A deep learning-based building engineering cost analysis method and system

Abstract

The application provides a deep learning-based building engineering cost analysis method and system, and relates to the technical field of machine learning. The application acquires historical engineering change cost data sets by collecting original size parameters of original building components and changed size parameters of changed building components after the original building components are changed. The application obtains comparison similarity based on the original size parameters and the changed size parameters. The application obtains cost adjustment rules through deep learning. The application obtains feature quantization vectors based on size parameter differences and cost adjustment amplitudes in target historical change cases. The application obtains cost adjustment coefficients based on the cost adjustment rules. The application extracts historical benchmark costs corresponding to the original building components from the target historical change cases to calculate building engineering changed costs, thereby realizing accurate, efficient and automatic calculation of building engineering changed costs.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of machine learning, in particular to a building engineering cost analysis method and system based on deep learning. BACKGROUND

[0002] In modern building engineering management, engineering changes caused by design optimization, construction condition changes or owner demand adjustment are common phenomena. Such changes will lead to changes in building component size, material consumption or construction process, and thus have a direct impact on the overall project cost. Therefore, there is an urgent need for a technical system that can integrate data collection, efficient information transmission and intelligent analysis and decision-making.

[0003] Currently, the existing mainstream scheme attempts to cope with the cost accounting challenges brought by engineering changes by integrating building information models and cloud big data analysis platforms. This scheme uses the component geometry and attribute information stored in the integrated building information model to automatically extract the data differences of the components before and after the change when a design change occurs, and uploads the change information to the cloud server through the 5G network. Based on the pre-set rule engine and statistical regression model, the server side combines similar change records in the historical project database to infer the cost change trend, and finally generates a change cost estimation report. However, the existing scheme has certain defects, for example, it relies on a fixed rule library and a linear statistical model, making it difficult to fully capture the complex rules of multi-factor coupling and non-linear correlation in engineering changes, and the model generalization ability is insufficient, resulting in a large deviation in the estimation result. Since all data processing is centralized in the cloud, the on-site measurement data needs to go through a long queuing analysis period after uploading, which cannot meet the feedback needs in high-time scenarios, and the matching of historical cases only relies on parameter threshold comparison, lacking the ability to identify the deep similarity of change patterns. SUMMARY

[0004] The present application aims to provide a building engineering cost analysis method and system based on deep learning to solve the problems of insufficient accuracy of building engineering cost estimation results in the prior art, inability to meet the feedback needs in high-time scenarios, and lack of ability to identify the deep similarity of change patterns.

[0005] To solve the above technical problems, in a first aspect, the present application provides a building engineering cost analysis method based on deep learning, comprising:

[0006] Collecting original size parameters of original building components and changed size parameters of changed building components after change, obtaining a historical engineering change cost dataset, the historical engineering change cost dataset including multiple change types, each change type corresponding to multiple historical change cases;

[0007] obtain a comparison similarity degree based on the original size parameter and the changed size parameter, and select a target historical change case matching the comparison similarity degree from the historical engineering change cost data set;

[0008] perform feature analysis on the historical original building component feature set, the historical changed building component feature set, and the historical cost data in the target historical change case through deep learning to obtain a cost adjustment rule;

[0009] perform feature quantization on a size difference between the original size parameter and the changed size parameter based on the size parameter difference and the cost adjustment range in the target historical change case to obtain a feature quantization vector;

[0010] perform matching calculation on the feature quantization vector based on the cost adjustment rule to obtain a cost adjustment coefficient, extract a historical benchmark cost corresponding to the original building component from the target historical change case, and calculate a building engineering changed cost by combining the cost adjustment coefficient.

[0011] Optionally, obtaining a comparison similarity degree based on the original size parameter and the changed size parameter, and selecting a target historical change case matching the comparison similarity degree from the historical engineering change cost data set, comprises:

[0012] calculate a length difference, a width difference, and a height difference based on the original size parameter and the changed size parameter, and combine the original size parameter, the changed size parameter, the length difference, the width difference, and the height difference to form a change feature vector;

[0013] compare the change feature vector with historical change feature vectors of historical change cases in the historical engineering change cost data set to obtain a plurality of comparison similarity degrees, and select a historical change case with a comparison similarity degree greater than a preset threshold as a target historical change case.

[0014] Optionally, performing feature analysis on the historical original building component feature set, the historical changed building component feature set, and the historical cost data in the target historical change case through deep learning to obtain a cost adjustment rule, comprises:

[0015] extract a historical original size parameter and a historical original material parameter from the historical original building component feature set of the target historical change case, extract a historical changed size parameter and a historical changed material parameter from the historical changed building component feature set, and extract a historical cost adjustment value from the historical cost data;

[0016] Calculate a size difference value of a historical original size parameter and a corresponding historical changed size parameter in a same target historical change case, a material difference degree of a historical original material parameter and a corresponding historical changed material parameter, and combine the size difference value, the material difference degree, and the historical cost adjustment value to form a plurality of feature cost sample pairs;

[0017] Perform associated feature analysis on the feature cost sample pairs through deep learning to obtain a plurality of intermediate feature vectors;

[0018] Calculate a similarity of any two intermediate feature vectors, combine the intermediate feature vectors with a similarity exceeding a preset vector similarity threshold to form a plurality of combined feature clusters, take the intermediate feature vectors corresponding to a similarity not exceeding the preset vector similarity threshold as separate feature clusters, and integrate all the combined feature clusters and all the separate feature clusters to obtain a plurality of feature clusters;

[0019] Perform regularity analysis on each feature cluster to form a plurality of preliminary cost adjustment rules, and integrate all the preliminary cost adjustment rules to obtain a cost adjustment rule.

[0020] Optionally, the associated feature analysis on the feature cost sample pairs through deep learning to obtain a plurality of intermediate feature vectors includes:

[0021] Perform vector conversion on the size difference value, the material difference degree, and the historical cost adjustment value corresponding to each feature cost sample pair to obtain a plurality of basic feature vectors;

[0022] Perform numerical distribution feature analysis on each basic feature vector through deep learning to obtain a numerical range feature, a fluctuation feature, and a change amplitude feature, combine the numerical range feature, the fluctuation feature, and the change amplitude feature to form a primary associated feature of each basic feature vector;

[0023] Perform interactive feature analysis on each basic feature vector to obtain a parameter co-occurrence frequency feature, a numerical change coordination feature, and a proportion associated feature, combine the parameter co-occurrence frequency feature, the numerical change coordination feature, and the proportion associated feature to form an intermediate associated feature of each basic feature vector;

[0024] Perform comprehensive feature analysis on each basic feature vector to obtain a parameter combination mode feature and a numerical linkage regularity feature, and combine the parameter combination mode feature and the numerical linkage regularity feature to form a high-level associated feature of each basic feature vector;

[0025] Integrate the primary associated feature, the intermediate associated feature, and the high-level associated feature corresponding to each basic feature vector to form a plurality of intermediate feature vectors.

[0026] Optionally, each feature cluster is regularly analyzed to form a plurality of preliminary cost adjustment rules, all preliminary cost adjustment rules are integrated to obtain the cost adjustment rule, including:

[0027] According to the intermediate feature vector corresponding to each feature cluster, the first size difference interval, the material difference range and the historical cost adjustment interval of the target historical change case associated with each feature cluster are determined;

[0028] The first size difference interval and the historical cost adjustment interval of each feature cluster are analyzed to obtain a first corresponding relationship, and the material difference range and the historical cost adjustment interval of each feature cluster are analyzed to obtain a second corresponding relationship;

[0029] Based on the first corresponding relationship and the second corresponding relationship, the combination rule of size difference, material difference degree and historical cost adjustment value in each feature cluster is determined to form a preliminary cost adjustment rule of each feature cluster;

[0030] The preliminary cost adjustment rules corresponding to the first size difference interval existing in the overlapping numerical range and the material difference range existing in the overlapping numerical range are integrated to obtain a plurality of intermediate cost adjustment rules;

[0031] If different intermediate cost adjustment rules exist rule conflict, the intermediate cost adjustment rule with more feature cluster quantity is reserved as a candidate cost adjustment rule, if the intermediate cost adjustment rule does not exist rule conflict, the intermediate cost adjustment rule is taken as a candidate cost adjustment rule, all candidate cost adjustment rules are integrated to obtain the cost adjustment rule.

