Machine learning method based on engineering project contract data

Abstract

The invention relates to the technical field of machine learning, and particularly discloses an engineering project contract data-based machine learning method, which comprises the following steps of S1, collecting contract data of historical engineering projects, including factor values of time nodes and project progress; calculating the influence sensitivity and the sensitive influence value of each factor, and distinguishing a high-sensitivity factor and a low-sensitivity factor; s2, setting a monitoring interval, and collecting each factor value and project progress data in real time; dynamically correcting the sensitive influence value of the high-sensitivity factor according to the real-time data, generating a correction parameter, and ensuring that the model adapts to the latest project state; s3, predicting a time interval of project completion in combination with the correction parameter of the high-sensitivity factor and the fixed sensitivity value of the low-sensitivity factor; calculating the probability value of the reasonable long account age of the project based on the comparison of the prediction time and the actual plan time; according to the method, key factor parameters are dynamically adjusted, so that the accuracy of engineering project progress prediction is improved.

Description

Technical Field

[0001] This invention relates to the field of machine learning technology, and more specifically to a machine learning method based on engineering project contract data. Background Technology

[0002] In engineering project management, the analysis of contract data and schedule forecasting are crucial for ensuring timely project completion. Traditional methods rely primarily on human experience or static models for schedule assessment, but these methods still have limitations. For example, they cannot dynamically respond to changes in external factors, nor can they distinguish the differentiated impacts of highly sensitive or insensitive factors on schedule, and important factors may be underestimated. To address this issue, some existing research attempts to optimize schedule management through regression analysis or BIM technology, but problems such as poor real-time performance and lack of correction mechanisms still exist. For instance, while BIM-based schedule simulation can visualize the project status, it cannot automatically correlate the impact of dynamic factors (such as temperature and precipitation). Although machine learning methods have been introduced, they mostly focus on training with single historical data and lack the ability to adaptively adjust to real-time monitoring data. At the same time, a reasonably long payment aging period can disrupt the original payment rhythm, leading to cash flow imbalance, increased costs, and increased risks in contract settlement and accounting. Therefore, if there is a risk of payment delays in the project progress, it is necessary to issue early warnings and adjust the funding plan accordingly. Therefore, there is an urgent need for a machine learning method that can dynamically integrate historical and real-time data, quantify factor sensitivity, and automatically correct predictions to improve the accuracy and timeliness of early warning of aging risks in engineering projects. Summary of the Invention

[0003] The purpose of this invention is to provide a machine learning method based on engineering project contract data to solve the following technical problems.

[0004] The objective of this invention can be achieved through the following technical solutions: A machine learning method based on engineering project contract data includes the following steps: Step S1: Set several factors and obtain contract data for all historical engineering projects to obtain historical engineering project data. The historical engineering project data includes the factor values ​​of each factor at each time node within the historical engineering project cycle and the progress of the historical engineering project. Based on the historical engineering project data, the sensitivity and sensitivity values ​​of each factor are obtained, and high-sensitivity factors and low-sensitivity factors are determined based on the sensitivity of each factor. Step S2: Set the monitoring interval for the project, and monitor the factor values ​​of each factor and the project progress at each time node in real time within the monitoring interval to obtain real-time monitoring data; based on the real-time monitoring data, correct the sensitivity impact value of the highly sensitive factor to obtain the correction parameter of the highly sensitive factor. Step S3: Based on the correction parameters of the high-sensitivity factors and the influence sensitivity values ​​of the low-sensitivity factors, obtain the predicted completion period interval of the project; obtain the predicted completion time, and based on the predicted completion time and the predicted completion period interval, obtain the reasonable long aging probability value of the project.

[0005] As a further aspect of the present invention: the process of obtaining the historical engineering project data includes: The start time and completion time of the historical project are obtained. The time period consisting of the start time and completion time is recorded as the historical project cycle. A time interval threshold is set, and several time nodes are selected by equally dividing the historical project cycle according to the time interval threshold.

