that the disclosed system, device and method can be implemented in other ways. For example, the above-described device embodiments are only schematic, for example, the division of the units is only a logical function division, and actual implementation can have another division manner, for example, a plurality of units or components can be combined or integrated into another system, or some features can be ignored or not executed. In addition, the coupling or direct coupling or communication connection between the units shown or discussed can be indirect coupling or communication connection through some interface, device or unit, and can be electrical, mechanical or other forms.

[0041] The units described as separate components may or may not be physically separate, and the components displayed as units may or may not be physical units, that is, may be located in one place, or may also be distributed to multiple network units. Part or all of the units can be selected to achieve the purpose of the embodiment of the present application according to actual needs.

[0042] The above is merely a specific implementation of the present application, but the protection scope of the present application is not limited thereto, and any person skilled in the art can easily think of changes or replacements within the technical range disclosed by the present application, which should be covered within the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the protection scope of the claims.

Claims

1. A method for correcting an engineering budget deviation based on a machine learning model, characterized by, The method comprises the following steps: Obtaining engineering budget data of a plurality of historical periods, analyzing the engineering budget data of the plurality of historical periods, obtaining lag relative deviation, lag numerical deviation and lag trend deviation of the engineering budget prediction model, and calculating the deviation index of the engineering budget prediction model in the historical period based on the lag relative deviation, the lag numerical deviation and the lag trend deviation, wherein the engineering budget prediction model is an LSTM network; Based on the deviation index, it is judged whether the engineering budget prediction model meets the current period engineering budget prediction requirement, if yes, the current period engineering budget data is input into the engineering budget prediction model, and an initial engineering budget prediction result is output; Obtaining point data corresponding to each sub-time in the current period, calculating the importance weight of the point data corresponding to each sub-time, embedding the weighted point data into the initial engineering budget prediction result in turn according to the high and low values of the importance weight, and generating a point data related engineering budget sequence corresponding to the current period; Calculating the trend compensation information of the point data related engineering budget sequence, determining the trend compensation value based on the trend compensation information, and correcting the prediction result based on the trend compensation value. 2.The method of claim 1, wherein, The analysis of the engineering budget data of the plurality of historical periods obtains the lag relative deviation, the lag numerical deviation and the lag trend deviation of the engineering budget prediction model, which comprises the following processes: Calculate the lag relative deviation LRE: ； Calculate the lag numerical deviation LVE: ； Calculate the lag trend deviation LTE: ； wherein, represents a real project budget value of the i-th historical period, represents a project budget prediction value of the i-th historical period obtained by the project budget prediction model, represents a real project budget value input to the project budget prediction model of the i-th historical period, and N represents the number of historical periods. 3.The method of claim 1, wherein, The deviation index of the engineering budget prediction model in the historical period is calculated based on the lag relative deviation, the lag numerical deviation and the lag trend deviation, which comprises the following processes: Calculate the average value of the lag relative deviation, the lag numerical deviation and the lag trend deviation in each historical period; A rectangular coordinate system is constructed with the execution time of the historical period as the X-axis and the average value as the Y-axis, all average values are marked as points in the rectangular coordinate system, and adjacent points in the rectangular coordinate system are connected to generate a deviation multiple curve; A vertical line is drawn from both ends of the deviation multiple curve to the X-axis to obtain four starting and ending line segments, and a closed figure is formed by the deviation multiple curve, the two starting and ending line segments and the X-axis. The total area of the closed figure is calculated and recorded as the deviation index of the engineering budget prediction model in the historical period.

4. The method of claim 1, wherein the method is based on a machine learning model. Based on the deviation index, it is judged whether the engineering budget prediction model meets the current period engineering budget prediction requirement, if yes, the current period engineering budget data is input into the engineering budget prediction model, and an initial engineering budget prediction result is output; The current period engineering budget data is input into the engineering budget prediction model, and an initial engineering budget prediction result is output, which comprises the following processes:

5. The method of claim 1, wherein, The importance weight of the point data corresponding to each sub-time is calculated, which comprises the following processes: The point data includes building engineering cost growth information, installation engineering cost growth information, equipment and tool purchase cost growth information and other cost growth information; extracting a feature vector of the project budget data of the current period , inputting the feature vector into an LSTM network to obtain a last state of a forward LSTM network and a last state of a backward LSTM network : ； ； and denote the forward and backward state operations, respectively; To concatenate as a vector , :​ ; wherein, denotes a concatenation operation; will be described below. The input is fed into a fully connected network and the predicted value label for the current time period is output through a softmax function: ； a probability distribution representing the predicted value label for the current time period, , denotes a training parameter; The objective function is: ; wherein, denotes a learnable parameter, denotes the probability that the engineering budget of the kth prediction value label for the is the number of time periods, is the number of prediction value range intervals.

6. The method of claim 1, wherein, ​ ​ ​ A hierarchical model is established layer by layer, wherein the hierarchy includes a target layer, a criterion layer and a measure layer; Constructing judgment matrix: determining the number of relevant influencing factors of each level , constructing the set of importance weight influencing factors , , wherein, is the subset of building engineering cost growth rate, is the subset of installation engineering cost growth rate, is the subset of equipment and tool purchase cost growth, is the subset of other cost growth rate, and two subsets in the same level are taken out from the set for comparison, the ratio of importance is represented by , and the corresponding importance is valued according to the preset proportion, and the importance of each level is combined to form a judgment matrix; Computing the largest eigenvalue of a judgment matrix : ； wherein, is a matrix obtained by normalizing each column vector of the judgment matrix, has values of 1, 2,... b, is a matrix whose elements are added row by row to obtain a vector, and then normalized, is a matrix whose elements are added column by column to obtain a vector, and then normalized. calculating an importance weight of the point data corresponding to each sub-time point : ； wherein denotes the order of the judgment matrix.

7. The method of claim 1, wherein the method is based on a machine learning model. The trend compensation information of the point data related engineering budget sequence includes the following processes: Acquisition point data related engineering budget sequence ; Each sub-time corresponding node is taken as a parent node in turn to construct j dependent classifiers; For each dependent sub-classifier, the posterior probability of the engineering budget sample is calculated : ; wherein represents an engineering budget trend category associated with the point data maximum conditional probability of represents a probability of occurrence of an engineering budget trend category in the point data associated engineering budget sequence ​ After obtaining each dependent sub-classifier posterior probability, calculate the posterior probability average value : ; j is the number of classifiers; selective posterior probability average The maximum engineering budget trend category is taken as the trend compensation information, wherein the engineering budget trend category includes a long-term continuous growth trend category, a periodic fluctuation trend category, and a project-specific fluctuation trend category.

8. The method of claim 1, wherein the method is based on a machine learning model. The trend compensation value is determined based on the trend compensation information, and the prediction result is corrected based on the trend compensation value, which includes the following processes: The trend compensation value is queried based on the trend compensation information, and each engineering budget trend category is preset with a trend compensation value, and the prediction result is corrected based on the trend compensation value.

Images