[0032] Optionally, based on the size parameter difference and the cost adjustment range in the target historical change case, the size difference of the original size parameter and the changed size parameter is quantified to obtain a feature quantization vector, including:

[0033] The size parameter difference between the historical original size parameter and the historical changed size parameter in the target historical change case is analyzed to determine the interval division rule, and the cost adjustment range in the target historical change case is extracted to obtain the cost adjustment range feature;

[0034] According to the interval division rule, the size parameter difference is divided into a plurality of second size difference intervals, so as to associate the second size difference interval and the cost adjustment range feature of the corresponding target historical change case to form a difference feature association table;

[0035] The size difference of the original size parameter and the changed size parameter is calculated, and the target size difference interval of the size difference is determined according to the difference feature association table.

[0036] quantify the cost adjustment amplitude feature quantity associated with the target size difference interval in the difference feature association table as a cost adjustment amplitude feature quantization value, to combine the size difference and the cost adjustment amplitude feature quantization value to obtain a feature quantization vector.

[0037] Optionally, based on the cost adjustment rule, a matching calculation is performed on the feature quantization vector to obtain a cost adjustment coefficient, a historical reference cost corresponding to the original building component is extracted from the target historical change case, and a building engineering change cost is calculated in combination with the cost adjustment coefficient, including:

[0038] The matching degrees of each candidate cost adjustment rule in the cost adjustment rule and the feature quantization vector are calculated respectively, and the candidate cost adjustment rule corresponding to the maximum matching degree is taken as a target cost adjustment rule;

[0039] An adjustment parameter corresponding to a third size difference interval and a material difference range matched with the feature quantization vector is extracted from the target cost adjustment rule to calculate a cost adjustment coefficient;

[0040] A matching historical change case with the same component type and specification parameter as the original building component is extracted from the target historical change case, an initial cost of the matching historical change case is taken as a historical reference cost, and a building engineering change cost is calculated in combination with the cost adjustment coefficient.

[0041] In a second aspect, the present application provides a building engineering cost analysis system based on deep learning, including:

[0042] The acquisition module is configured to acquire original size parameters of an original building component and changed size parameters of a changed building component after change, and obtain a historical engineering change cost dataset, the historical engineering change cost dataset including a plurality of change types, each change type corresponding to a plurality of historical change cases;

[0043] The selection module is configured to obtain a comparison similarity based on the original size parameters and the changed size parameters, and select a target historical change case matched with the comparison similarity from the historical engineering change cost dataset;

[0044] The analysis module is configured to perform feature analysis on a historical original building component feature set, a historical changed building component feature set, and historical cost data in the target historical change case through deep learning to obtain a cost adjustment rule;

[0045] a quantification module configured to quantize a size difference between the original size parameter and the changed size parameter based on a size parameter difference and a cost adjustment range in the target historical change case, to obtain a feature quantization vector;

[0046] a calculation module configured to perform matching calculation on the feature quantization vector based on the cost adjustment rule, to obtain a cost adjustment coefficient, extract a historical reference cost corresponding to the original building component from the target historical change case, and calculate a changed cost of the building project based on the cost adjustment coefficient.

[0047] In a third aspect, the present application provides an electronic device, comprising:

[0048] a memory configured to store a computer program;

[0049] a processor configured to execute the computer program to implement the steps of the building project cost analysis method based on deep learning according to the first aspect.

[0050] In a fourth aspect, the present application provides a computer readable storage medium, wherein the computer readable storage medium stores a computer program, and the computer program is executable by a processor to implement the steps of the building project cost analysis method based on deep learning according to the first aspect.

[0051] The building engineering cost analysis method based on deep learning provided in the application collects original size parameters of original building components and changed size parameters of changed building components after change, obtains a historical engineering change cost dataset, the historical engineering change cost dataset includes multiple change types, and each change type corresponds to multiple historical change cases; based on the original size parameters and the changed size parameters, a comparison similarity is obtained, and a target historical change case matching the comparison similarity is selected from the historical engineering change cost dataset; through deep learning, feature analysis is performed on a historical original building component feature set, a historical changed building component feature set and historical cost data in the target historical change case, and a cost adjustment rule is obtained; based on a size parameter difference value and a cost adjustment range in the target historical change case, a size difference value of the original size parameters and the changed size parameters is quantified, and a feature quantization vector is obtained; based on the cost adjustment rule, matching calculation is performed on the feature quantization vector, a cost adjustment coefficient is obtained, a historical benchmark cost corresponding to the original building component is extracted from the target historical change case, and the original building component is combined with the cost adjustment coefficient to calculate a building engineering changed cost. By collecting the original size parameters and the changed size parameters of the changed building components, obtaining the historical engineering change cost dataset, the key physical parameters in the building engineering change process and the corresponding cost response data are systematically collected; the reference cases with comparability are accurately selected from the massive historical data, the mechanicalness and one-sidedness of the traditional threshold matching method are avoided, the intelligence and relevance of the case matching are improved, the generalization ability and decision-making scientificity of the cost analysis model are enhanced, the differences of different change types and ranges can be quantitatively analyzed and matched in a unified space, the intelligent migration and dynamic adaptation from the historical law to the current scene are realized, the dependence on cloud centralized calculation is avoided, the high-precision cost recalculation is quickly completed on the edge side or locally, and the demand of the engineering site for the change cost response is met; further, the difference between the original and changed size parameters in the same case and the difference degree of the material parameters are calculated to construct multiple feature cost sample pairs; the deep learning model is used to mine the correlation features of the sample pairs, and multiple intermediate feature vectors are generated; the sample mode in each feature cluster is inducted to generate multiple preliminary cost adjustment rules, and all the preliminary rules are integrated into a unified cost adjustment rule through a fusion optimization strategy; the accuracy and intelligence level of the cost deduction are improved, and the efficient and accurate local analysis of the engineering change cost is realized without relying on cloud centralized processing. BRIEF DESCRIPTION OF DRAWINGS

[0052] In order to more clearly illustrate the technical solutions of the embodiments of the present application or the prior art, the accompanying drawings required by the embodiments or the prior art description will be briefly introduced below. Obviously, the described embodiments are only some of the embodiments of the present application, not all. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative labor are within the scope of protection of the present application.

[0053] Figure 1 A flowchart of a building engineering cost analysis method based on deep learning provided by an embodiment of the present application is shown in the figure.

[0054] Figure 2 A specific implementation schematic diagram of a building engineering cost analysis method based on deep learning provided by an embodiment of the present application is shown in the figure.

[0055] Figure 3 A structural schematic diagram of a building engineering cost analysis system based on deep learning provided by an embodiment of the present application is shown in the figure. DETAILED DESCRIPTION

[0056] For the scenario that after the size of building components, the amount of materials, etc. are changed due to design optimization, construction condition change or owner demand adjustment, the cost needs to be quickly and accurately calculated, the existing scheme relies on building information model and cloud big data platform, combines with preset rule engine and linear statistical model to perform cost deduction, but has defects such as difficulty in capturing nonlinear correlation law under multi-factor coupling, insufficient model generalization ability, response delay caused by cloud centralized processing, and lack of deep similarity recognition ability by only comparing parameter threshold value with historical change cases; the present application filters out matched target historical change cases from historical engineering change cost data set by constructing a comprehensive data base, improves the intelligence and relevance of case selection; through deep learning, the historical original building component feature set, the historical changed building component feature set and the historical cost data in the target historical change case are analyzed, the complex nonlinear mapping relationship is mined, the cost adjustment rule with adaptive ability is formed, and the efficient and accurate calculation of the cost after the building engineering change is completed, so as to realize the intelligent and scientific analysis of the engineering change cost.