[0006] As a further aspect of the present invention: the process of obtaining the sensitivity of the influence of the aforementioned factor includes: Based on the historical project progress at each time node in the historical project data, several data points are obtained. Each data point is fitted to generate a historical progress change curve. The slope of the tangent line at each data point is obtained on the historical progress change curve and recorded as the progress change rate at the corresponding time node. For any factor, all other factors besides the stated factor are denoted as "remaining factors," and the factor values ​​of these remaining factors are denoted as "remaining factor values." In the historical project data, the factor values ​​of the stated factor are selected when all remaining factor values ​​are consistent, and the progress change rate corresponding to each factor value at this time is obtained, resulting in a data set. Linear regression fitting is performed on the data set based on the least squares method to obtain a regression line. The slope of the regression line is obtained and denoted as the sensitive influence value, and the absolute value of the slope of the regression line is obtained and denoted as the influence sensitivity of the stated factor.

[0007] As a further aspect of the present invention: the process for determining the highly sensitive factors and the low-sensitivity factors includes: Obtain the sensitivity of each factor's influence {Aff1, Aff2, ..., Aff} n}, where Aff n This represents the sensitivity of the nth factor, where n represents the total number of factors; a multiple threshold is set to obtain the sensitivity threshold. Aff i This represents the sensitivity of the influence of the i-th factor, where i ∈ [1, n] and i is a positive integer. l The threshold is a multiple; factors whose influence sensitivity exceeds the sensitivity threshold are categorized as high-sensitivity factors, and factors whose influence sensitivity is less than or equal to the sensitivity threshold are categorized as low-sensitivity factors.

[0008] As a further aspect of the present invention: the process of obtaining the historical engineering project data further includes: An engineering simulation model is established based on BIM technology. A three-dimensional model of the historical project at the time of completion is established through the engineering simulation model and denoted as the as-built three-dimensional model. The three-dimensional model of the historical project at the time node is also obtained and denoted as the real-time three-dimensional model. The model areas of the as-built three-dimensional model and the real-time three-dimensional model are obtained respectively and denoted as the completed area and the real-time area. The progress of the historical project at the time node is obtained as Poh = Fs / Ra × 100%, where Fs is the completed area and Ra is the real-time area.

[0009] As a further aspect of the present invention, the process of correcting the sensitivity influence value of the highly sensitive factor includes: Based on the real-time monitoring data, the factor values ​​of all low-sensitivity factors at each time point within the monitoring interval are obtained. Based on the sensitivity impact value of each low-sensitivity factor, the low-sensitivity interference value at each time point is obtained. K r Eff represents the sensitivity value of the r-th low-sensitivity factor. r Let r represent the value of the r-th low-sensitivity factor, where r ∈ [1, m] and r is a positive integer, and m is the total number of low-sensitivity factors; Based on the real-time monitoring data, the project progress at each time node is obtained, several real-time data points are obtained, a real-time progress change curve is generated, and the slope of the tangent at each real-time data point on the real-time progress change curve is obtained, which is recorded as the real-time progress change rate at the time node. Based on the low-sensitivity interference values ​​at each time node, the correction progress change rate Rps = Ps + Lsi is obtained at each time node within the monitoring interval, where Ps is the real-time progress change rate at the time node. The factor values ​​and correction progress change rates of high-sensitivity factors at each time node are obtained, and the sensitivity impact values ​​of these high-sensitivity factors are recorded as high-impact sensitivity values. Correction parameters for each high-impact sensitivity value are obtained based on a genetic algorithm, such that... Reaching the minimum value, where oh x Kh represents the x-th high-influence sensitivity correction parameter. x His represents the high impact sensitivity value of the x-th highly sensitive factor. x Rps represents the value of the x-th highly sensitive factor, t represents the t-th time point, and Rps represents the value of the x-th highly sensitive factor. t This represents the rate of change of the correction progress at the t-th time node, where T is the total number of time nodes within the monitoring interval, and num is the total number of high-impact sensitivity values.

