Rail transit project cost intelligent assessment method based on big data

Abstract

The invention belongs to the technical field of rail transit cost assessment, and particularly relates to a rail transit project cost intelligent assessment method based on big data, which comprises the steps of inputting a cost file of a project to be assessed, including sub-items, a list, a quota and labor, material and machine information; performing index physical examination early warning, matching historical sample data, rejecting abnormal values through quartile, calculating a confidence interval, and detecting cost deviation; checking the features, comparing with a standard feature configuration library, and verifying the integrity and normalization of the engineering features; through cooperation of index physical examination early warning and feature examination, the problem of engineering type recognition errors can be quickly responded, so that the defects of a static model are overcome, construction cost files with engineering type labeling errors can be quickly responded, the false alarm frequency can be reduced, and the defects of the construction cost files can be quickly analyzed based on a multi-stage examination mechanism. Compared with manual checking, the method and the device have the advantages that the speed and the efficiency are higher, and the checking is more comprehensive.

Description

Technical Field

[0001] This invention belongs to the field of rail transit cost assessment technology, specifically a big data-based intelligent assessment method for rail transit engineering costs. Background Technology

[0002] During the editing stage, the data in the cost documents for rail transit projects often need to undergo multiple rounds of review to check the rationality of the cost, the completeness and accuracy of the information, etc.

[0003] Currently, the solutions mainly involve manual inspection by design institute personnel, manual summarization, calculation by multiple people using spreadsheets, judging the reasonableness of the cost based on experience, applying for leadership approval offline, and then sending the legal contract for review. As the above shows, rail transit engineering cost estimation involves numerous disciplines and comprises a wide range of components, including itemized costs, lists of quantities, quotas, and information on labor, materials, and machinery costs, as well as regulatory fees, taxes, and project characteristics. During the preparation of the preliminary estimate and contract cost documents, each item needs to be checked to ensure the reasonableness of costs, compliance with project characteristics, and completeness. In the past, relying on manual labor and human experience to evaluate rail transit engineering cost documents was time-consuming, inefficient, lacked scientific basis, and heavily depended on experienced cost estimators.

[0004] Therefore, this invention provides an intelligent cost assessment method for rail transit engineering based on big data. Summary of the Invention

[0005] In order to overcome the shortcomings of the prior art, at least one technical problem raised in the background art is solved.

[0006] The technical solution adopted by this invention to solve its technical problem is: a big data-based intelligent cost assessment method for rail transit engineering, comprising the following steps: S1: Input the cost documents for the project to be evaluated, including sub-items, bill of quantities, quotas, and information on labor, materials, and machinery; S2: Execute health check alerts based on performance indicators, match historical sample data, remove outliers using quartiles, and calculate confidence intervals. If the unit cost exceeds the confidence interval, it is determined that there is a cost deviation; When S2 detects a cost deviation, it sends a first instruction signal to S3 to trigger a project type matching verification. S3: Perform feature checks to verify the completeness and compliance of engineering features, and encode the verification results as feedback signals to send to S2. When S2 receives a feedback signal, it dynamically adjusts the confidence interval. S4: Perform valid data checks, scan the validity of key fields based on the validation rule base, and if invalid data is identified, send a second instruction model carrying the node path to S5; S5: Responds to the second instruction model, locates the node path and associates with policy clauses, generates a specification amendment instruction and pushes it to S4; S6: Generate a comprehensive assessment report, summarizing the inspection results and corrective suggestions.

[0007] Preferably, in step S2, the method for early warning of health indicators includes the following steps: S21: Select similar historical samples based on project type, region, and year characteristics, and verify the consistency of data sources through blockchain to generate reliable sample labels; S22: Use quartiles to remove outliers and calculate the mean of the remaining samples. and standard deviation ; S23: Construct a multi-dimensional confidence interval system based on the classification of rail transit engineering elements, including: Material price range: ; Labor cost range: ; The comprehensive cost range is expressed as follows: ; S24: Establish a dynamic interval prediction model for multi-dimensional confidence intervals: Predicted value calculation: ; in, As a time decay weight; For viewing; Interval half-width calculation: ; in, This is the error amplification factor; Output prediction interval: ; S25: When the current indicator exceeds the range, generate a first instruction signal containing the deviation code; S26: Receive the feature verification results fed back by S3 and dynamically adjust the confidence interval.

[0008] Preferably, in step S3, the method for implementing feature inspection includes the following steps: S31: Construct a standard feature configuration library, defining the mandatory fields, data types, and format rules for feature items; S32: Verify feature values ​​using enumeration matching and regular expressions; S33: Generate semantic correction suggestions for missing items; S34: In response to the first instruction signal, prioritize verifying the matching of the project type; S35: Encode the feature verification result into a feature feedback signal and send it to S2.

[0009] Preferably, in step S4, the valid data check includes the following steps: S41: Use a tree-based recursive traversal algorithm to scan multi-level nodes in the cost breakdown file; S42: Use a dynamic rule engine to validate key fields and determine whether the key field data is invalid; S43: When the data is invalid, generate a second instruction signal containing the node path; The dynamic rule engine has a built-in validation rule library for validating key fields.

[0010] Preferably, in step S5, the specification check includes the following steps: S51: Built-in specification library, which includes bill of quantities specifications and tax calculation rules; S52: Respond to the second command signal and locate the abnormal node path; S53: Related policy provisions and explanations; S54: Generate specification amendment instructions based on policy provisions and push them to the dynamic rule engine of S4 to update the verification rule base.

[0011] Preferably, the verification rule base is executed as follows: S61: Update rules in response to new policies; S62: Receives the protocol correction command from S54 and updates the verification rule base in real time.

[0012] Preferably, it also includes data security processing steps: S71: Homomorphically encrypt sensitive fields, including contract amount and supplier information; S72: Generate a desensitization assessment report for authorized users to view the complete data.

