Construction method of project investment budget full-cycle model

Abstract

The invention discloses an engineering investment budget full-period model construction method, and belongs to the technical field of engineering investment budget. The method comprises the following steps: obtaining engineering project data; summarizing the engineering project data to form a wide table required by modeling; performing data preprocessing on the data in the wide table to obtain a standby data set; according to the preprocessed data, taking actual funds of each key link of the project as dependent variables, taking other indexes as independent variables, and constructing a random forest regression model; according to the random forest regression model, engineering investment factor analysis at a full-period perspective is carried out, and engineering investment influence factors are mined; extracting principal component data by adopting principal component analysis based on the influence factors to obtain training set principal component data and test set principal component data; and constructing a multiple regression analysis model, and performing model training by using the principal component data of the training set. The method achieves the evaluation and analysis of the whole period of the project investment budget, and assists a power grid company in scientific decision-making of the project investment budget.

Description

Technical Field

[0001] The present invention relates to the technical field of engineering investment budget, and particularly relates to a method for constructing a full-cycle model of engineering investment budget. Background Art

[0002] With the development of social economy, the full-cycle cost is no longer an abstract concept at the beginning of the last century, but a relatively mature economic and technical method. On the premise of ensuring project quality and meeting the basic functions of the project, it organically combines all stages in the whole process of the project, realizes the safety, reliability and economy of the project, and makes the investment of the whole project reach the best input-output ratio.

[0003] Power engineering is an important part of the national infrastructure construction project. In order to meet the electricity demand of the people and the development demand of a low-carbon society, the number of power engineering constructions is gradually increasing, and the construction scope is also gradually expanding. How to improve the economic benefits of power engineering construction and strengthen the budget management of power engineering has become an important issue related to the quality and efficiency development of power enterprises.

[0004] To build a value ecosystem for the whole-process management of power grid investment and construction projects, the intelligent control of budget funds is one of its core contents. Engineering budget is an important part of project management, and its significance lies in improving the scientific nature of capital decision-making, reasonably arranging and effectively using funds in all aspects, and obtaining the maximum investment return. Engineering projects with project investment generally have the characteristics of large investment amount and long investment time span, and the accurate calculation of its engineering budget has always been a difficult point. Based on the whole-cycle process of the project, study an engineering investment budget model with high practical application value and strong operability to achieve effective control of investment.

[0005] Before the implementation of power engineering project investment, it is necessary to prepare a budget for it, which is an indispensable part of the cost accounting of power grid companies. For power grid companies with many infrastructure projects, large quantities of work and high investment amount, the accuracy of project engineering investment budget is related to the capital safety of the whole enterprise. Improving the accuracy of investment budget can reduce the capital risk of the enterprise.

[0006] The investment prediction scheme adopted by Chinese power enterprises needs to use uncertain data and expert experience for calculation, which has poor adaptability to different projects, a large amount of prediction work, and is affected by subjective factors, resulting in inaccurate prediction results. Summary of the Invention

[0007] The present invention provides a method for constructing a full-cycle model of engineering investment budget, and uses a big data machine learning algorithm to construct a full-cycle model of engineering investment budget to solve problems such as inaccurate engineering investment budget, poor adaptability, and large amount of prediction work.

[0008] According to one aspect of the present disclosure, a method for constructing a full-cycle model of a project investment budget is provided, the method comprising: (1) Acquire engineering project data; summarize the engineering project data to form a wide table required for modeling; (2) Preprocess the data in the wide table to obtain a backup data set. The data preprocessing includes: splicing the wide table, missing value processing, outlier processing, and normalization processing; (3) Based on the pre-processed data, the actual funds of each key link of the project were used as the dependent variable, and the remaining indicators were used as independent variables to construct a random forest regression model; the remaining indicators included: the name of the applied project, the project cycle, the project type, the estimated budget, the project milestones, and the corresponding data on pre-construction preparation, material declaration, material procurement, construction, settlement, and final settlement of the project; (4) Conducting a full-cycle engineering investment factor analysis based on the random forest regression model to explore the factors affecting engineering investment; (5) Based on the influencing factors, principal component analysis is used to extract the principal component data to obtain the principal component data of the training set and the principal component data of the test set; (6) Construct a multivariate regression analysis model and use the principal component data of the training set to train the model. The multivariate regression analysis model is used for the comprehensive cost estimation of the entire project cycle; (7) Use the principal component data of the test set to evaluate the effectiveness of the multivariate regression analysis model.