[0010] As a further aspect of the present invention: the process of obtaining the predicted completion period interval of the engineering project includes: The end time of the monitoring interval is recorded as the start time. The project progress of the project at the start time is obtained and recorded as the start progress. Each time node after the start time is obtained and recorded as the prediction node. The factor value of each factor at each prediction node is obtained and recorded as the prediction factor value. The time node corresponding to the starting time is denoted as t0, and each prediction node is numbered and denoted as t. q , t q This represents the q-th prediction node; it obtains the progress of the first project at the q-th prediction node. , where P tq-1 H represents the project progress at the (q-1)th prediction node. x This represents the predictor value of the x-th highly sensitive factor; and the second project progress at the q-th prediction node is obtained. E r This represents the predictor value of the r-th low-sensitivity factor; Obtain the timestamp corresponding to the prediction node where the progress of the first project first satisfies firP≥100%, denoted as T. fir ; Obtain the timestamp corresponding to the predicted node where the progress of the second project first satisfies thP≥100%, denoted as T. th ; Obtain the predicted completion period interval [T] fir T th ].

[0011] As a further aspect of the present invention: the process of obtaining the reasonable long aging probability value of the project includes: If the predicted completion time is denoted as "time", then the reasonable long aging probability value is obtained. , where W is the preset attenuation coefficient and W>0, and e is the Euler number.

[0012] The beneficial effects of this invention are: This invention distinguishes between highly sensitive and low-sensitive factors and dynamically adjusts the sensitivity impact values ​​of highly sensitive parameters based on real-time monitoring data, enabling the model to adapt to the latest project status and significantly improving the accuracy of schedule prediction. By comparing the predicted completion time interval with the planned time, it calculates a reasonable long aging probability value, providing intuitive quantitative evidence for project management and facilitating the early development of risk response strategies. It utilizes machine learning algorithms (such as least squares and genetic algorithms) to automatically analyze historical data and optimize parameters, reducing manual intervention and improving efficiency. Simultaneously, it achieves 3D schedule visualization through BIM technology, enhancing data reliability. By differentiating critical and non-critical factors through sensitivity thresholds, it helps managers focus on highly sensitive factors (such as weather conditions) that significantly impact schedule allocation, thus optimizing resource allocation. Through data-driven and dynamic optimization, this invention achieves scientific and intelligent project schedule management, effectively reducing the risk of reasonable long aging. Attached Figure Description

[0013] The invention will now be further described with reference to the accompanying drawings.

[0014] Figure 1 This is a schematic diagram of the structure of a machine learning method based on engineering project contract data according to the present invention. Detailed Implementation