[0013] Preferably, the historical sample data is integrated from multiple sources, including publicly available government databases, private enterprise databases, and building materials platform data; The historical sample data is verified for cross-source data consistency through blockchain, and the confidence weight factor for S21 sample screening is calculated based on the verification results.

[0014] Preferably, in step S6, the evaluation report generation step includes: S61: Build multi-dimensional visualization dashboards, including: The index deviation radar chart is used to display the S2 results; The collaborative path flow diagram is used to show the linkage relationship between S2 and S3 and between S4 and S5; A data error distribution tree diagram is used to highlight the S4 abnormal node.

[0015] Preferably, the data error distribution tree diagram is also used to integrate three-dimensional analysis coordinate axes, including: X-axis positioning: Determine the spatial criticality level of the abnormal node based on its hierarchical depth in the sub-item cost document tree and the proportion of nodes within its influence range; Y-axis tracing: Based on blockchain evidence storage, historical modification records are extracted, and the time anomaly index of the node is calculated by comparing the deviation of the current value from the historical fluctuation trend. Z-axis assessment: Based on the matching degree between policy provisions and node semantics, combined with the policy timeliness and violation consequences database, a compliance impact rating is generated; When a node’s spatial criticality level, time anomaly index, and compliance impact rating all exceed the industry benchmark threshold, the compliance check process will be automatically executed.

[0016] The beneficial effects of this invention are as follows: 1. The intelligent cost assessment method for rail transit engineering based on big data described in this invention can quickly respond to the problem of incorrect engineering type identification through the synergy of indicator health check and feature inspection, thereby solving the defects of static models. For cost documents with incorrect engineering type labeling, it can respond quickly and reduce the frequency of false alarms. Based on the multi-level inspection mechanism, it can quickly analyze the defects of cost documents and output the error content simultaneously. Compared with manual inspection, it is faster, more efficient and more comprehensive.

[0017] 2. The intelligent cost assessment method for rail transit engineering based on big data described in this invention utilizes a multi-level inspection mechanism to assess cost documents. If a second instruction signal is generated during the valid data inspection, it indicates that invalid data exists in the cost document. For complex data types in the cost document, a second instruction signal can be generated when invalid data is identified and transmitted to the specification inspection. The specification inspection is used to achieve feedback and updates, enabling real-time reverse-driving of the dynamic rule engine to update the verification rule base based on invalid data. Attached Figure Description

[0018] The invention will now be further described with reference to the accompanying drawings.

[0019] Figure 1 This is a flowchart of the present invention; Detailed Implementation

[0020] To make the technical means, creative features, objectives and effects of this invention easier to understand, the invention will be further described below in conjunction with specific embodiments.

[0021] like Figure 1 As shown in the embodiment of the present invention, an intelligent cost assessment method for rail transit engineering based on big data includes the following steps: S1: Input the cost documents for the project to be evaluated, including sub-items, bill of quantities, quotas, and information on labor, materials, and machinery; S2: Execute health check alerts based on performance indicators, match historical sample data, remove outliers using quartiles, and calculate confidence intervals. If the unit cost exceeds the confidence interval, it is determined that there is a cost deviation; When S2 detects a cost deviation, it sends a first instruction signal to S3 to trigger a project type matching verification. S3: Perform feature checks to verify the completeness and compliance of engineering features, and encode the verification results as feedback signals to send to S2. When S2 receives a feedback signal, it dynamically adjusts the confidence interval. S4: Perform valid data checks, scan the validity of key fields based on the validation rule base, and if invalid data is identified, send a second instruction model carrying the node path to S5; S5: Responds to the second instruction model, locates the node path and associates with policy clauses, generates a specification amendment instruction and pushes it to S4; S6: Generate a comprehensive assessment report, summarizing inspection results and corrective recommendations; The cost estimation of rail transit projects involves numerous disciplines, and the cost documents consist of a wide range of contents, including itemized costs, lists of quantities, quotas, and information on labor, materials, and machinery costs, as well as regulatory fees, taxes, and project characteristics. During the preparation of the preliminary estimate and contract cost documents, each item needs to be checked to ensure the reasonableness of costs and the compliance and completeness of project characteristics. In the past, the evaluation of rail transit project cost documents relied on manual judgment and human experience, which was time-consuming, inefficient, lacked scientific basis, and heavily depended on experienced cost estimators.