[0009] In a possible implementation, the method further includes: (8) if the result of the effect evaluation of the multivariate regression analysis model is that the model effect is good, then the multivariate regression analysis model is applied to the comprehensive cost estimation of the entire cycle of the engineering project; if the result of the effect evaluation of the multivariate regression analysis model is that the model effect is poor, then re-execute steps (4) to (7).

[0010] In a possible implementation, (4) the random forest regression model is used to perform a full-cycle engineering investment factor analysis to explore the factors affecting engineering investment, including: The random forest algorithm is used to analyze, mine and learn the backup data set, and the influencing factors of engineering investment corresponding to the key stages of the entire investment budget cycle are obtained according to the importance ranking of the indicators of the random forest; The analyzed influencing factor data are divided into training set and test set.

[0011] In one possible implementation, (6) a multivariate regression analysis model is constructed and the principal component data of the training set is used for model training. The multivariate regression analysis model is used for comprehensive cost estimation of the entire project cycle, including: Based on the principal component data of the training set, a comprehensive cost estimation model for the whole life cycle of engineering projects is constructed through the multiple regression analysis algorithm, and the weights of various indicators of the principal components are obtained by training the model.

[0012] In a possible implementation, (9) if the result of the effect evaluation of the multiple regression analysis model shows that the model effect is not good, return to step (4) to re-select the indicators of the random forest for importance ranking, and obtain the engineering investment influencing factors corresponding to the key stages of the whole life cycle of the investment budget.

[0013] In a possible implementation, (10) if the result of the effect evaluation of the multiple regression analysis model shows that the model effect is good, the least squares polynomial regression method is used to fit the proportion of each period's cost of the engineering project over time and the cumulative cost proportion over time, so as to show the time trend of the occurrence of engineering costs and construct the financial target curve of the engineering project.

[0014] Compared with the prior art, the beneficial effects of the present invention are as follows: Through the random forest algorithm (Random Forest, RF), the engineering investment factors from the perspective of the whole life cycle are analyzed to mine the engineering investment influencing factors. Based on these influencing factors as data, the principal component analysis and multiple regression analysis methods are used to construct a comprehensive cost estimation model for the whole life cycle of engineering projects. Using its powerful self-learning ability, the data of the influencing factors of the engineering investment projects of the power grid company are fused, and the multiple linear relationship between the influencing factors and the project funds is trained and established using historical data. Thus, a more accurate budget prediction value for the investment stage of the power engineering project is predicted by the model, improving the accuracy of budget allocation.

[0015] By using algorithms such as random forest, principal component analysis method, and multiple regression analysis, with reference to engineering estimates and budgets, combined with the engineering milestone plan, key stages such as project pre-stage, engineering pre-stage, construction implementation, and project settlement are selected, and the corresponding relationship between the time periods of each key stage and the financial cost progress is calculated, and the monthly theoretical cost curve of the whole life cycle of the project is drawn to complete the construction of the comprehensive cost estimation model for the whole life cycle of engineering projects.

[0016] Based on machine learning and big data analysis technologies, a new model for the whole life cycle of engineering investment budget is constructed and applied research is carried out, forming a set of models for the whole life cycle of engineering investment budget applicable to the power grid company, guiding early warning in advance and effectively preventing potential risks in the key stages of engineering investment, effectively improving the accuracy of the engineering investment budget of the power grid company and the company's capital risk prevention and control ability, realizing the evaluation and analysis of the whole life cycle of the engineering investment budget, and assisting the scientific decision-making of the engineering investment budget of the power grid company. Description of the Drawings

[0017] Figure 1A flowchart of a method for constructing a full-cycle model of a project investment budget according to an embodiment of the present disclosure is shown.

[0018] Figure 2 A curve showing the relationship between the project implementation time (month) and the proportion of funds issued according to an embodiment of the present disclosure is shown. DETAILED DESCRIPTION

[0019] Various exemplary embodiments, features and aspects of the present disclosure will be described in detail below with reference to the accompanying drawings. The same reference numerals in the accompanying drawings represent elements with the same or similar functions. Although various aspects of the embodiments are shown in the accompanying drawings, the drawings are not necessarily drawn to scale unless otherwise specified.