[0015] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0016] Please see Figure 1 As shown, this invention is a machine learning method based on engineering project contract data, comprising the following steps: Step S1: Set several factors and obtain contract data for all historical engineering projects to obtain historical engineering project data. The historical engineering project data includes the factor values ​​of each factor at each time node within the historical engineering project cycle and the progress of the historical engineering project. Based on the historical engineering project data, the sensitivity and sensitivity values ​​of each factor are obtained, and high-sensitivity factors and low-sensitivity factors are determined based on the sensitivity of each factor. In a preferred embodiment of the present invention, the factors are a number of predictable external factors, including temperature, precipitation, and wind. In a preferred embodiment of the present invention, the process of obtaining the historical engineering project data includes: The start time and completion time of the historical project are obtained. The time period consisting of the start time and completion time is recorded as the historical project cycle. A time interval threshold is set, and several time nodes are selected by equally dividing the historical project cycle according to the time interval threshold. The process of obtaining the historical engineering project data also includes: An engineering simulation model is established based on BIM technology. A three-dimensional model of the historical project at the completion time is established through the engineering simulation model and denoted as the as-built three-dimensional model. The three-dimensional model of the historical project at the time node is also obtained and denoted as the real-time three-dimensional model. The model areas of the as-built three-dimensional model and the real-time three-dimensional model are obtained respectively and denoted as the completed area and the real-time area. The progress of the historical project at the time node is obtained as Poh = Fs / Ra × 100%, where Fs is the completed area and Ra is the real-time area. The factor value is a quantitative value of the factor, and the factor value includes temperature, precipitation and wind level; Specifically, the factors considered in the method of this invention are not limited to weather factors such as temperature, precipitation, and wind speed, but also include a variety of quantifiable engineering-related factors such as material supply stability, human resource allocation, equipment operating status, design change frequency, policy and regulatory changes, construction technical difficulty, on-site safety management level, supply chain reliability, geological condition changes, and contract clause execution. These factor values ​​are obtained from historical contract data at different time points and participate in the calculation and classification of impact sensitivity. The factors can be flexibly selected and set according to the specific project type and project management needs to ensure that the model can comprehensively reflect various variables affecting project progress. In a preferred embodiment of the present invention, the process of obtaining the sensitivity of the influence of the factor includes: Based on the historical project progress at each time node in the historical project data, several data points are obtained. Each data point is fitted to generate a historical progress change curve. The slope of the tangent line at each data point is obtained on the historical progress change curve and recorded as the progress change rate at the corresponding time node. For any factor, all other factors are denoted as "remaining factors," and the factor values ​​of these remaining factors are denoted as "remaining factor values." In the historical project data, the factor values ​​of the factor are selected when all remaining factor values ​​are consistent, and the progress change rate corresponding to each factor value at this time is obtained to obtain a data set. Linear regression fitting is performed on the data set based on the least squares method to obtain a regression line. The slope of the regression line is obtained and denoted as the sensitive influence value. The absolute value of the slope of the regression line is obtained and denoted as the influence sensitivity of the factor. It should be noted that, through BIM technology or contract records, the start and end times of historical projects are obtained, time nodes are divided at fixed intervals, and the factor values ​​and project progress of each node are recorded; historical progress data are fitted to generate change curves, and the progress change rate of each time node is calculated through the slope of the tangent line. By fixing the values ​​of other factors, we select the rate of change of progress corresponding to different values ​​of the target factor (such as temperature) to form a dataset. We then fit the dataset using the least squares method, and the slope of the regression line is the sensitive impact value (positive or negative indicates that progress is promoted or inhibited). The absolute value of the slope is the impact sensitivity (measures the strength of the impact). For example, if a 1 unit increase in temperature leads to a 0.5% acceleration in progress, the sensitive impact value is +0.5, and the sensitivity is 0.5. In a preferred embodiment of the present invention, the process of determining the highly sensitive factors and the low-sensitivity factors includes: Obtain the sensitivity of each factor's influence {Aff1, Aff2, ..., Aff} n}, where Aff n This represents the sensitivity of the nth factor, where n represents the total number of factors; a multiple threshold is set to obtain the sensitivity threshold. Aff i This represents the sensitivity of the influence of the i-th factor, where i ∈ [1, n] and i is a positive integer. l The threshold is a multiple; factors whose influence sensitivity exceeds the sensitivity threshold are categorized as high-sensitivity factors, and factors whose influence sensitivity is less than or equal to the sensitivity threshold are categorized as low-sensitivity factors. Identify key influencing factors from historical project data and quantify their sensitivity to project schedule to lay a data foundation for subsequent dynamic forecasting and risk control; organize contract data from historical projects to extract time nodes, factor values ​​(such as temperature, precipitation, etc.) and corresponding schedules; distinguish between highly sensitive factors (significantly affecting schedule) and low-sensitive factors (with less impact) to help managers focus on key variables; Step S2: Set the monitoring interval for the project, and monitor the factor values ​​of each factor and the project progress at each time node in real time within the monitoring interval to obtain real-time monitoring data; based on the real-time monitoring data, correct the sensitivity impact value of the highly sensitive factor to obtain the correction parameter of the highly