[0022] In this embodiment, a database is used as the sample base, combined with algorithm calculations, to detect the rationality of the current project cost. Compared with manual detection, it is more objective. Based on the configured multi-level inspection mechanism, the computer can quickly check the project cost content, including compliance, correctness, and completeness. It is faster, more efficient, and more comprehensive than manual inspection. In practical application, a multi-level inspection mechanism is constructed using indicator health check and early warning, feature inspection, valid data inspection, and specification inspection. When the cost document is input, historical samples are first filtered based on the indicator health check and early warning according to the marked project type and other data. Then, the unit cost range is calculated based on the unit cost recorded in the historical samples. Then, the project value in the current cost document, that is, whether the unit cost range falls within the unit cost range calculated by the historical samples, is identified. If it falls within the range, it is marked as "reasonable". Otherwise, a first instruction signal is sent to S3 to trigger the project type matching review. It can be understood that, for example, the project type recorded in the cost document... If the project type is elevated, then based on this project type, historical samples are selected, for example, 50 historical projects are selected. Then, based on identification and calculation, the cost per square meter of these 50 historical projects is output, forming a cost range per square meter, for example, "1800, 2200" yuan. If, after identification, the project value in this cost document, that is, the unit cost, is 2000 yuan, it is marked as reasonable. If the project value in this cost document is 2400 yuan, it is unreasonable. At this point, the first instruction signal is sent to the feature check, triggering the project type matching review. The feature check identifies the content of the cost document and finds that the actual project type should be an underground tunnel, simultaneously generating the signal: {Error type: "Project type mismatch", Actual value: "Underground tunnel"}. Because of the change in project type, in S2, the confidence interval calculation and the detection cost deviation also need to change accordingly, that is, the unit cost range needs to be changed. After selecting historical samples of "underground tunnel", the correct cost range calculated should be "2200,If the project value in this cost document is 2400 yuan, it is marked as reasonable. Based on the logical feedback of S2 and S3, it can identify project type labeling errors that cannot be identified by the static model, promptly identify errors, and thus cancel the warning. Among them, the feature check in S3, as mentioned above, is used to verify the standardization and completeness of the cost document. For example, it verifies the project type. If the elevated structure verification passes, the project type is correct. If the cost range does not fall within the unit cost range, it is marked as unreasonable; otherwise, it is reasonable. In addition, other features need to be verified, such as the seismic resistance level. If the seismic resistance level recorded in this cost document is level 7, but it is recorded as level 8 according to the new policy regulations, the verification fails, and a signal is generated, such as {Error type: "Feature value expired", field: "Seismic resistance level", required value: "level 8"}. Then, a valid data check is performed based on scanning key data items. For example, scanning galvanized steel ducts, if the verification finds that the project quantity is correct, but the project quantity unit is missing, a signal is generated simultaneously: {Error type: "Unit missing", node path: "Ventilation-galvanized"}. Based on the above, the evaluation of traditional cost documents relies on manual labor and requires manual tracking of nodes. However, in this embodiment, based on valid data checking, missing data can be accurately located and node paths can be marked for easy modification later. Then, a specification check is performed, including: receiving the second instruction signal output by the valid data check. As mentioned above, if a unit missing problem is found during the valid data check, the second instruction signal is received in the specification check step, and the "unit missing" node is located. Then, the specification library is matched. For example, according to Article 3.2.1 of GB50500-2013, the unit is required to be filled in. The associated policy is: Jianbiao

[2023] No. 18 strengthens unit verification, and the output correction suggestion is: "Automatically associate unit 'square meters'". Based on the above, a multi-level check mechanism is used, from indicator experience to feature check to valid data check to specification check, to achieve rapid check of cost documents. Compared with manual check, it is faster and can reduce check omissions. Finally, a report is generated according to the above multi-level check mechanism, for example: Composite Anomaly Report: [Critical] Project type incorrectly labeled (labeled: elevated section, actual: underground tunnel). This leads to: distorted indicator warnings (actual construction cost is lower than the underground tunnel standard). - Recommendation: Update the project type and recalculate. [WARNING] Seismic resistance level not met (current: level 7, new standard: level 8) [Error] Unit missing for galvanized steel sheet - Automatically corrected to 'square meters'.

[0023] Based on the above, in this embodiment, the collaboration of S2 and S3 can quickly respond to the problem of incorrect project type identification, thereby solving the defects of the static model. For cost documents with incorrect project type labeling, it can respond quickly, reduce the frequency of false alarms, and based on the multi-level inspection mechanism, it can quickly analyze the defects of cost documents and output the error content synchronously. Compared with manual inspection, it is faster and more efficient, and the inspection is more comprehensive.

[0024] In one embodiment, the method for early warning of health indicators in step S2 includes the following steps: S21: Select similar historical samples based on project type, region, and year characteristics, and verify the consistency of data sources through blockchain to generate reliable sample labels; S22: Use quartiles to remove outliers and calculate the mean of the remaining samples. and standard deviation ; S23: Construct a multi-dimensional confidence interval system based on the classification of rail transit engineering elements, including: Material price range: ,in, This represents the average unit price of the sample materials. The standard deviation of the unit price of the sample materials. This is the material unit price amplification factor; Labor cost range: ,in, This represents the average unit price of labor costs. This represents the standard deviation of the unit price of labor costs. This is the multiplier for labor costs; The comprehensive cost range is expressed as follows: ,in, This is the average of the overall construction costs. For the comprehensive cost standard deviation, This is the engineering complexity coefficient, with a value range of [value missing]. ; S24: Establish a dynamic interval prediction model for multi-dimensional confidence intervals: Predicted value calculation: ; in, As a time decay weight; For viewing; Interval half-width calculation: ; in, This is the error amplification factor; Output prediction interval: S25: When the current indicator exceeds the range, generate a first instruction signal containing the deviation code; S26: Receive the feature verification results fed back by S3 and dynamically adjust the confidence interval.