[0057] In order to enable the personnel in the technical field to better understand the present application scheme, the present application will be further described in detail below in combination with the accompanying drawings and specific embodiments. Obviously, the described embodiments are only some of the embodiments of the present application, not all. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative labor are within the scope of protection of the present application.

[0058] The core of the present application is to provide a building engineering cost analysis method based on deep learning, and a flowchart of one specific implementation method is shown in the figure.Figure 1 The method comprises the following steps:

[0059] Step 101: Collecting original size parameters of an original building component and changed size parameters of a changed building component formed by changing the original building component, and obtaining a historical engineering change cost dataset, wherein the historical engineering change cost dataset comprises a plurality of change types, and each change type corresponds to a plurality of historical change cases.

[0060] In this step, the original building component refers to a building component that has not been changed. The original size parameters refer to parameters such as length, width, and height of the original building component, which describe the size of the original building component. The changed building component refers to a building component formed by changing the original building component. The changed size parameters refer to parameters such as length, width, and height of the changed building component, which describe the size of the changed building component. The historical engineering change cost dataset refers to a set containing a plurality of past engineering change cases and corresponding cost information. The change type refers to the category of building component change, such as size adjustment, material replacement, etc. The historical change case refers to a specific engineering change instance recorded in the historical engineering change cost dataset.

[0061] In the embodiments of the present application, the original size parameters of the original building component are collected, and the changed size parameters of the changed building component formed by changing the original building component are also collected. By collecting the change records and cost data of past projects, a historical engineering change cost dataset is formed. The dataset is divided into a plurality of change types according to different categories of changes, and each change type contains a plurality of corresponding historical change cases.

[0062] Step 102: Based on the original size parameters and the changed size parameters, a comparison similarity is obtained, and a target historical change case matching the comparison similarity is selected from the historical engineering change cost dataset.

[0063] In this step, the comparison similarity refers to the degree of similarity between the change feature vector and the historical change feature vector. The target historical change case refers to the historical change case selected from the historical engineering change cost dataset and matching the comparison similarity.

[0064] Step 103: Through deep learning, the historical original building component feature set, the historical changed building component feature set, and the historical cost data in the target historical change case are analyzed to obtain a cost adjustment rule.

[0065] In this step, deep learning refers to a method of feature learning on data through a multi-layer neural network. The historical original building component feature set refers to the feature set of the building component before the change in the target historical change case. The historical changed building component feature set refers to the feature set of the building component after the change in the target historical change case. The historical cost data refers to the cost record related to the change in the target historical change case. The cost adjustment rule refers to a rule obtained based on feature analysis for guiding cost adjustment.

[0066] Step 104: Based on the size parameter difference value and the cost adjustment amplitude in the target historical change case, the size difference value of the original size parameter and the changed size parameter is quantified, and a feature quantization vector is obtained.

[0067] In this step, the size parameter difference value refers to the difference between the historical original size parameter and the historical changed size parameter. The cost adjustment amplitude refers to the cost adjustment amount caused by the change of the building component. The size difference value refers to the difference between the original size parameter and the changed size parameter. The feature quantization vector refers to a vector formed after quantizing the size difference value and the corresponding cost adjustment amplitude feature.

[0068] Step 105: Based on the cost adjustment rule, the feature quantization vector is matched and calculated to obtain a cost adjustment coefficient. The historical reference cost corresponding to the original building component is extracted from the target historical change case, and the cost adjustment coefficient is combined to calculate the building engineering change cost.

[0069] In this step, the cost adjustment coefficient refers to a coefficient for adjusting the historical reference cost. The historical reference cost refers to the initial cost corresponding to the original building component in the target historical change case. The building engineering change cost refers to the building engineering cost after the change obtained by adjustment calculation.

[0070] The embodiment of the present application realizes accurate cost analysis based on historical data and deep learning by collecting parameters before and after the change of the building component, combining historical data to select a target case, using deep learning to analyze features to form a rule, quantizing features and matching calculation, and finally obtaining the change cost.

[0071] The present application provides a specific embodiment, step 102, based on the original size parameter and the changed size parameter, the comparison similarity is obtained, and the target historical change case matching the comparison similarity is selected from the historical engineering change cost data set, specifically including the following steps:

[0072] Step 201: based on the original size parameter and the changed size parameter, the length difference value, the width difference value and the height difference value are calculated, and the original size parameter, the changed size parameter, the length difference value, the width difference value and the height difference value are combined to form a change feature vector.

[0073] In this step, the change feature vector refers to the vector formed by combining the original size parameter, the changed size parameter, the length difference value, the width difference value and the height difference value.

[0074] In the embodiments of the present application, the original length parameter, the original width parameter and the original height parameter are extracted from the original size parameter, the changed length parameter, the changed width parameter and the changed height parameter are extracted from the changed size parameter, the length difference value = original length parameter-changed length parameter is calculated, the width difference value = original width parameter-changed width parameter is calculated, and the height difference value = original height parameter-changed height parameter is calculated. The original size parameter, the changed size parameter, the length difference value, the width difference value and the height difference value are arranged in a preset order to form a change feature vector.

[0075] Step 202: comparing the change feature vector with the historical change feature vectors of each historical change case in the historical engineering change cost data set, obtaining a plurality of comparison similarities, and taking the historical change case with a comparison similarity greater than a preset threshold as a target historical change case.

[0076] In this step, the historical change feature vector refers to the vector formed by combining the corresponding original size parameter, the changed size parameter, the length difference value, the width difference value and the height difference value in the historical change case. The preset threshold refers to a value preset for judging whether the comparison similarity meets the standard.

[0077] In the embodiments of the present application, the historical change feature vectors of each historical change case are retrieved from the historical engineering change cost data set, the change feature vector is compared with each historical change feature vector, the similarity of the two is calculated to obtain a plurality of comparison similarities, each comparison similarity is compared with the preset threshold, and the historical change case with a comparison similarity greater than the preset threshold is selected as the target historical change case.

[0078] The embodiments of the present application realize the accurate selection of the target historical change case by calculating the size difference value and combining to form the feature vector, and comparing and screening with the feature vectors of the historical cases, which provides a reliable case basis for subsequent cost analysis.

[0079] For example, the original size parameters of the A component are length 5 m, width 2 m, and height 3 m, and the changed size parameters are length 6 m, width 2 m, and height 3 m. The length difference is calculated as 5-6=-1 m, the width difference is calculated as 2-2=0 m, and the height difference is calculated as 3-3=0 m. The original size parameters, the changed size parameters, the length difference, the width difference, and the height difference are combined to form a changed feature vector ; the historical changed feature vector of a historical changed case is called from the historical engineering changed cost data set . The cosine similarity is used to calculate the comparison similarity, and the calculation formula is: , wherein, is the vector dot product, which is calculated as 5x4+2x2+3x3+6x5+2x2+3x3+(-1)x(-1)+0x0+0x0=20+4+9+30+4+9+1+0+0=77; |X| is the length of vector X, which is calculated as sqrt(5^2+2^2+3^2+6^2+2^2+3^2+(-1)^2+0^2+0^2)=sqrt(25+4+9+36+4+9+1)=sqrt(88)≈9.38; |Y| is the length of vector Y, which is calculated as sqrt(4^2+2^2+3^2+5^2+2^2+3^2+(-1)^2+0^2+0^2)=sqrt(16+4+9+25+4+9+1)=sqrt(68)≈8.24; and the comparison similarity is 77 / (9.38x8.24)≈77 / 77.3≈0.9. The preset threshold is 0.8. Since 0.9>0.8, the historical case is taken as the target historical changed case.