[0020] The word “exemplary” is used exclusively herein to mean “serving as an example, example, or illustration.” Any embodiment described herein as “exemplary” is not necessarily to be construed as preferred or advantageous over other embodiments.

[0021] In addition, in order to better illustrate the present disclosure, numerous specific details are given in the following specific embodiments. It should be understood by those skilled in the art that the present disclosure can also be implemented without certain specific details. In some examples, methods, means, components and circuits well known to those skilled in the art are not described in detail in order to highlight the subject matter of the present disclosure.

[0022] refer to Figure 1 - Figure 2 According to one aspect of the present disclosure, a method for constructing a full-cycle model of a project investment budget is provided, the method comprising: (1) Acquire engineering project data; summarize the engineering project data to form a wide table required for modeling; (2) Preprocess the data in the wide table to obtain a backup data set. The data preprocessing includes: splicing the wide table, missing value processing, outlier processing, and normalization processing; (3) Based on the pre-processed data, the actual funds of each key link of the project were used as the dependent variable, and the remaining indicators were used as independent variables to construct a random forest regression model; the remaining indicators included: the name of the applied project, the project cycle, the project type, the estimated budget, the project milestones, and the corresponding pre-construction preparation, material declaration, material procurement, construction, settlement, and final settlement data of the project; (4) Conducting a full-cycle engineering investment factor analysis based on the random forest regression model to explore the factors affecting engineering investment; (5) Based on the influencing factors, principal component analysis is used to extract the principal component data to obtain the principal component data of the training set and the principal component data of the test set; (6) Construct a multivariate regression analysis model and use the principal component data of the training set to train the model. The multivariate regression analysis model is used for the comprehensive cost estimation of the entire project cycle; (7) Use the principal component data of the test set to evaluate the effect of the multiple regression analysis model.

[0023] For example, collect the data required for model construction, summarize the collected data to form a wide table for modeling analysis, and complete the cleaning and transformation of the wide table data, including splicing the wide table data, handling missing values, outliers, and noise. For example, taking all the engineering projects within the jurisdiction of a certain company as the research object based on data requirements, obtain business data such as the name of the applied engineering project, project cycle, project type, estimated budget, project milestones, etc., as well as key stage data or status corresponding to the project, such as pre-construction preparation, material declaration, material procurement, construction, settlement, final accounts, etc., and complete data extraction and collection.

[0024] Splice the wide table data: Taking the project name or code as the entity, splice the data such as project cycle and estimated budget involved in data preparation into a detailed wide table that integrates multi-source data items to support subsequent data mining and analysis.

[0025] Handling missing values: When the data is severely missing, it will have a greater impact on the results of data mining and analysis. Therefore, for missing values, reasonable methods are used to fill them according to the attributes to which the missing values belong. Common methods include mean filling, K-nearest neighbor method, regression method, etc.

[0026] Handling outliers: This model uses common methods for identifying abnormal data, namely physical discrimination method and statistical discrimination method. For abnormal data, filling or deletion methods are adopted. The physical discrimination method is to judge the measured data that deviates from the normal result due to external interference, human error, etc. based on people's existing understanding of objective things, business, etc., and judge outliers. For example, the weighted average power consumption is a value greater than or equal to 0 in real life. If a negative value appears, it can be judged as an outlier according to the physical discrimination method.

[0027] The statistical discrimination method is to give a confidence probability and determine a confidence limit. Any error exceeding this limit is considered not to belong to the range of random errors and is regarded as an outlier. This model uses the 3σ criterion to handle the outliers of continuous data in the model wide table. The values outside the interval (μ - 3σ, μ + 3σ), where μ is the mean and σ is the standard deviation.

[0028] Normalization processing is a dimensionless operation for specific algorithms. In this project, the Min-Max normalization method is used to normalize continuous numerical values, standardize the continuous numerical values in the interval [0, 1], and convert them into dimensionless numerical values to facilitate the comparison and weighting of indicators with different units or magnitudes. The conversion function is as follows: ; Among them, max is the maximum value of the sample data, and min is the minimum value of the sample data.