[0025] After inputting the cost estimate document, a multi-level inspection mechanism is used to identify defects in the document. The first step is to identify cost deviations based on indicator-based early warning. This can be simply understood as judging the reasonableness of the project type and unit cost recorded in the cost estimate document. The basis for judging the reasonableness of the unit cost is to identify similar historical samples and construct a unit cost range based on the unit costs recorded in these samples. Then, it is determined whether the unit cost recorded in this cost estimate document falls within this range. If it does, it is marked as "reasonable." However, due to the historical database containing projects from different years and inflation, unit cost documents can deviate significantly. In this case, the confidence interval constructed using similar historical samples cannot respond to market fluctuations, making the unit cost range difficult to apply directly. Therefore, in this embodiment, when performing indicator-based early warning, similar historical samples are first screened based on project type, price, and year characteristics. Then, the cost range is classified according to dimensions into: material price range, labor cost range, and comprehensive cost range. Taking the comprehensive cost range as an example: For example, the project type shown in this cost estimate document is "land..." Given the project is a "tunnel," constructed in 2024, and located in East China, the following screening criteria are set: Project Type = Underground Tunnel + Region = East China + Year ∈ [2020, 2024]. After screening, the matching results are the unit price data of 50 historical project samples, as follows: {1820, 1950, ..., 3200, 4500}. The outlier sample "450" ​​is then removed using interquartile range. The mean and standard deviation of the remaining samples are calculated, yielding a confidence interval of [1630, 3000]. This indicates that the unit price in this cost document is correct. If the price is "2800 yuan / m³", then the unit cost recorded in this cost document falls within the confidence interval and is marked as reasonable. If the unit cost recorded in this cost document is 3200 yuan / m³, then after comparison, it does not fall within the calculated confidence interval. At this time, the first instruction signal is generated. After feature inspection, it is found that the project uses a new type of fireproof coating. The confidence interval is then adjusted. After adjustment, the confidence interval is [1870, 3450]. Then, after comparison again, the unit cost in this cost document falls within the above-adjusted confidence interval and is marked as reasonable. Furthermore, the material prices and labor costs shown in the cost estimates are verified using the corresponding material price ranges or labor cost ranges to determine if there are any cost discrepancies. Specifically, the material unit price magnification factor is used in the material prices. All figures were obtained from historical data regression analysis; the labor cost amplification factor in labor costs... It is also obtained from historical data regression analysis; for the engineering complexity coefficient It is also derived from regression analysis of historical data; Understandably, the multi-dimensional confidence interval system constructed based on historical data can detect cost deviations in multiple data points included in the cost estimate. However, this multi-dimensional execution interval system can only indicate whether current project indicators fall within the confidence interval—that is, a static confidence interval obtained from historical data. To determine whether these project indicators pose a risk in the future, a dynamic interval prediction model is needed for further identification. Exemplary: Taking the construction of power supply equipment as an example, based on historical time-series data: time 2023Q1 2023Q2 2023Q3 2023Q4 Unit price / yuan 2750 2810 2920 2850 Based on the historical time-series data mentioned above, the dynamic range of unit construction cost for the next 9 months is calculated as follows: First, calculate the weighted average predicted value. ; ; The weighting rule is that the most recent quarter has the highest weight, set to 1.0; the weighting formula is: ; in, Indicates the most recent quarter; According to calculations: ; Next, calculate the mean absolute error: Based on the actual value of 2750 in Q1 2023, the forecast for the next period (Q2 2023) is as follows: Predicted 2023Q2 = 2750 * 1 / 1 = 2750 The actual value is 2810, so the absolute error is... ; Predicted 2023Q3: 2810 * 1.0 + 2750 * 0.8 / 1.0 + 0.8 = 2810 + 2200 / 1.8 = 2783.3 The actual value is 2920, so the absolute error is... ; Predicting Q4 2023: 2920 * 1.0 + 2810 * 0.8 + 2750 * 0.64 / (1.0 + 0.8 + 0.64) = 6928 / 2.44 = 2839.3 The actual value is 2850, so the absolute error is... ; but: ; Since there is no forecast for Q1 2023, the above only uses forecasts for three quarters. Then, the interval half-width is calculated: Set the basic error amplification factor ; but: ; The output prediction interval is then: ; Based on the output prediction interval, it can be defined as a dynamic confidence interval for a future period. Using historical data as a basis, it predicts the fluctuations in the future period, thereby determining whether there is a risk of exceeding the limits for the engineering indicators in the cost document in the future period. Even if the engineering indicators in the current engineering cost document can fall within the corresponding comprehensive cost interval, if the project cost exceeds the prediction interval, it indicates that there is a risk of exceeding the limit in the future. Therefore, in this embodiment, the multi-dimensional confidence interval system as described above can be used to detect whether the current engineering indicators are reasonable. Combined with the dynamic prediction interval, it can determine whether there is a risk for the engineering indicators in the future period. Thus, by using the multi-dimensional confidence interval system and the dynamic prediction interval, a dynamic and static confidence interval is formed, which can judge the reasonableness of the current engineering indicators and provide early warning of future exceedances, so as to achieve preventive intervention. For long-cycle equipment procurement scenarios, the static confidence interval cannot reflect the future situation, so it may cause the risk of overspending. The prediction interval can make up for this risk. Among them, time decay weight The setting rules are as described in the above embodiments: the closer the observation window, the greater the time decay weight. The larger the value of , the better, based on the formula ,in The value is set based on the number of observation windows, and the nearest observation window is specified as [value]. In this way, the weight of the nearest observation window This is inevitable. Based on this, after acquiring historical data, it's possible to sort it by time, assigning the lowest weight to the oldest observation window and the highest weight (1) to the most recent observation window; other observation windows are then sorted by time, increasing or decreasing sequentially. This strengthens the influence of short-term trends and weakens long-term fluctuations. Secondly, regarding the error amplification factor... This is used to cover the uncertainties of rail transit engineering. The value selection rules depend on the type of project. For the construction of power supply equipment for rail engineering mentioned in the above embodiments, it is necessary to set... To cover 95% of the fluctuations, and The larger the value, the greater the coverage of the new interval obtained through comprehensive calculation; however, excessively large values... This can also lead to an overly conservative range, resulting in an increased false alarm rate. Therefore, based on historical sample backtesting, a value of 1.5 can maintain a range coverage rate of over 90% while reducing the false alarm rate, achieving the optimal balance between pick-up coverage and false alarm rate.

[0026] Based on the above, by identifying similar historical samples based on features, calculating the unit cost in the similar historical samples, obtaining a confidence interval, and then comparing it with the unit cost in the current cost document, it is determined whether the unit cost in the current cost document falls within the confidence interval. If it does, it is marked as reasonable; otherwise, a first instruction signal is generated. The confidence interval is then adjusted based on the results of feature checks. The comparison is performed again based on the adjusted confidence interval. If it falls within the confidence interval, it is marked as reasonable; otherwise, it is unreasonable. Based on this, the false alarm frequency caused by static models or manual checks is reduced, making the cost document index check faster and the error frequency lower.