sensitive factor. In a preferred embodiment of the present invention, the monitoring interval is a time period of a preset length after the start of the project; In a preferred embodiment of the present invention, the process of correcting the sensitivity influence value of the highly sensitive factor includes: Based on the real-time monitoring data, the factor values ​​of all low-sensitivity factors at each time point within the monitoring interval are obtained. Based on the sensitivity impact value of each low-sensitivity factor, the low-sensitivity interference value at each time point is obtained. K r Eff represents the sensitivity value of the r-th low-sensitivity factor. r Let r represent the value of the r-th low-sensitivity factor, where r ∈ [1, m] and r is a positive integer, and m is the total number of low-sensitivity factors; Based on the real-time monitoring data, the project progress at each time node is obtained, several real-time data points are obtained, a real-time progress change curve is generated, and the slope of the tangent at each real-time data point on the real-time progress change curve is obtained, which is recorded as the real-time progress change rate at the time node. Based on the low-sensitivity interference values ​​at each time node, the correction progress change rate Rps = Ps + Lsi is obtained at each time node within the monitoring interval, where Ps is the real-time progress change rate at the time node. The factor values ​​and correction progress change rates of high-sensitivity factors at each time node are obtained, and the sensitivity impact values ​​of these high-sensitivity factors are recorded as high-impact sensitivity values. Correction parameters for each high-impact sensitivity value are obtained based on a genetic algorithm, such that... Reaching the minimum value, where oh x Kh represents the x-th high-influence sensitivity correction parameter. x His represents the high impact sensitivity value of the x-th highly sensitive factor. x Rps represents the value of the x-th highly sensitive factor, t represents the t-th time point, and Rps represents the value of the x-th highly sensitive factor. t This represents the rate of change of the correction progress at the t-th time node, where T is the total number of time nodes within the monitoring interval, and num is the total number of high-impact sensitivity values. It should be noted that after the project starts, a fixed time period (such as the first month) is designated as a monitoring window to collect factor values ​​(such as temperature and wind force) and project progress at each time node in real time. This ensures that the timestamps of the factor values ​​and progress data are strictly aligned to avoid lag or deviation. The calculation of the interference value of low-sensitivity factors is used to quantify the comprehensive impact of low-sensitivity factors on the progress and serves as a benchmark value for subsequent correction. A real-time progress change curve is generated based on the progress data within the monitoring interval. The derivative of each time node on the curve is calculated to obtain the real-time progress change rate, which reflects the current actual progress trend. The optimization of the high-sensitivity factor correction parameters is used to eliminate the interference of low-sensitivity factors and separate the independent influence of high-sensitivity factors. The corrected high-sensitivity parameters are obtained through iterative optimization using a genetic algorithm (such as crossover and mutation), so that the model can dynamically adapt to the latest data. By continuously calibrating the model with real-time data, prediction failures caused by environmental changes or initial parameter deviations are avoided. Progress deviations are attributed to highly sensitive factors, and key variables are adjusted accordingly. The genetic algorithm has strong global search capabilities and is suitable for multi-parameter nonlinear optimization problems. Step S3: Based on the correction parameters of the high-sensitivity factors and the influence sensitivity values ​​of the low-sensitivity factors, obtain the predicted completion period interval of the project; obtain the predicted completion time, and based on the predicted completion time and the predicted completion period interval, obtain the reasonable long aging probability value of the project. In a preferred embodiment of the present invention, the process of obtaining the predicted completion period interval of the engineering project includes: The end time of the monitoring interval is recorded as the start time. The project progress of the project at the start time is obtained and recorded as the start progress. Each time node after the start time is obtained and recorded as the prediction node. The factor value of each factor at each prediction node is obtained and recorded as the prediction factor value. The time node corresponding to the starting time is denoted as t0, and each prediction node is numbered and denoted as t. q , t q This represents the q-th prediction node; it obtains the progress of the first project at the q-th prediction node. , where P tq-1 H represents the project progress at the (q-1)th prediction node. x This represents the predictor value of the x-th highly sensitive factor; and the second project progress at the q-th prediction node is obtained. E r This represents the predictor value of the r-th low-sensitivity factor; Obtain the timestamp corresponding to the prediction node where the progress of the first project first satisfies firP≥100%, denoted as T. fir ; Obtain the timestamp corresponding to the predicted node where the progress of the second project first satisfies thP≥100%, denoted as T. th ; Obtain the predicted completion period interval [T] fir T th ]; It should be noted that the end time of the monitoring interval is used as the prediction starting point to record the current project progress; the first project progress only considers the impact of the corrected high-sensitivity factors, and the second project progress only considers the impact of the low-sensitivity factors. The prediction interval is formed by finding the timestamps when the two paths first reach 100% or first exceed 100%; this represents the fastest completion time (optimal for high-sensitivity factors) and the slowest completion time (only low-sensitivity factors are affected). In a preferred embodiment of the present invention, the process of obtaining the reasonable long aging probability value of the engineering project includes: If the predicted completion time is denoted as "time", then the reasonable long aging probability value is obtained. , where W is the preset attenuation coefficient and W>0, and e is the Euler number, with an approximate value of 2.71828; The decay coefficient controls the probability growth rate and needs to be set according to the project's risk appetite. If time ≥ T th The probability of a reasonably long aging period is close to 100%, and the planned time exceeds the most pessimistic forecast; if time ≤ T th The probability is 0, and the planned time is earlier than the optimal prediction.