[0080] The present application provides a specific embodiment, as shown in Figure 2 step 103, the historical original building component feature set, the historical changed building component feature set, and the historical cost data in the target historical changed case are analyzed by deep learning to obtain a cost adjustment rule, which specifically includes the following steps:

[0081] Step 301: The historical original size parameters and the historical original material parameters are extracted from the historical original building component feature set of the target historical changed case, the historical changed size parameters and the historical changed material parameters are extracted from the historical changed building component feature set, and the historical cost adjustment value is extracted from the historical cost data.

[0082] In this step, the historical original size parameter refers to a size parameter of a historical original building component in a target historical change case. The historical original material parameter refers to a material parameter of the historical original building component in the target historical change case. The historical changed size parameter refers to a size parameter of a historical changed building component in the target historical change case. The historical changed material parameter refers to a material parameter of the historical changed building component in the target historical change case. The historical cost adjustment value refers to a cost adjustment value generated due to a change of the building component in the target historical change case.

[0083] In the embodiments of the present application, the historical original size parameter and the historical original material parameter are selected and extracted from the historical original building component feature set of the target historical change case, the historical changed size parameter and the historical changed material parameter are selected and extracted from the historical changed building component feature set, and the historical cost adjustment value is selected and extracted from the historical cost data, so as to ensure that the extracted parameters correspond to the same target historical change case.

[0084] Step 302: calculating a size difference value of the historical original size parameter and the corresponding historical changed size parameter in the same target historical change case, a material difference degree of the historical original material parameter and the corresponding historical changed material parameter, and combining the size difference value, the material difference degree and the historical cost adjustment value to form a plurality of feature cost sample pairs.

[0085] In this step, the material difference degree refers to a difference degree between the historical original material parameter and the corresponding historical changed material parameter. The feature cost sample pair refers to paired data formed by combining the size difference value, the material difference degree and the historical cost adjustment value.

[0086] In the embodiments of the present application, for the historical original size parameter and the corresponding historical changed size parameter in the same target historical change case, the size difference value = historical original size parameter - historical changed size parameter is calculated; according to the historical original material parameter and the corresponding historical changed material parameter, the material difference degree = (number of difference items of the historical original material parameter and the historical changed material parameter / total number of items of the historical original material parameter) x 100% is calculated; and the size difference value, the material difference degree and the historical cost adjustment value of the same case are combined in order to form a plurality of feature cost sample pairs.

[0087] Step 303: performing associated feature analysis on the feature cost sample pairs by deep learning to obtain a plurality of intermediate feature vectors.

[0088] In this step, the intermediate feature vector refers to a vector obtained after performing associated feature analysis on the feature cost sample pairs.

[0089] In the embodiments of the present application, the size difference, the material difference degree and the historical cost adjustment value in each feature cost sample pair are analyzed by deep learning, the internal correlation and feature performance among the three are mined, the analysis results are presented in the form of vectors, and a plurality of intermediate feature vectors are obtained.

[0090] Step 304: Similarity of any two intermediate feature vectors is calculated, intermediate feature vectors with similarity exceeding a preset vector similarity threshold are combined to form a plurality of combined feature clusters, intermediate feature vectors corresponding to similarity not exceeding the preset vector similarity threshold are taken as separate feature clusters, all combined feature clusters and all separate feature clusters are integrated to obtain a plurality of feature clusters.

[0091] In this step, the similarity refers to the degree of similarity between two intermediate feature vectors. The preset vector similarity threshold refers to a similarity value preset for judging whether the intermediate feature vectors can be combined. The combined feature cluster refers to a set formed by combining a plurality of intermediate feature vectors with similarity exceeding the preset vector similarity threshold. The separate feature cluster refers to a set formed by an intermediate feature vector alone without similarity exceeding the preset vector similarity threshold with other intermediate feature vectors. The feature cluster refers to a set formed by integrating all combined feature clusters and all separate feature clusters.

[0092] In the embodiments of the present application, the similarity of any two intermediate feature vectors is calculated, the similarity is compared with the preset vector similarity threshold, the intermediate feature vectors with similarity exceeding the threshold are grouped together to form a plurality of combined feature clusters, the intermediate feature vectors not exceeding the threshold are each taken as a group to form a plurality of separate feature clusters, and all combined feature clusters and separate feature clusters are integrated together to obtain a plurality of feature clusters.

[0093] Step 305: Each feature cluster is regularly analyzed to form a plurality of preliminary cost adjustment rules, and all preliminary cost adjustment rules are integrated to obtain a cost adjustment rule.

[0094] In this step, the preliminary cost adjustment rule refers to the initial cost adjustment rule formed after regular analysis of the feature cluster.

[0095] In the embodiments of the present application, the intermediate feature vectors in each feature cluster are regularly analyzed, the correlation rules between the size difference, the material difference degree and the historical cost adjustment value in the cluster are refined, and these rules are converted into preliminary cost adjustment rules. All preliminary cost adjustment rules are sorted and integrated to remove repeated or conflicting contents, and a cost adjustment rule is obtained.

[0096] The embodiment of the application realizes accurate generation of cost adjustment rules based on historical data by extracting key parameters from target historical change cases, calculating difference degrees and forming sample pairs, obtaining feature vectors through deep learning analysis, and refining rules after clustering to form feature clusters, thereby providing a scientific basis for cost analysis.

[0097] The application provides an embodiment, step 303, which comprises the following steps:

[0098] Step 311: Vector conversion is performed on the size difference, material difference degree and historical cost adjustment value corresponding to each feature cost sample pair to obtain a plurality of basic feature vectors.

[0099] In this step, the basic feature vector refers to a vector obtained by vector conversion of the size difference, material difference degree and historical cost adjustment value in the feature cost sample pair.

[0100] In the embodiment of the application, for each feature cost sample pair, the size difference, material difference degree and historical cost adjustment value are converted into a numerical sequence in a preset order to form a vector form, thereby obtaining a plurality of basic feature vectors, each of which corresponds to a feature cost sample pair.

[0101] Step 312: Numerical distribution feature analysis is performed on each basic feature vector through deep learning to obtain a numerical range feature, a fluctuation feature and a change amplitude feature, and the numerical range feature, the fluctuation feature and the change amplitude feature are combined to form a primary association feature of each basic feature vector.

[0102] In this step, the numerical range feature refers to the numerical value range feature of each parameter in the basic feature vector. The fluctuation feature refers to the numerical fluctuation situation feature of each parameter in the basic feature vector. The change amplitude feature refers to the numerical change amplitude feature of each parameter in the basic feature vector. The primary association feature refers to a feature formed by combination of the numerical range feature, the fluctuation feature and the change amplitude feature.

[0103] In the embodiment of the application, numerical distribution feature analysis is performed on each basic feature vector through deep learning, and according to the size difference, a parameter value set is determined, thereby obtaining a numerical range feature = [min (parameter value set), max (parameter value set)], wherein min (parameter value set) is the minimum value in the parameter value set, and max (parameter value set) is the maximum value in the parameter value set. The fluctuation feature = √[Σ(x i -μ)² / n] is calculated, wherein x i is the i-th value in the parameter value set, and μ is the mean value of the parameter value set (μ = Σx in is the number of values in the parameter value set; the variation range feature is calculated as max (parameter value set) - min (parameter value set); similarly, the value range feature, the fluctuation feature, and the variation range feature of the material difference degree and the historical cost adjustment value are calculated according to the above formula, and the three types of features corresponding to the size difference value, the material difference degree, and the historical cost adjustment value of the same basic feature vector are combined in order to form the primary associated features of the basic feature vector.