[0027] In one embodiment, the feature inspection method in step S3 includes the following steps: S31: Construct a standard feature configuration library, defining the mandatory fields, data types, and format rules for feature items; S32: Verify feature values ​​using enumeration matching and regular expressions; S33: Generate semantic correction suggestions for missing items; S34: In response to the first instruction signal, prioritize verifying the matching of the project type; S35: Encode the feature verification result into a feature feedback signal and send it to S2.

[0028] After an indicator health check and early warning, feature checks are performed. However, situations may arise where cross-regional engineering projects and conflicting regional regulations make it difficult to identify feature items in cost documents. For example, a cross-provincial subway project may involve Province A and Province B. Province A requires a seismic resistance level of 7, while Province B requires 8, leading to confusion of feature items across different sections of the same project. In this embodiment, after the cost document is input, the project type, region, seismic resistance level, and other feature items are identified first. Then, feature item resources corresponding to different sections of the same project are integrated to build a dynamic feature library. As mentioned above, local regulations from Province A, new regulations from Province B, and national standards (GB) are input into the feature library to build a dynamic feature library. Then, intelligent feature verification is performed. For example, when performing regional sensitivity verification, regular expressions are used to verify seismic resistance, etc. The verification output result is level 7, while the dynamic feature library records that the new regulations of Province B require a seismic resistance level of level 8. Therefore, the regional sensitivity verification fails. Then, the timeliness verification is performed. For example, the policy is automatically associated with the time. If the current time is January 1, 2024, and according to the new concrete regulations, the concrete grade recorded in the cost document is C30, which does not conform to the [C35, C36] recorded in the new concrete regulations, then the detection result also fails. Then, the contradiction feature detection is performed. By constructing a construction method verification rule library, for example: shield tunneling method: [segment support, grouting support], while open cut method: [sheet pile support, soil nailing wall support]. After inspection, the construction method feature in this cost document is shield tunneling method, but the support type is open cut support. Therefore, it does not match the segment support and grouting support recorded in the construction method verification rule library and is marked as contradictory. Based on the above, structured repair instructions are generated, such as: 1. [Section B] Seismic resistance grade incorrect. Current value: Level 7 Required value: Level 8 (according to Clause 5.2.1 of DB42 / T2023-2023) Correction: Automatically replace with level 8 2. [Throughout the entire line] Concrete grade mismatch. Current value: C30 The new national standard requires: C35 (GB 55006-2024, Clause 3.1.3) Impact Analysis: The unit price needs to be increased by ¥120 / m³ 3. [Contradiction between construction methods] Shield tunneling method and open-cut support method are contradictory. Recommended solution: Option 1: Modify the construction method to open-cut method Option 2: Modify the support structure to segment support. Finally, if the indicator health check outputs the first instruction signal, it is necessary to prioritize verifying whether the project type is accurate and checking whether the concrete grade has been updated. For example, if the first instruction signal shows that the unit price of ventilation engineering exceeds the limit by 15%, then verify whether the project type is accurate, excluding ground buildings, and then check whether the concrete grade has been updated. For example, the new national standard uses C35 concrete, which will lead to an increase in cost. Based on the verification, it was found that because the historical sample used the old standard, the concrete grade was C30, which lowered the unit cost. However, this cost document uses C35 concrete because the new regulations, which leads to an increase in the unit cost and exceeds the limit. At this time, the root cause is: the unit cost is distorted because the concrete grade has not been updated. Based on the above, this embodiment uses enumeration matching and regular expression to verify feature values, which can solve the problem of difficulty in implementing cross-provincial standard differences. Secondly, it can respond to false alarms caused by policy changes in a timely manner. In addition, based on signal response, it can locate feature blind spots when the indicator is abnormal, and the root cause can be located faster.

[0029] In one embodiment, the valid data check in step S4 includes the following steps: S41: Use a tree-based recursive traversal algorithm to scan multi-level nodes in the cost breakdown file; S42: Use a dynamic rule engine to validate key fields and determine whether the key field data is invalid; S43: When the data is invalid, generate a second instruction signal containing the node path; The dynamic rule engine has a built-in validation rule library for validating key fields.

[0030] When evaluating cost estimates, it is also necessary to identify missing items in the cost estimates. In the traditional method based on manual verification, there are problems such as data omissions, policy disconnects, and inability to identify data chain reactions, which lead to data errors in the cost estimates and cause overall data distortion. In this embodiment, when identifying cost estimates, a tree recursive traversal algorithm is used to scan the multi-level nodes in the sub-item data. Taking a tunnel project as an example, the sub-item is a shield section, and the multi-level nodes include segment installation and grouting. Segment installation includes reinforced concrete segments, and more specifically, steel bars and coolant. For example, it is assumed that the steel bar quantity is recorded as 1200t and the coolant quantity is recorded as 85m³. Grouting includes synchronous grouting, but the units are not identified. Based on a tree-based recursive traversal algorithm, the system then calls a dynamic rule engine to perform verification by loading a validation rule library. First, if a grouting unit is missing, the node is marked as an abnormal node. During the valid data check of the cost document, the dynamic rule engine and tree-based recursive traversal algorithm enable rapid scanning, identification, and validation of key fields. Specifically, the quantity must be numerical and greater than zero, the unit field must be non-empty, and the unit price field must be greater than zero. If a key field in the cost document does not meet these requirements, it is defined as an abnormal node. Furthermore, when the dynamic rule engine calls the validation rule library, if it identifies that the construction method is shield tunneling and the material is [reinforcing steel, concrete] instead of the cryogenic fluid recorded in the cost document, then the shield tunneling method and cryogenic fluid combination is incorrect. When invalid data is identified, a second instruction signal is generated and transmitted to the specification check, calling the specification library for verification, and outputting the verification result. For example, if the shield tunneling method and cryogenic fluid combination is missing and does not conform to the specification, it is also defined as an abnormal node.