[0017] It should be further explained that "reasonable long aging" refers to the situation in material contracts where, for equipment and materials, especially large main equipment with complex technology and many parts, there may be long aging accounts payable that are of concern to the finance department due to objective and reasonable reasons such as the quality guarantee period stipulated in the contract or the quality problems of the equipment (parts).

[0018] The project progress prediction results of this invention, quantified by BIM technology as a real-time progress percentage, can serve as an objective basis for applying for progress payments, reducing errors from manual calculations. Based on real-time monitoring data, the system dynamically corrects parameters of highly sensitive factors, significantly improving the accuracy of progress predictions. If the probability of a reasonably long payment period for the project is low, early payment can be made; if the probability of a reasonably long payment period is high, it prompts the need to reserve funds for late payment penalties (to address contract delays) and adjust the advance payment plan or apply for a buffer advance payment. Sensitivity thresholds guide resources towards key factors (such as investment in other equipment), indirectly ensuring timely payment.

[0019] The foregoing has provided a detailed description of one embodiment of the present invention, but this description is merely a preferred embodiment and should not be construed as limiting the scope of the invention. All equivalent variations and modifications made within the scope of the claims of this invention should still fall within the patent coverage of this invention.

Claims

1. A machine learning method based on engineering project contract data, characterized in that, Includes the following steps: Step S1: Set several factors and obtain contract data for all historical engineering projects to obtain historical engineering project data. The historical engineering project data includes the factor values ​​of each factor at each time node within the historical engineering project cycle and the progress of the historical engineering project. Based on the historical engineering project data, the sensitivity and sensitivity values ​​of each factor are obtained, and high-sensitivity factors and low-sensitivity factors are determined based on the sensitivity of each factor. Step S2: Set the monitoring interval for the project, and monitor the factor values ​​of each factor and the project progress at each time node in real time within the monitoring interval to obtain real-time monitoring data; based on the real-time monitoring data, correct the sensitivity impact value of the highly sensitive factor to obtain the correction parameter of the highly sensitive factor. Step S3: Based on the correction parameters of the high-sensitivity factors and the influence sensitivity values ​​of the low-sensitivity factors, obtain the predicted completion period interval of the project; obtain the predicted completion time, and based on the predicted completion time and the predicted completion period interval, obtain the reasonable long aging probability value of the project.

2. The machine learning method based on engineering project contract data according to claim 1, characterized in that, In step S1, the process of obtaining the historical engineering project data includes: The start time and completion time of the historical project are obtained. The time period consisting of the start time and completion time is recorded as the historical project cycle. A time interval threshold is set, and several time nodes are selected by equally dividing the historical project cycle according to the time interval threshold.

3. The machine learning method based on engineering project contract data according to claim 1, characterized in that, In step S1, the process of obtaining the sensitivity of the influence of the factor includes: Based on the historical project progress at each time node in the historical project data, several data points are obtained. Each data point is fitted to generate a historical progress change curve. The slope of the tangent line at each data point is obtained on the historical progress change curve and recorded as the progress change rate at the corresponding time node. For any factor, all other factors besides the stated factor are denoted as "remaining factors," and the factor values ​​of these remaining factors are denoted as "remaining factor values." In the historical project data, the factor values ​​of the stated factor are selected when all remaining factor values ​​are consistent, and the progress change rate corresponding to each factor value at this time is obtained, resulting in a data set. Linear regression fitting is performed on the data set based on the least squares method to obtain a regression line. The slope of the regression line is obtained and denoted as the sensitive influence value, and the absolute value of the slope of the regression line is obtained and denoted as the influence sensitivity of the stated factor.

4. The machine learning method based on engineering project contract data according to claim 1, characterized in that, In step S1, the process of determining the highly sensitive factors and the low-sensitivity factors includes: Obtain the sensitivity of each factor's influence {Aff1, Aff2, ..., Aff} n }, where Aff n This represents the sensitivity of the nth factor, where n represents the total number of factors; a multiple threshold is set to obtain the sensitivity threshold. Aff i This represents the sensitivity of the influence of the i-th factor, where i ∈ [1, n] and i is a positive integer. λ The threshold is a multiple; factors whose influence sensitivity exceeds the sensitivity threshold are categorized as high-sensitivity factors, and factors whose influence sensitivity is less than or equal to the sensitivity threshold are categorized as low-sensitivity factors.