[0104] Step 313: Inter-feature analysis is performed on each basic feature vector to obtain parameter co-occurrence frequency features, value change coordination features, and proportion association features, and the parameter co-occurrence frequency features, the value change coordination features, and the proportion association features are combined to form the intermediate associated features of each basic feature vector.

[0105] In this step, the parameter co-occurrence frequency feature refers to the frequency feature of the simultaneous occurrence of different parameters in the basic feature vector. The value change coordination feature refers to the coordination feature of the value change of different parameters in the basic feature vector. The proportion association feature refers to the proportion relationship feature between different parameter values in the basic feature vector. The intermediate associated feature refers to the feature formed by the combination of the parameter co-occurrence frequency feature, the value change coordination feature, and the proportion association feature.

[0106] In the embodiments of the present application, the inter-feature analysis is performed on each basic feature vector, the number of times that the size difference value and the material difference degree are in a specific range at the same time is counted to obtain the parameter co-occurrence frequency feature, the consistency of the size difference value change direction and the material difference degree change direction is analyzed to obtain the value change coordination feature, and the ratio of the size difference value to the historical cost adjustment value and the ratio of the material difference degree to the historical cost adjustment value are calculated to obtain the proportion association feature, and the three types of features of the same basic feature vector are combined in order to form the intermediate associated features of the basic feature vector.

[0107] Step 314: Comprehensive feature analysis is performed on each basic feature vector to obtain parameter combination mode features and value linkage rule features, and the parameter combination mode features and the value linkage rule features are combined to form the senior associated features of each basic feature vector.

[0108] In this step, the parameter combination mode feature refers to the combination mode feature of different parameters in the basic feature vector. The value linkage rule feature refers to the rule feature of the linkage change of different parameter values in the basic feature vector. The senior associated feature refers to the feature formed by the combination of the parameter combination mode feature and the value linkage rule feature.

[0109] In the embodiments of the present application, the comprehensive feature analysis is performed on each basic feature vector, common numerical combination forms of size difference value and material difference degree are summarized to obtain parameter combination mode features, the change of size difference value and material difference degree how to jointly affect the change rule of historical cost adjustment value is analyzed to obtain numerical linkage rule features, and the two types of features of the same basic feature vector are combined in sequence to form senior associated features of the basic feature vector.

[0110] Step 315: integrating the primary associated features, the middle associated features and the senior associated features corresponding to each basic feature vector to form a plurality of intermediate feature vectors.

[0111] In this step, in the embodiments of the present application, the primary associated features, the middle associated features and the senior associated features corresponding to each basic feature vector are spliced and integrated in a preset dimension order to form vectors containing multi-level associated information, and a plurality of intermediate feature vectors are obtained, each intermediate feature vector corresponding to a basic feature vector.

[0112] The embodiments of the present application gradually and deeply mine the association rules between parameters through multi-level associated feature analysis on the feature cost sample pairs, from the basic vector to the associated features at each level to the intermediate feature vectors, enrich the expression dimension of the features, and provide more comprehensive feature support for the formation of subsequent feature clusters and rule extraction.

[0113] The embodiments of the present application gradually and deeply mine the association rules between parameters through multi-level associated feature analysis on the feature cost sample pairs, from the basic vector to the associated features at each level to the intermediate feature vectors, enrich the expression dimension of the features, and provide more comprehensive feature support for the formation of subsequent feature clusters and rule extraction.

[0114] For example, the basic feature vector is [-1, 0, 1], corresponding to the size difference value -1m, the material difference degree 0%, and the historical cost adjustment value 10,000 yuan, and the size difference value parameter value set of the basic feature vector is {-1, -2, 0}, the material difference degree parameter value set is {0%, 0%, 1%}, and the historical cost adjustment value parameter value set is {10,000 yuan, 20,000 yuan, 5,000 yuan}. The numerical range feature of the size difference value is calculated as [min{-1, -2, 0}, max{-1, -2, 0}]=[-2, 0]; the fluctuation feature is √[((-1+1)²+(-2+1)²+(0+1)²) / 3]=√[(0+1+1) / 3]=√(2 / 3)≈0.82; the change amplitude feature is 0-(-2)=2. The numerical range feature of the material difference degree is [0%, 1%], the fluctuation feature is √[((0-0.33%)²+(0-0.33%)²+(1-0.33%)²) / 3]≈0.58%, and the change amplitude feature is 1%-0%=1%. The numerical range feature of the historical cost adjustment value is [5,000 yuan, 20,000 yuan], the fluctuation feature is √[((1-1.17)²+(2-1.17)²+(0.5-1.17)²) / 3]≈7,600 yuan, and the change amplitude feature is 2-0.5=1.5,000 yuan. These features are combined in order to obtain the primary association feature [[-2, 0], 0.82, 2, [0%, 1%], 0.58%, 1%, [5,000 yuan, 20,000 yuan], 7,600 yuan, 1.5,000 yuan]].

[0115] The present application provides an embodiment, step 305, regular analysis of each feature cluster to form a plurality of preliminary cost adjustment rules, and integrating all preliminary cost adjustment rules to obtain the cost adjustment rule, specifically including the following steps:

[0116] Step 321: According to the intermediate feature vector corresponding to each feature cluster, the first size difference interval, the material difference range, and the historical cost adjustment interval of the target historical change case associated with each feature cluster are determined.

[0117] In this step, the first size difference interval refers to the value range of the size difference value in the target historical change case associated with the feature cluster. The material difference range refers to the value range of the material difference degree in the target historical change case associated with the feature cluster. The historical cost adjustment interval refers to the value range of the historical cost adjustment value in the target historical change case associated with the feature cluster.

[0118] In the embodiments of the present application, for each feature cluster, the size difference value of the target historical change case associated with the corresponding intermediate feature vector is extracted, a first size difference interval is calculated, the first size difference interval is [min(size difference value set), max(size difference value set)], the size difference value set is a set composed of all size difference values of the target historical change cases associated with the feature cluster, min(size difference value set) is the minimum value in the set, and max(size difference value set) is the maximum value in the set; similarly, the material difference degree set is composed of the material difference degrees, and a material difference range is calculated, the material difference range is [min(material difference degree set), max(material difference degree set)]; the historical cost adjustment value set is composed of the historical cost adjustment values, and a historical cost adjustment interval is calculated, the historical cost adjustment interval is [min(historical cost adjustment value set), max(historical cost adjustment value set)], and the first size difference interval, the material difference range, and the historical cost adjustment interval of each feature cluster are determined.

[0119] Step 322: The first size difference interval and the historical cost adjustment interval of each feature cluster are analyzed to obtain a first corresponding relationship, and the material difference range and the historical cost adjustment interval of each feature cluster are analyzed to obtain a second corresponding relationship.

[0120] In this step, the first corresponding relationship refers to the mapping relationship between the first size difference interval and the historical cost adjustment interval of the feature cluster. The second corresponding relationship refers to the mapping relationship between the material difference range and the historical cost adjustment interval of the feature cluster.

[0121] In the embodiments of the present application, for each feature cluster, the first size difference interval and the historical cost adjustment interval are analyzed, and the historical cost adjustment interval corresponding to the historical cost adjustment value when the size difference value is in the first size difference interval is recorded to form the first corresponding relationship; at the same time, the material difference range and the historical cost adjustment interval are analyzed, and the historical cost adjustment interval corresponding to the historical cost adjustment value when the material difference degree is in the material difference range is recorded to form the second corresponding relationship.

[0122] Step 323: Based on the first corresponding relationship and the second corresponding relationship, the combination rule of the size difference value, the material difference degree, and the historical cost adjustment value in each feature cluster is determined to form a preliminary cost adjustment rule of each feature cluster.