[0029] Table 1 Data Traceability Details: ; The data is extracted and collected into an original data table, and the index items in the original data table are shown in Table 2.

[0030] Table 2 Original Index Items: Serial Number Index Name Remarks 1 Project Type 2 Voltage Level 3 Construction Scale - Line Length 4 Construction Scale - Substation Capacity 5 Construction Nature New Construction / Reconstruction / Expansion 6 Project Planned Cycle Duration of Each Cycle of Project Preparation, Engineering Preparation, Construction Implementation, and Project Settlement 7 Actual Project Cycle Duration of Each Cycle of Project Preparation, Engineering Preparation, Construction Implementation, and Project Settlement 8 Number of Participants 9 Material Consumption 10 Total Project Investment Amount 11 Static Investment 12 Equipment Consumption 13 Construction Engineering Cost Including: Direct Project Cost, Measure Fee, Regulatory Fee, Enterprise Management Fee 14 Installation Engineering Cost 15 Power Supply Subsidy Fee 16 Temporary Facilities Fee 17 Taxes 18 Equipment Purchase Fee 19 Project Supervision Fee 20 Differential Price Reserve Fund 21 Loan Interest 22 Additional Expenses Land Acquisition and Clearance Fee, Project Construction Management Fee, Technical Support Fee, Production Preparation Fee, Temporary Labor Fee, etc. 23 Engineering Insurance Fee 24 Commissioning and Trial Operation Fee for the Whole Set 25 Research and Experimentation Fee 26 Monthly Budget Data 27 Name of Key Project Phases Preparation before Start - up, Material Declaration, Material Procurement, Construction, Settlement, Final Account, etc. 28 Key Project Nodes (from the Perspective of Funds) Project Start - up, Construction, Project Commissioning, Project Settlement ; The wide table data after data preprocessing is standardized to form a standby data set, and the standby data set is divided into a training set and a test set. The training set is used to train the model and determine the weights of various indicators, and the test set is used to verify the model effect. The model parameters are adjusted according to the test set effect, and finally the comprehensive cost estimation model for the whole life cycle of the engineering project is constructed.

[0031] Select representative engineering investment projects for example verification and application. Integrate the verification results and expert suggestions, and iteratively improve the evaluation indicators, weight optimization, etc. of the whole life cycle model of engineering investment budget, providing a set of feasible, reasonable and efficient theoretical basis and foundation for the company's engineering investment budget.

[0032] Use the random forest algorithm to analyze, mine and learn the standby data set. According to the importance ranking of the indicators of the random forest, obtain the influencing factors corresponding to the key stages of the whole life cycle of investment budget; according to the data collection and processing results, select the actual funds of each key link as the dependent variable, and the other indicators as the independent variables to construct a random forest regression model. The implementation steps are as follows: (a) Randomly sample m samples from the original training set with replacement, and perform n samplings (the number of decision trees included in the random forest) in total, and finally generate n training sets; (b) For the n training sets, train n decision tree models respectively; (c) For a single decision tree model, assume that the number of training sample features is m, then each time of splitting, select the best feature according to the information gain (or information gain ratio and Gini index) to split the node; (d) Each tree is known to split like this until all the training examples of this node belong to the same class. In addition, pruning is not required during the splitting process of the decision tree; (e) Combine the generated multiple decision trees into a random forest. Since this project is a regression problem, the final prediction result is determined by the mean of the predicted values of multiple trees.

[0033] The importance of an indicator represents the degree of influence of the indicator on the prediction result. The greater the importance of an indicator, the greater the influence of the indicator on the prediction result; the smaller the importance, the smaller the influence of the indicator on the prediction result. The calculation steps are as follows: S1. Calculate the score of the training data on the random forest model according to a certain metric, denoted as S0; S2. Traverse each indicator in the training dataset. Each time, perform a shuffle operation (disrupt operation) on the corresponding indicator based on the original training dataset, and then use the random forest model to obtain the score of the shuffled dataset, denoted as S i ; S3. The importance of the i-th indicator is: .

[0034] S4. Sort according to the importance of the indicators. According to the sorting result, select the indicators with higher importance. The selected indicators are the influencing factors of project investment as shown in Table 3.