5. The machine learning method based on engineering project contract data according to claim 1, characterized in that, In step S1, the process of obtaining the historical engineering project data further includes: An engineering simulation model is established based on BIM technology. A three-dimensional model of the historical project at the time of completion is established through the engineering simulation model and denoted as the as-built three-dimensional model. The three-dimensional model of the historical project at the time node is also obtained and denoted as the real-time three-dimensional model. The model areas of the as-built three-dimensional model and the real-time three-dimensional model are obtained respectively and denoted as the completed area and the real-time area. The progress of the historical project at the time node is obtained as Poh = Fs / Ra × 100%, where Fs is the completed area and Ra is the real-time area.

6. The machine learning method based on engineering project contract data according to claim 1, characterized in that, In step S2, the process of correcting the sensitivity value of the highly sensitive factor includes: Based on the real-time monitoring data, the factor values ​​of all low-sensitivity factors at each time point within the monitoring interval are obtained. Based on the sensitivity impact value of each low-sensitivity factor, the low-sensitivity interference value at each time point is obtained. K r Eff represents the sensitivity value of the r-th low-sensitivity factor. r Let r represent the value of the r-th low-sensitivity factor, where r ∈ [1, m] and r is a positive integer, and m is the total number of low-sensitivity factors; Based on the real-time monitoring data, the project progress at each time node is obtained, several real-time data points are obtained, a real-time progress change curve is generated, and the slope of the tangent at each real-time data point on the real-time progress change curve is obtained, which is recorded as the real-time progress change rate at the time node. Based on the low-sensitivity interference values ​​at each time node, the correction progress change rate Rps = Ps + Lsi is obtained at each time node within the monitoring interval, where Ps is the real-time progress change rate at the time node. The factor values ​​and correction progress change rates of high-sensitivity factors at each time node are obtained, and the sensitivity impact values ​​of these high-sensitivity factors are recorded as high-impact sensitivity values. Correction parameters for each high-impact sensitivity value are obtained based on a genetic algorithm, such that... Reaching the minimum value, where ω x Kh represents the x-th high-influence sensitivity correction parameter. x His represents the high impact sensitivity value of the x-th highly sensitive factor. x Rps represents the value of the x-th highly sensitive factor, t represents the t-th time point, and Rps represents the value of the x-th highly sensitive factor. t This represents the rate of change of the correction progress at the t-th time node, where T is the total number of time nodes within the monitoring interval, and num is the total number of high-impact sensitivity values.

7. The machine learning method based on engineering project contract data according to claim 1, characterized in that, In step S3, the process of obtaining the predicted completion period interval of the project includes: The end time of the monitoring interval is recorded as the start time. The project progress of the project at the start time is obtained and recorded as the start progress. Each time node after the start time is obtained and recorded as the prediction node. The factor value of each factor at each prediction node is obtained and recorded as the prediction factor value. The time node corresponding to the starting time is denoted as t0, and each prediction node is numbered and denoted as t. q , t q This represents the q-th prediction node; it obtains the progress of the first project at the q-th prediction node. , where P tq-1 H represents the project progress at the (q-1)th prediction node. x This represents the predictor value of the x-th highly sensitive factor; and the second project progress at the q-th prediction node is obtained. E r This represents the predictor value of the r-th low-sensitivity factor; Obtain the timestamp corresponding to the prediction node where the progress of the first project first satisfies firP≥100%, denoted as T. fir ; Obtain the timestamp corresponding to the predicted node where the progress of the second project first satisfies thP≥100%, denoted as T. th ; Obtain the predicted completion period interval [T] fir T th ].

8. The machine learning method based on engineering project contract data according to claim 1, characterized in that, In step S3, the process of obtaining the reasonable long aging probability value of the project includes: If the predicted completion time is denoted as "time", then the reasonable long aging probability value is obtained. , where W is the preset attenuation coefficient and W>0, and e is the Euler number.

Images