[0031] In one embodiment, the specification check in S5 includes the following steps: S51: Built-in specification library, which includes bill of quantities specifications and tax calculation rules; S52: Respond to the second command signal and locate the abnormal node path; S53: Related policy provisions and explanations; S54: Generate specification amendment instructions based on policy provisions and push them to the dynamic rule engine of S4 to update the verification rule base.

[0032] Based on the above, cost documents may involve cross-regional projects, which can lead to inconsistent specifications. Since the verification rule base in the dynamic rule engine may not be updated in a timely manner, using the specifications of any region for key field verification can result in distorted inspection results and may lead to false alarms. In this embodiment, during the evaluation of cost documents using a multi-level inspection mechanism, if a second instruction signal is generated during valid data inspection, it indicates that invalid data exists in the cost document. For complex data types in cost documents, the invalid data mentioned here needs to be analyzed in conjunction with the contextual semantics. For example, when the dynamic rule engine calls the verification rule base for verification, it identifies that when the construction method is shield tunneling, the material is [steel reinforcement, concrete], instead of the coolant recorded in the cost document. Therefore, the shield tunneling method and coolant are determined to be an incorrect combination according to the content recorded in the verification rule base, and the coolant is invalid data. At this time, it is necessary to call the specification base to verify whether there is a new specification that can verify the use of shield tunneling method with coolant. Furthermore, let's take an inter-provincial tunnel project as an example to illustrate: Suppose a cross-provincial tunnel project involves Province A and Province B. Province A's local regulations require a seismic resistance level of 7, while Province B has issued a new local regulation mandating a seismic resistance level of 8. Using traditional manual review processes, the different local regulations of the two provinces are easily overlooked, leading to underestimation or omission of project costs and impacting later construction progress. In this embodiment, after the cost estimate document is input, a multi-level checking mechanism identifies multiple layers of data in the cost estimate document, for example: Section: Section B; Items: Code: 010302005; Name: Tunnel concrete lining; Seismic resistance rating: 7; Concrete grade: C35; Material unit price: 520 yuan / m³; Because the seismic resistance level in the cost document does not match the seismic resistance level stipulated in the new local standard of Province B, a second instruction signal will be generated during the valid data check. The specification check mechanism will then interpret the reason for the anomaly by associating it with the policy clauses. In this embodiment, after the specification check receives the second instruction signal, it first defines the abnormal node path and automatically identifies the location of the contract section based on the abnormal node path. In this embodiment, it is the B contract section. Then, it checks whether the new local standard document exists for the B contract section. According to the record in the new local standard document, the seismic resistance level requirement is level 8. Then, a specification correction instruction is generated according to the specific clauses in the new local standard and sent to the dynamic rule engine of S4 to update the verification rules. Furthermore, the specific process for the specification library and validation rule library to handle the above scenarios is as follows: Step 1: The rule base updates the basic rule "Seismic resistance level ≥ 8" through S61, and then performs a valid data check on the cost file. A level 7 anomaly is detected, but the reason cannot be explained. Step 2: S4 generates a second command signal, carrying the node path: [Section B / Structural Engineering / Seismic Resistance Grade]; Step 3: Check the S52 positioning node and the new S53 policy and explanation: New regulations in Province B include: Article 5.2.1 of DB42 / T2023-2023 (Mandatory Level 8). Secondly, according to the geological survey report, the peak ground acceleration in this area is >0.3g; Step 4: S54 generates differentiated solutions: The new regulations mandate an upgrade to Level 8 (increasing costs by 12 million). Step 5: Push the specification modification command to the rule base: If the project location is Province B and the policy version is greater than 2023: Add a correlation check for "matching between seismic resistance level and geological fortification intensity"; Based on the above, it is understandable that when checking the effective data of the cost documents, the results of the effective data check are distorted due to the problem of lagging updates of local regulations. Therefore, when invalid data is identified, a second instruction signal can be generated and transmitted to the regulation check. The regulation check is used to realize feedback and updates, and to realize real-time reverse driving of the dynamic rule engine to update the verification rule base based on invalid data.

[0033] In one embodiment, the verification rule base is executed as follows: S61: Update rules in response to new policies; S62: Receives the protocol correction command from S54 and updates the verification rule base in real time.

[0034] It is understood that in this embodiment, the validation rule base is used to verify the validity of key fields in the cost estimate document. It is also understood that the validation rule base can be used to validate key fields, such as significant features, which need to meet the following requirements: The quantities are numerical and greater than zero; The unit field is not empty and exists in the unit dictionary; Unit price is greater than zero; For erroneous data, such as cross-provincial tunnel projects where the seismic resistance requirements of Province A and Province B are different, if a seismic resistance level of 7 is uniformly entered for both Section A and Section B, then for Section B, the location file of the region described in Section B will show a mismatch when the verification rule base retrieves the location file. To address the above, the validation rule base needs to update its rules in response to new policies, forcibly overriding old rules. Furthermore, for cross-provincial projects, the S54 specification amendment instructions can be used to automatically adapt to local regulations by associating the project location. While there is a connection between the validation rule base and the specification library, there are also differences. It's understandable that the rule base is used to identify explicit conditions; the validation rule base intelligently judges whether a value exists and is correct, but it cannot analyze the cause of abnormal values. The purpose of setting up specification checks is to analyze policies and optimize the validation rule base and update validation rules in a timely manner, such as: "When a project adopts a prefabricated structure, the seismic resistance level is automatically increased by 1 level." The validation rule base is used to reverse-verify and update its own rules, resolving blind spots in semantic understanding and conflict determination, and enabling root cause analysis and solution output for errors reported by the validation rule base.