[0123] In the embodiments of the present application, the first corresponding relationship and the second corresponding relationship of each feature cluster are combined, the specific distribution rule of the historical cost adjustment value in the historical cost adjustment interval when the size difference value is in the first size difference interval and the material difference degree is in the material difference range is analyzed, the fixed combination mode between the size difference value, the material difference degree, and the historical cost adjustment value is refined, the mode is converted into a rule provision, and a preliminary cost adjustment rule of each feature cluster is formed.

[0124] Step 324: The preliminary cost adjustment rules corresponding to the first size difference interval and the material difference range with the overlapping numerical range are integrated to obtain a plurality of intermediate cost adjustment rules.

[0125] In this step, the overlapping numerical range refers to the overlapping numerical part between two or more intervals. The intermediate cost adjustment rule refers to the rule formed by integrating the preliminary cost adjustment rules with the first size difference interval and the material difference range both having the overlapping numerical range.

[0126] In the embodiments of the present application, all the preliminary cost adjustment rules are compared, the intersection of the first size difference intervals of any two rules is calculated, if the intersection is not empty, it is determined that there is an overlapping numerical range, and the intersection of the material difference ranges of the two rules is calculated, if the intersection is also not empty, the two preliminary cost adjustment rules are integrated into one rule, and the first size difference interval, the material difference range and the historical cost adjustment interval are combined to obtain an intermediate cost adjustment rule; the process is repeated until all the preliminary cost adjustment rules meeting the conditions are integrated to obtain a plurality of intermediate cost adjustment rules.

[0127] Step 325: If different intermediate cost adjustment rules exist rule conflicts, the intermediate cost adjustment rule with more feature clusters is retained as a candidate cost adjustment rule, if the intermediate cost adjustment rules do not exist rule conflicts, the intermediate cost adjustment rules are taken as candidate cost adjustment rules, and all the candidate cost adjustment rules are integrated to obtain a cost adjustment rule.

[0128] In this step, the number of feature clusters refers to the number of feature clusters supporting a certain intermediate cost adjustment rule. The candidate cost adjustment rule refers to the intermediate cost adjustment rule retained after the conflict processing.

[0129] In the embodiments of the present application, all the intermediate cost adjustment rules are queried, if there are two or more intermediate cost adjustment rules corresponding to different historical cost adjustment intervals in the same first size difference interval and material difference range, it is determined that there is a rule conflict; for the rules with conflicts, the number of feature clusters corresponding to each rule is counted, and the rule with more feature clusters is retained as a candidate cost adjustment rule; for the intermediate cost adjustment rules without conflicts, they are directly taken as candidate cost adjustment rules; all the candidate cost adjustment rules are integrated to remove the repeated contents to obtain a cost adjustment rule.

[0130] The embodiments of the present application determine the interval range of the feature clusters, analyze the interval corresponding relationship, refine and combine the rules to form preliminary rules, integrate the overlapping rules and handle the conflicts, and finally obtain a unified cost adjustment rule, which ensures the consistency and effectiveness of the rule and provides a reliable rule basis for cost analysis.

[0131] The application provides an embodiment, step 104, based on the size parameter difference and the cost adjustment range in the target historical change case, feature quantization is performed on the size difference between the original size parameter and the changed size parameter to obtain a feature quantization vector, and the feature quantization vector specifically includes the following steps:

[0132] Step 401: Distribution feature analysis is performed on the size parameter difference between the historical original size parameter and the historical changed size parameter in the target historical change case, interval division rules are determined, and cost adjustment range features are extracted from the target historical change case.

[0133] In this step, the interval division rules refer to rules for dividing the size parameter difference into multiple intervals. The cost adjustment range features refer to features of the cost adjustment range in the target historical change case.

[0134] In the embodiment of the application, the difference between the historical original size parameter and the historical changed size parameter in the target historical change case is analyzed, the frequency distribution and the density are counted, the standard for dividing the size parameter difference into multiple continuous intervals is determined according to the distribution result, the interval division rules are obtained, and the cost adjustment range features are extracted from the target historical change case, the numerical range, the change trend and other features are extracted, and the cost adjustment range features are obtained.

[0135] Step 402: According to the interval division rules, the size parameter difference is divided into multiple second size difference intervals, the second size difference intervals and the corresponding cost adjustment range features of the target historical change case are associated, and a difference feature association table is formed.

[0136] In this step, the second size difference interval refers to an interval obtained by dividing the size parameter difference according to the interval division rules. The difference feature association table refers to a table associating the second size difference interval and the corresponding cost adjustment range feature.

[0137] In the embodiment of the application, according to the interval division rules, all size parameter differences in the target historical change case are divided into multiple continuous and non-overlapping second size difference intervals. Each second size difference interval is respectively corresponding to the cost adjustment range feature of the corresponding target historical change case, and is arranged into a table in a preset format to form a difference feature association table.

[0138] Step 403: The size difference between the original size parameter and the changed size parameter is calculated, and the target size difference interval of the size difference is determined according to the difference feature association table.

[0139] In this step, the target size difference interval refers to a second size difference interval to which the size difference belongs.

[0140] In the embodiment of the present application, the size difference is calculated as size difference = original size parameter - changed size parameter; the second size difference interval containing the size difference is found by consulting the difference feature association table, and the interval is taken as the target size difference interval of the size difference.

[0141] Step 404: quantifying the cost adjustment amplitude feature value associated with the target size difference interval in the difference feature association table into a cost adjustment amplitude feature quantification value, to combine the size difference and the cost adjustment amplitude feature quantification value to obtain a feature quantification vector.

[0142] In this step, the cost adjustment amplitude feature quantification value refers to a quantification value converted from the cost adjustment amplitude feature.

[0143] In the embodiment of the present application, the cost adjustment amplitude feature associated with the target size difference interval in the difference feature association table is converted into a specific value according to a preset quantification rule to obtain a cost adjustment amplitude feature quantification value; the size difference and the quantification value are combined in a preset order to form a feature quantification vector.

[0144] The embodiment of the present application realizes accurate feature quantification of the size difference by interval division and feature association of the size parameter difference, calculates the size difference and determines the interval to which it belongs, and combines the quantified cost adjustment amplitude feature to form a feature quantification vector, thereby providing a standardized feature basis for subsequent matching calculation with the cost adjustment rule.

[0145] The present application provides a specific embodiment, step 501, based on the cost adjustment rule, the feature quantification vector is matched and calculated to obtain the cost adjustment coefficient, the historical reference cost corresponding to the original building component is extracted from the target historical change case, and the cost adjustment coefficient is combined to calculate the cost of building engineering after change, specifically including the following steps:

[0146] Step 501: respectively calculate the matching degree of each candidate cost adjustment rule in the cost adjustment rule and the feature quantification vector, and take the candidate cost adjustment rule corresponding to the matching degree with the largest value as the target cost adjustment rule.

[0147] In this step, the matching degree refers to the matching degree between the feature quantification vector and the candidate cost adjustment rule. The target cost adjustment rule refers to the rule with the largest matching degree between the feature quantification vector and the multiple candidate cost adjustment rules.

[0148] In the embodiments of the present application, for each candidate cost adjustment rule in the cost adjustment rules, a corresponding candidate rule feature vector is constructed for each candidate cost adjustment rule; a matching degree of the feature quantization vector and each candidate rule feature vector is calculated wherein U is the feature quantization vector, V is the candidate rule feature vector, is the dot product of the two vectors, |U| is the length of U, and |V| is the length of V; the maximum matching degree is selected from the plurality of calculated matching degrees, and the candidate cost adjustment rule corresponding to the matching degree is determined as the target cost adjustment rule.

[0149] Step 502: Extracting an adjustment parameter corresponding to a third size difference interval and a material difference range matching the feature quantization vector from the target cost adjustment rule to calculate a cost adjustment coefficient.

[0150] In this step, the third size difference interval refers to a size difference interval contained in the target cost adjustment rule. The adjustment parameter refers to a parameter in the target cost adjustment rule for calculating the cost adjustment coefficient.