[0035] Table 3 Influencing factors of project investment after screening ; For the key factor data that has been analyzed, divide the training set and the test set, and extract the principal components of the training set and the test set respectively through the principal component analysis algorithm; The engineering project has a wide range of aspects, and there are many factors that affect the cost of the engineering project. If all influencing factors are analyzed one by one, not only is the calculation volume large, but also errors are likely to occur. Within a certain time range, the consumption of materials and labor in the engineering project is basically stable, and there are also certain rules in the changes of the unit prices of each sub-project. That is, the expected cost of each stage of the engineering project to be estimated can be determined by the cost of similar completed projects and the project stage cycle.

[0036] The principal component analysis algorithm is the most commonly used linear dimensionality reduction method. Its goal is to map high-dimensional data to a low-dimensional space through a certain linear projection, and it is expected that the information volume of the data is the largest (the variance is the largest) on the projected dimension, so as to use fewer data dimensions while retaining the characteristics of more original data points. The purpose of dimensionality reduction by the principal component analysis method is to reduce the dimension of the original features while trying to ensure that "the information volume is not lost", that is, to project the original features onto the dimension with the largest projected information volume as much as possible, project the original features onto these dimensions, and minimize the information volume loss after dimensionality reduction.

[0037] Therefore, adopting the idea of principal component analysis, several fitting comprehensive factors are selected to replace the original numerous influencing factors, achieving the effect of both simplifying calculations and improving accuracy. Using the principal component analysis method, the dataset in different dimensions is converted into the first n principal components with the cumulative contribution rate of principal components reaching ≥ 85%. The main steps of using the principal component analysis method in this project are as follows: The first step: For the key factor data analyzed except for the financial cost and time cycle of the engineering key stage, divide the training set and the test set with the engineering key stage as the reference, and the ratio of the training set to the test set is 3:1; The second step: Construct matrices for the training set and the test set respectively. Assume that the current dataset has m samples, and each sample has k index descriptions, forming the original matrix A, and standardize the original matrix; The second step: Obtain the correlation coefficient matrix Q of the standardized matrix A, and obtain the eigenvalues λ of the correlation coefficient matrix Q, getting k eigenvalues λ1 ≥ λ2... λ k ≥ 0; The third step: Solve the equation λ i a = aQ, and obtain the eigenvector a i ; The fourth step: Taking as the standard, determine the principal components.

[0038] When using principal component analysis, select the first 4 principal components. At this time, the information volume of the principal components is far greater than 85%.

[0039] For the above-mentioned principal component data of the training set, construct a comprehensive cost estimation model for the whole cycle of the engineering project through the multiple regression analysis algorithm, and train the model to obtain the weights of each index of the principal components; Use the principal components of the test set to verify the model effect. If the model effect is not good, return to the first step to reselect the indicators for importance ranking. If the model effect is good, proceed to the next step; The least squares polynomial regression method is used to fit the proportion of the cost of each period of the power grid project over time and the cumulative cost proportion over time to show the time trend of the engineering cost occurrence, and the financial target curve of this type of project is constructed.

[0040] Regression analysis is a statistical analysis method that, based on the study of the correlation relationships among various phenomena, fits a mathematical model to the changing trends of the dependent variable and independent variables for quantitative calculation. Regression analysis includes simple linear regression analysis and multiple regression analysis. Since there are many factors affecting project cost, the method of multiple regression analysis is adopted in this project. Multiple regression analysis uses a multiple regression model to study the mutual relationship between one dependent variable and multiple independent variables, so as to deduce or predict the future value of the dependent variable. In this project, the principal components obtained in the previous step are used for regression analysis, and principal component analysis and multiple regression analysis are comprehensively applied. Through the historical data of completed projects, the principal components are fitted and a regression equation is established, so as to estimate the project cost to be appraised. At the same time, compared with the traditional calculation method, several characteristic variables are used to replace its original influencing factors, unnecessary redundant factors are eliminated, time is saved, the amount of calculation is reduced, it is simple and convenient, and it is easy to understand and convenient to comprehend. The regression equation is constructed as follows: ; where y is the dependent variable (i.e., the financial cost and time period at each key stage), that is, the prediction object; x is the independent variable (i.e., the extracted principal components), and in this project, n takes the value of 4.