[0035] In one embodiment, a data security processing step is also included: S71: Homomorphically encrypt sensitive fields, including contract amount and supplier information; S72: Generate a desensitization assessment report for authorized users to view the complete data.

[0036] When cost estimates are reviewed on a third-party review platform, they are prone to being transmitted without proper anonymization, leading to the extraction of supplier lists and potential losses. To address this, this embodiment employs homomorphic encryption for end-to-end encrypted calculation during the cost estimate review and evaluation process. This prevents data leakage and generates an anonymized evaluation report, which authorized users can access when needed, thus preventing information leakage and potential losses.

[0037] In one embodiment, the historical sample data is fused from multiple sources, including publicly available government databases, private enterprise databases, and building materials platform data; The multi-source historical sample data is verified for cross-source data consistency through blockchain, and the confidence weight factor for S21 sample screening is calculated based on the verification results.

[0038] In the process of cost assessment for rail transit projects, it is foreseeable that the quality of historical sample data directly determines the accuracy of indicator health check and early warning. However, the quality of historical sample data is subject to problems such as mixed sources and version conflicts. In this embodiment, based on historical sample data from multiple sources, specifically including government, enterprise and platform data, and combined with blockchain to verify data consistency, a three-in-one trusted data pool can be constructed to avoid the above defects. In addition, the basis for setting the confidence weight factor is set: if the data is verified by blockchain, it is counted as 1.2; otherwise, it is counted as 0.5. Confidence weighting factor = Blockchain verification score × Time decay coefficient × Location feature matching degree; Among these methods, combining the time-related decay coefficient with the confidence weight factor can automatically reduce the weight of old samples. The time-related decay coefficient is expressed as follows: ; Example: Taking three sets of historical sample data from the government, enterprises, and platforms respectively as examples: The corresponding confidence weighting factor is: Government database: 1.2 × 0.92 × 0.95 = 1.05 Company A: 0.5 × 0.85 × 0.88 = 0.37 Based on the above, it is stipulated that when the weight is less than 0.6, the corresponding sample is excluded. Therefore, in the above example, only the government database sample is retained. Calculation shows that: mean ; Standard deviation ; The confidence interval is expressed as [320, 320]; Based on the above, sample contamination that could cause abnormal confidence intervals and lead to false alarms in health checkups can be prevented. Furthermore, by calculating confidence weighting factors, false, outdated, and irrelevant samples from the selected historical samples are eliminated. Simultaneously, the calculation ensures the mean... with standard deviation Reflecting a true and reliable data distribution, the accuracy of the final output confidence interval depends on the quality control of the confidence weight factor. Based on the calculation of the confidence weight factor in this embodiment, the confidence interval forms a closed-loop process from data purification to statistical decision-making, thereby reducing misjudgments caused by defects in the quality of historical samples.

[0039] In one embodiment, the evaluation report generation step in S6 includes: S61: Build multi-dimensional visualization dashboards, including: The index deviation radar chart is used to display the S2 results; The collaborative path flow diagram is used to show the linkage relationship between S2 and S3 and between S4 and S5; A data error distribution tree diagram is used to highlight the S4 abnormal node.

[0040] In one embodiment, the data error distribution tree diagram is further used to integrate three-dimensional analysis coordinate axes, including: X-axis positioning: Determine the spatial criticality level of the abnormal node based on its hierarchical depth in the sub-item cost document tree and the proportion of nodes within its influence range; Y-axis tracing: Based on blockchain evidence storage, historical modification records are extracted, and the time anomaly index of the node is calculated by comparing the deviation of the current value from the historical fluctuation trend. Z-axis assessment: Based on the matching degree between policy provisions and node semantics, combined with the policy timeliness and violation consequences database, a compliance impact rating is generated; When a node’s spatial criticality level, time anomaly index, and compliance impact rating all exceed the industry benchmark threshold, the compliance check process will be automatically executed.

[0041] In the cost assessment process of rail transit projects, traditional assessment methods can only display information such as abnormal indicators, incorrect features, and invalid data. Manual analysis is required to determine if there are correlations between various defects, leading to fragmented defects that are difficult to locate and establish connections. In this embodiment: In the X-axis, the criticality level of a node is determined based on the hierarchical depth and the proportion of nodes in the abnormal node. The root node in the data tree is defined as level 1, and the depth increases by 1 for each additional level. The proportion of the impact range is calculated by the number of child nodes affected by the abnormal node / the total number of nodes. The criticality level includes level 1, level 2, level 3, and level 4. The above is the existing technology and can be referred to in the "Data Tree Structure Specification at Home" GB / T51269-2017. In the Y-axis, the time anomaly index is calculated based on the degree of deviation, specifically obtained as |current value - historical mean| / historical standard deviation. A deviation greater than [a certain value] is defined as [a certain value]. At that time, it was considered a severe abnormality; in the state of... When the value is between 0 and 1, it is considered moderately abnormal; when it is less than 1, it is considered moderately At that time, it is considered a mild anomaly. The above is existing technology and can be referenced in the "Standard for Identifying Price Fluctuations" JT / T1320-2020. In the Z-axis, compliance impact rating requires obtaining semantic matching degree, timeliness, and consequences of violations. When all three are high, the rating is Level 1 impact; when two are high, it is Level 2 impact; and when one is high, it is Level 3 impact. For the above, please refer to the "Guidelines for Prevention and Control of Major Policy Risks" Jianbiao

[2023] No. 18. The logic for triggering a compliance check is as follows: when the spatial criticality is greater than level two, the time anomaly index is greater than moderate anomaly, and the compliance impact rating is greater than level two impact, then a compliance check is triggered. Based on three-dimensional analysis coordinates, a three-dimensional diagnostic process of space-time-policy is established by forcibly associating multi-source data. Among them, based on the correlation between the blockchain history on the Y-axis and the policy on the Z-axis, spatiotemporal policy penetration analysis is realized. When a high-impact anomaly is detected, a specification check is automatically triggered to reduce the loop of missed detections and feedback.