[0151] In the embodiments of the present application, the third size difference interval matching the size difference value in the feature quantization vector and the material difference range corresponding to the interval are selected from the target cost adjustment rule; the adjustment parameter associated with the third size difference interval and the material difference range is extracted, and the cost adjustment coefficient is calculated by combining the size difference value in the feature quantization vector and the cost adjustment amplitude quantization value, i.e., cost adjustment coefficient = adjustment parameter × (size difference value / third size difference interval span) × (cost adjustment amplitude quantization value / interval amplitude mean).

[0152] Step 503: Extracting a matching historical change case with the same component type and specification parameter as the original building component from the target historical change case, taking the initial cost of the matching historical change case as a historical reference cost, and calculating the cost of the building engineering after the change by combining the cost adjustment coefficient.

[0153] In this step, the component type refers to the category to which the building component belongs. The specification parameter refers to the parameter describing the specification of the building component. The matching historical change case refers to the target historical change case with the same component type and specification parameter as the original building component. The initial cost refers to the cost of the building component before the change in the matching historical change case.

[0154] In the embodiments of the present application, the cases with the same component type and specification parameter as the original building component are selected from the target historical change case as matching historical change cases; the initial cost in the matching historical change case is extracted and taken as a historical reference cost to calculate the cost of the building engineering after the change, i.e., cost of building engineering after change = historical reference cost × cost adjustment coefficient.

[0155] The embodiment of the application filters the target rule by calculating the matching degree, extracts the parameter to calculate the adjustment coefficient, and obtains the changed cost combined with the initial cost of the matching case, realizes the accurate cost calculation based on the rule and historical data, and improves the accuracy and reliability of the cost analysis.

[0156] For example, the feature quantization vector U of the A component is [-1, 2.5], wherein -1 is the size difference value, and 2.5 is the cost adjustment amplitude feature quantization value; the third size difference interval [-2, 0] of rule 1 in the candidate rule in the cost adjustment rule corresponds to the cost adjustment amplitude feature quantization value range [1, 3], the candidate rule feature vector V1 of rule 1 is [-1, 2], that is, the interval midpoint, the third size difference interval [0, 2] of rule 2 corresponds to the cost adjustment amplitude feature quantization value range [2, 4], and the candidate rule feature vector V2 of rule 2 is [1, 3]. Based on rule 1, = (-1) * (-1) + 2.5 * 2 = 1 + 5 = 6; |U| = sqrt[(-1)² + 2.5²] = sqrt(1 + 6.25) = sqrt 7.25 ≈ 2.69; |V1| = sqrt[(-1)² + 2²] = sqrt(1 + 4) = sqrt 5 ≈ 2.24; matching degree = 6 / (2.69 * 2.24) ≈ 6 / 6.03 ≈ 0.995. Based on rule 2, = (-1) * 1 + 2.5 * 3 = -1 + 7.5 = 6.5; |V2| = sqrt(1² + 3²) = sqrt 10 ≈ 3.16; matching degree = 6.5 / (2.69 * 3.16) ≈ 6.5 / 8.5 ≈ 0.765. Rule 1 is selected as the target cost adjustment rule; the third size difference interval [-2, 0] of rule 1 is extracted, the span = 0 - (-2) = 2, the adjustment parameter 1.2, the interval amplitude mean value = (1 + 3) / 2 = 2 ten thousand yuan, the cost adjustment coefficient = 1.2 * (size difference value / span) * (cost adjustment amplitude feature quantization value / interval amplitude mean value) = 1.2 * (-1 / 2) * (2.5 / 2) = 1.2 * 0.5 * 1.25 = 0.75; the matching case with the same type and consistent specification parameters of the A component is extracted from the target historical change case, the initial cost is 10,000 yuan, and the changed cost of the building engineering is calculated as 10 * 0.75 = 7.5 ten thousand yuan.

[0157] Figure 3 FIG. 1 is a structural schematic diagram of one specific embodiment of a building engineering cost analysis system based on deep learning provided by the embodiment of the application, which is referred to Figure 3 The system can include:

[0158] The acquisition module 21 is used to acquire the original size parameters of the original building components and the modified size parameters of the modified building components after the original building components have been modified, and to obtain a historical engineering change cost dataset. The historical engineering change cost dataset includes multiple change types, and each change type corresponds to multiple historical change cases.

[0159] Selection module 22 is used to obtain the comparison similarity based on the original size parameters and the changed size parameters, and select the target historical change case that matches the comparison similarity from the historical engineering change cost dataset;

[0160] Analysis module 23 is used to perform feature analysis on the original historical building component feature set, the post-historical building component feature set, and historical cost data in the target historical change case through deep learning, and to obtain cost adjustment rules;

[0161] Quantization module 24 is used to perform feature quantization on the size difference between the original size parameter and the changed size parameter based on the size parameter difference and cost adjustment range in the target historical change case, and obtain a feature quantization vector;

[0162] The calculation module 25 is used to perform matching calculation on the feature quantization vector based on the cost adjustment rules to obtain the cost adjustment coefficient, extract the historical benchmark cost corresponding to the original building component from the target historical change case, and calculate the cost after the building project change by combining the cost adjustment coefficient.

[0163] This application provides a deep learning-based construction cost analysis system to implement the aforementioned deep learning-based construction cost analysis method. Therefore, the specific implementation of the deep learning-based construction cost analysis system can be found in the previous section on the embodiment of the deep learning-based construction cost analysis method. The specific implementation can be referred to the description of the corresponding embodiments, which will not be repeated here.

[0164] This application also provides an electronic device, comprising: a memory for storing a computer program; and a processor for executing the computer program to implement the steps of any of the above-described deep learning-based construction cost analysis methods.

[0165] This application also provides a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the steps of any of the above-described deep learning-based construction engineering cost analysis methods.

[0166] In an example embodiment, the computer readable storage medium described above can include, but is not limited to, a U disk, a read-only memory, a random access memory, a mobile hard disk, a magnetic disk or an optical disk, and various media that can store computer programs.

[0167] Embodiments of the present application also provide a computer program product, the computer program product comprising a computer program, the computer program being executed by a processor to implement the steps in any of the above deep learning-based construction engineering cost analysis method embodiments.

[0168] The skilled person can further realize that the units and algorithm steps of the examples described in connection with the embodiments disclosed herein can be realized in electronic hardware, computer software or a combination of both. In order to clearly illustrate the interchangeability of hardware and software, the components and steps of the examples have been described in general terms in the above description. Whether the functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. The skilled person 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.

[0169] The above provides a deep learning-based construction engineering cost analysis method and system. The principles and implementation modes of the present application are described in this paper using specific examples. The above example descriptions are only used to help understand the method and its core idea. It should be pointed out that for ordinary skilled persons in the technical field, without departing from the principles of the present application, the present application can be improved and modified in several ways. These improvements and modifications also fall within the scope of the present application.