[0041] Taking the data of the same type of projects and the financial fund utilization rate and time period at each key stage as the independent variable and the dependent variable respectively, after principal component analysis, the method of multiple regression analysis is adopted, and through training and learning with the training set, an engineering project fund budget estimation model is obtained to realize the estimation of the financial cost at the key stages of the engineering project.

[0042] Finally, the principal components of the test set are used to verify the model effect. If the model effect is not good, the index coverage range or model parameters are adjusted.

[0043] Select representative engineering investment projects for example verification and application, and iteratively improve the optimization of the comprehensive cost estimation model for the whole life cycle of the engineering project by integrating the verification results and expert suggestions.

[0044] After the financial cost ratio and time period at each key stage of each project are predicted by the above multiple regression model, a specific scenario application of the comprehensive cost estimation model for the engineering project in the financial department of Lanzhou Power Supply Company is formed. Based on the results of this model, a theoretical budget curve for the whole life cycle of the engineering project is drawn for management business scenarios such as engineering budget review in the finance department.

[0045] (1) Determine the basic input data and substitute the corresponding data according to the principal component indicators.

[0046] (2) Conduct a trial calculation of the model to form a theoretical budget release curve for the whole life cycle of the project.

[0047] (3)Determine the measurement interval [start month, end month] of the budget release cycle to obtain the theoretical recommended value of budget release.

[0048] In a possible implementation manner, the method further includes: (8) If the result of the effect evaluation of the multiple regression analysis model is that the model effect is good, apply the multiple regression analysis model to the comprehensive cost estimation of the whole life cycle of the engineering project. If the result of the effect evaluation of the multiple regression analysis model is that the model effect is poor, re-execute from step (4) to step (7).

[0049] In a possible implementation manner, (4) Analyze the engineering investment factors from the perspective of the whole life cycle according to the random forest regression model, and mine the engineering investment influencing factors, including: Use the random forest algorithm to analyze, mine and learn the backup data set, and obtain the engineering investment influencing factors corresponding to the key stages of the whole life cycle of the investment budget according to the importance ranking of the random forest indicators; Divide the analyzed influencing factor data into a training set and a test set.

[0050] In a possible implementation manner, (6) Construct a multiple regression analysis model and use the principal component data of the training set for model training. The multiple regression analysis model is used for the comprehensive cost estimation of the whole life cycle of the engineering project, including: According to the principal component data of the training set, construct a comprehensive cost estimation model for the whole life cycle of the engineering project through the multiple regression analysis algorithm, and train the model to obtain the weights of each index of the principal component.

[0051] In a possible implementation manner, (9) If the result of the effect evaluation of the multiple regression analysis model is that the model effect is poor, return to step (4) to re-select the indicators of the random forest for importance ranking, and obtain the engineering investment influencing factors corresponding to the key stages of the whole life cycle of the investment budget.

[0052] In a possible implementation manner, (10) If the result of the effect evaluation of the multiple regression analysis model is that the model effect is good, use the least square polynomial regression method to fit the proportion of each period's cost of the engineering project over time and the proportion of the cumulative cost over time to show the time trend of the occurrence of the engineering cost, and construct the financial target curve of the engineering project.

[0053] Application example: Select a 35kV substation step-up new construction project of Lanzhou Power Supply Company as the trial calculation scenario for application verification.

[0054] (1) Input the initial information data Data collection is carried out according to the main component indicators of the cost estimation model. Project type: New construction; Total project amount: RMB 38.5251 million; Construction scale: Line length 31.36; Transformer capacity 10, Initial design approval time: April 11, 2019; Project start time: July 29, 2019; Project commissioning time: August 18, 2020; Project settlement time: September 30, 2020; Project planned total investment budget (tax included): RMB 38.2903 million; (2) Model curve generation The cost estimation model generates the estimation operators and values ​​for each engineering stage. The theoretical budget release curve for the entire project cycle is formed based on the model trial calculation. The time period and funding release ratio of each order are calculated. Figure 2 A curve is drawn showing the relationship between the project implementation time (month) and the proportion of funds allocated.