[0042] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely illustrative of the principles of the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the present invention as claimed. The scope of protection of the present invention is defined by the appended claims and their equivalents.

Claims

1. A big data-based intelligent cost assessment method for rail transit engineering, characterized in that: Includes the following steps: S1: Input the cost documents for the project to be evaluated, including sub-items, bill of quantities, quotas, and information on labor, materials, and machinery; S2: Execute health check alerts based on performance indicators, match historical sample data, remove outliers using quartiles, and calculate confidence intervals. If the unit cost exceeds the confidence interval, it is determined that there is a cost deviation; When S2 detects a cost deviation, it sends a first instruction signal to S3 to trigger a project type matching verification. S3: Perform feature checks to verify the completeness and compliance of engineering features, and encode the verification results as feedback signals to send to S2. When S2 receives a feedback signal, it dynamically adjusts the confidence interval. S4: Perform valid data checks, scan the validity of key fields based on the validation rule base, and if invalid data is identified, send a second instruction model carrying the node path to S5; S5: Responds to the second instruction model, locates the node path and associates with policy clauses, generates a specification amendment instruction and pushes it to S4; S6: Generate a comprehensive assessment report, summarizing the inspection results and corrective suggestions.

2. The intelligent cost assessment method for rail transit engineering based on big data as described in claim 1, characterized in that: The method for early warning of health indicators in S2 includes the following steps: S21: Select similar historical samples based on project type, region, and year characteristics, and verify the consistency of data sources through blockchain to generate reliable sample labels; S22: Use quartiles to remove outliers and calculate the mean of the remaining samples. and standard deviation ; S23: Construct a multi-dimensional confidence interval system based on the classification of rail transit engineering elements, including: Material price range: ; Labor cost range: ; The comprehensive cost range is expressed as follows: ; S24: Establish a dynamic interval prediction model for multi-dimensional confidence intervals: Predicted value calculation: ; in, As a time decay weight; For viewing; Interval half-width calculation: ; in, This is the error amplification factor; Output prediction interval: ; S25: When the current indicator exceeds the range, generate a first instruction signal containing the deviation code; S26: Receive the feature verification results fed back by S3 and dynamically adjust the confidence interval.

3. The intelligent cost assessment method for rail transit engineering based on big data as described in claim 1, characterized in that: In step S3, the feature inspection method includes the following steps: S31: Construct a standard feature configuration library, defining the mandatory fields, data types, and format rules for feature items; S32: Verify feature values ​​using enumeration matching and regular expressions; S33: Generate semantic correction suggestions for missing items; S34: In response to the first instruction signal, prioritize verifying the matching of the project type; S35: Encode the feature verification result into a feature feedback signal and send it to S2.

4. The intelligent cost assessment method for rail transit engineering based on big data as described in claim 1, characterized in that: In step S4, the valid data check includes the following steps: S41: Use a tree-based recursive traversal algorithm to scan multi-level nodes in the cost breakdown file; S42: Use a dynamic rule engine to validate key fields and determine whether the key field data is invalid; S43: When the data is invalid, generate a second instruction signal containing the node path; The dynamic rule engine has a built-in validation rule library for validating key fields.

5. The intelligent cost assessment method for rail transit engineering based on big data according to claim 1, characterized in that: In S5, the specification check includes the following steps: S51: Built-in specification library, which includes bill of quantities specifications and tax calculation rules; S52: Respond to the second command signal and locate the abnormal node path; S53: Related policy provisions and explanations; S54: Generate specification amendment instructions based on policy provisions and push them to the dynamic rule engine of S4 to update the verification rule base.

6. The intelligent cost assessment method for rail transit engineering based on big data according to claim 4, characterized in that: The verification rule base is executed as follows: In response to new policies, rules have been updated. Receive S54 protocol correction instructions and update the verification rule base in real time.

7. The intelligent cost assessment method for rail transit engineering based on big data according to claim 1, characterized in that: It also includes data security processing steps: S71: Homomorphically encrypt sensitive fields, including contract amount and supplier information; S72: Generate a desensitization assessment report for authorized users to view the complete data.

8. The intelligent cost assessment method for rail transit engineering based on big data according to claim 1, characterized in that: The historical sample data is integrated from multiple sources, including publicly available government databases, private enterprise databases, and building materials platform data. The historical sample data is verified for cross-source data consistency through blockchain, and the confidence weight factor for S21 sample screening is calculated based on the verification results.

9. The intelligent cost assessment method for rail transit engineering based on big data according to claim 1, characterized in that: In step S6, the evaluation report generation step includes: S61: Build multi-dimensional visualization dashboards, including: The index deviation radar chart is used to display the S2 results; The collaborative path flow diagram is used to show the linkage relationship between S2 and S3 and between S4 and S5; A data error distribution tree diagram is used to highlight the S4 abnormal node.

10. The intelligent cost assessment method for rail transit engineering based on big data according to claim 9, characterized in that: The data error distribution tree diagram is also used to integrate three-dimensional analysis coordinate axes, including: X-axis positioning: Determine the spatial criticality level of the abnormal node based on its hierarchical depth in the sub-item cost document tree and the proportion of nodes within its influence range; Y-axis tracing: Based on blockchain evidence storage, historical modification records are extracted, and the time anomaly index of the node is calculated by comparing the deviation of the current value from the historical fluctuation trend. Z-axis assessment: Based on the matching degree between policy provisions and node semantics, combined with the policy timeliness and violation consequences database, a compliance impact rating is generated; When a node’s spatial criticality level, time anomaly index, and compliance impact rating all exceed the industry benchmark threshold, the compliance check process will be automatically executed.

Images