Claims

1. A deep learning-based construction engineering cost analysis method, characterized by, The method comprises the following steps: obtain a historical engineering change cost dataset, the historical engineering change cost dataset comprising a plurality of change types, each change type corresponding to a plurality of historical change cases; based on the original size parameters of the original building component and the changed size parameters corresponding to the original building component after the change, obtain a similarity comparison, and select a target historical change case from the historical engineering change cost dataset that matches the similarity comparison; through deep learning, analyze the features of the historical original building component feature set, the historical changed building component feature set, and the historical cost data in the target historical change case to obtain a cost adjustment rule; based on the cost adjustment rule, extract the historical benchmark cost corresponding to the original building component from the target historical change case, and calculate the changed cost of the building engineering based on the cost adjustment coefficient; through deep learning, analyze the features of the historical original building component feature set, the historical changed building component feature set, and the historical cost data in the target historical change case to obtain a cost adjustment rule, comprising: extract the historical original size parameters and historical original material parameters from the historical original building component feature set of the target historical change case, extract the historical changed size parameters and historical changed material parameters from the historical changed building component feature set, and extract the historical cost adjustment value from the historical cost data; calculate the size difference between the historical original size parameters and the corresponding historical changed size parameters, the material difference between the historical original material parameters and the corresponding historical changed material parameters in the same target historical change case, and combine the size difference, the material difference, and the historical cost adjustment value to form a plurality of feature cost sample pairs; through deep learning, analyze the associated features of the feature cost sample pairs to obtain a plurality of intermediate feature vectors to form a plurality of feature clusters; analyze each feature cluster to form a plurality of preliminary cost adjustment rules, integrate all preliminary cost adjustment rules to obtain a cost adjustment rule; after analyzing the features of the historical original building component feature set, the historical changed building component feature set, and the historical cost data in the target historical change case through deep learning to obtain a cost adjustment rule, the method further comprises the following steps: based on the size parameter difference and the cost adjustment amplitude in the target historical change case, quantize the size difference between the original size parameters and the changed size parameters to obtain a feature quantization vector; based on the cost adjustment rule, perform matching calculation on the feature quantization vector to obtain a cost adjustment coefficient; based on the cost adjustment rule, perform matching calculation on the feature quantization vector to obtain a cost adjustment coefficient, comprising: calculate the matching degree of each candidate cost adjustment rule in the cost adjustment rule and the feature quantization vector respectively, and select the candidate cost adjustment rule corresponding to the maximum matching degree as the target cost adjustment rule; Extract the third size difference interval, material difference range corresponding adjustment parameter matched with the feature quantitative vector from the target cost adjustment rule, to calculate the cost adjustment coefficient.

2. The method of claim 1, wherein, Based on the original size parameters of the original building component and the changed size parameters corresponding to the changed original building component, a comparison similarity is obtained, and a target historical change case matching the comparison similarity is selected from the historical engineering change cost data set, including: Based on the original size parameters and the changed size parameters, a length difference, a width difference and a height difference are calculated, and the original size parameters, the changed size parameters, the length difference, the width difference and the height difference are combined to form a change feature vector; The change feature vector and the historical change feature vector of each historical change case in the historical engineering change cost data set are compared to obtain a plurality of comparison similarities, and the historical change case with a comparison similarity greater than a preset threshold is taken as a target historical change case.

3. The method of claim 1, wherein, Through deep learning, the intermediate feature vectors are obtained by correlating the feature cost sample pairs, including: The size difference, material difference degree and historical cost adjustment value corresponding to each feature cost sample pair are vector converted to obtain a plurality of basic feature vectors; Through deep learning, the numerical distribution characteristics of each basic feature vector are analyzed to obtain the primary correlation characteristics, the intermediate correlation characteristics and the high-level correlation characteristics of each basic feature vector. The primary correlation characteristics are formed by combining the numerical range characteristics, the fluctuation characteristics and the change amplitude characteristics. The intermediate correlation characteristics are formed by combining the parameter co-occurrence frequency characteristics, the numerical change coordination characteristics and the proportion correlation characteristics. The high-level correlation characteristics are formed by combining the parameter combination mode characteristics and the numerical linkage law characteristics. The primary correlation characteristics, the intermediate correlation characteristics and the high-level correlation characteristics corresponding to each basic feature vector are integrated to form a plurality of intermediate feature vectors.

4. The method of claim 1, wherein, Each feature cluster is analyzed to form a plurality of preliminary cost adjustment rules, and all preliminary cost adjustment rules are integrated to obtain a cost adjustment rule, including: According to the intermediate feature vectors corresponding to each feature cluster, the first size difference interval, the material difference range and the historical cost adjustment interval of the target historical change case associated with each feature cluster are determined. The first size difference interval and the historical cost adjustment interval of each feature cluster are analyzed, and the material difference range and the historical cost adjustment interval of each feature cluster are analyzed to determine the combination rule of the size difference, the material difference degree and the historical cost adjustment value in each feature cluster, so as to form the preliminary cost adjustment rule of each feature cluster. The preliminary cost adjustment rules corresponding to the first size difference interval and the material difference range with overlapping numerical ranges are integrated to obtain a plurality of intermediate cost adjustment rules. If different intermediate cost adjustment rules exist rule conflicts, the intermediate cost adjustment rule with more feature clusters is reserved as a candidate cost adjustment rule, if the intermediate cost adjustment rule does not exist rule conflicts, the intermediate cost adjustment rule is taken as a candidate cost adjustment rule, all candidate cost adjustment rules are integrated to obtain the cost adjustment rule. 5.A deep learning-based construction engineering cost analysis system, characterized by, Comprise: The acquisition module is used for obtaining a historical engineering change cost dataset, and the historical engineering change cost dataset comprises a plurality of change types, and each change type corresponds to a plurality of historical change cases; The selection module is used for obtaining a comparison similarity based on an original size parameter of an original building component and a changed size parameter corresponding to the original building component after change, and selecting a target historical change case matching the comparison similarity from the historical engineering change cost dataset; The analysis module is used for performing feature analysis on a historical original building component feature set, a historical changed building component feature set and historical cost data in the target historical change case through deep learning to obtain a cost adjustment rule; The calculation module is used for extracting a historical benchmark cost corresponding to the original building component from the target historical change case based on the cost adjustment rule, and calculating a building engineering changed cost by combining a cost adjustment coefficient; Through deep learning, feature analysis is performed on a historical original building component feature set, a historical changed building component feature set and historical cost data in the target historical change case to obtain a cost adjustment rule, comprising: From the historical original building component feature set of the target historical change case, historical original size parameters and historical original material parameters are extracted, from the historical changed building component feature set, historical changed size parameters and historical changed material parameters are extracted, and from the historical cost data, a historical cost adjustment value is extracted; The size difference between the historical original size parameters and the corresponding historical changed size parameters in the same target historical change case, and the material difference degree between the historical original material parameters and the corresponding historical changed material parameters are calculated, and the size difference, the material difference degree and the historical cost adjustment value are combined to form a plurality of feature cost sample pairs; Through deep learning, associated feature analysis is performed on the feature cost sample pairs to obtain a plurality of intermediate feature vectors to form a plurality of feature clusters; Each feature cluster is analyzed to form a plurality of preliminary cost adjustment rules, and all preliminary cost adjustment rules are integrated to obtain a cost adjustment rule; After the feature analysis on the historical original building component feature set, the historical changed building component feature set and the historical cost data in the target historical change case through deep learning to obtain the cost adjustment rule, further comprising: Based on the size parameter difference value and the cost adjustment amplitude in the target historical change case, the size difference between the original size parameter and the changed size parameter is quantified to obtain a feature quantization vector; Based on the cost adjustment rule, the feature quantization vector is matched and calculated to obtain a cost adjustment coefficient; Based on the cost adjustment rule, the feature quantization vector is matched to obtain a cost adjustment coefficient, comprising: Respectively calculate the matching degree of each candidate cost adjustment rule in the cost adjustment rule and the feature quantization vector, and take the candidate cost adjustment rule corresponding to the maximum matching degree as the target cost adjustment rule; From the target cost adjustment rule, the third size difference interval and the material difference range corresponding to the adjustment parameter matched with the feature quantization vector are extracted to calculate the cost adjustment coefficient.

6. An electronic device, comprising: Comprise: Memory for storing computer programs; Processor for executing the computer program to realize the steps of the building engineering cost analysis method based on deep learning in any one of claims 1 to 4.

7. A computer-readable storage medium, characterized in that, The computer readable storage medium stores a computer program, and the computer program is executed by the processor to realize the building engineering cost analysis method based on deep learning in any one of claims 1 to 4.

Images