[0055] Table 4 Estimation operators and values ​​for each engineering stage generated by the cost estimation model ; (3) Customized cycle cost calculation The theoretical cost curve of the project's full cycle is formed based on the model calculation. The input budget release cycle calculation interval is 248 days, including 112 days in the early stage of the project and 184 days in the construction implementation. The estimated cost ratio of this cycle stage (296 days) is 18.49% in the early stage of the project, 32.35% in the construction implementation, and 19.01% in the construction implementation. The total budget release ratio is 51.36%.

[0056] Table 5 Theoretical cost of the project life cycle based on model calculation ; (4) Deviation analysis The absolute deviations of the three main phases of the project, namely the pre-project, construction implementation and project settlement, from the actual time period forecast were 2.75%, 6.22% and 18.60% respectively. The estimated values ​​are close to the phase time in the actual project and can be used to measure the correlation between the actual project cost and the time period; ; Optimize the full-cycle cost prediction method for power engineering projects, create a comprehensive cost estimation model for the full cycle of engineering projects, use its self-learning ability to integrate the data of influencing factors of the power engineering project budget, use historical data training to establish the relationship between influencing factors and the financial costs and time periods of key stages of the project, and select influencing factors through random forest indicator importance to avoid overfitting of model training.

[0057] From the perspective of project implementation, the concept and method of the full-cycle model of project investment budget are more scientific and reasonable in guiding the evaluation of construction organization design plans, the overall planning of project contracts, and the determination of project construction plans, etc., on the premise of comprehensively considering the proportion of life-cycle costs. It can directly affect the smooth progress of engineering projects, has far-reaching significance for the entire project, and also enhances the ability of power grid companies to use funds safely.

[0058] A relatively accurate predicted value of the proportion of the power engineering cycle cost can be estimated from this model, which can be used to assist the financial department in the annual and monthly budget review of projects, so as to improve the accuracy of fund utilization.

[0059] Under the development trend of the construction of a new power system, the investment concept of the company's project management also pays more attention to the realization of the benefit goal of precise investment, and realizes the optimization of capital allocation through budget control and issuance.

[0060] Compared with traditional prediction methods, the research results of the construction and application of the full-cycle model of project investment budget are more adaptable to different projects, have less prediction workload, are not easily affected by subjective factors, and the results are also more accurate. It can guide early warning in advance and effectively prevent potential risks in the key stages of project investment, and effectively improve the accuracy of project investment budget and the company's capital risk prevention and control ability of power grid companies. The results of this project not only have wide applicability, but also have high promotion value.

[0061] During the research process of this project, due to the influence of existing data support and data quality, it was not fully concretized. In the actual application of the subsequent model results, according to the actual situation, the key index data of each engineering project in the specific implementation process can be further refined, and the comprehensive cost estimation model and application method of the engineering project full cycle can be optimized, so as to better conform to the business reality of the engineering project budget management of the financial department of the prefecture-level company.

[0062] The above has described the embodiments of the present disclosure. The above description is exemplary, not exhaustive, and is not limited to the disclosed embodiments. Many modifications and variations are obvious to those of ordinary skill in the art in the technical field without departing from the scope and spirit of the described embodiments. The choice of terms used herein is intended to best explain the principles of the embodiments, the practical application, or the improvement of the technology in the market, or to enable other ordinary technical personnel in the technical field to understand the embodiments disclosed herein.

Claims

1. A method for constructing a full-cycle model of engineering investment budget, characterized in that, The method comprises: (1) Acquire engineering project data; summarize the engineering project data to form a wide table required for modeling; (2) Preprocess the data in the wide table to obtain a backup data set. The data preprocessing includes: splicing the wide table, missing value processing, outlier processing, and normalization processing; (3) Based on the pre-processed data, the actual funds of each key link of the project were used as the dependent variable, and the remaining indicators were used as independent variables to construct a random forest regression model; the remaining indicators included: the name of the applied project, the project cycle, the project type, the estimated budget, the project milestones, and the corresponding data on pre-construction preparation, material declaration, material procurement, construction, settlement, and final settlement of the project; (4) Conducting a full-cycle engineering investment factor analysis based on the random forest regression model to explore the factors affecting engineering investment; (5) Based on the influencing factors, principal component analysis is used to extract the principal component data to obtain the principal component data of the training set and the principal component data of the test set; (6) Construct a multivariate regression analysis model and use the principal component data of the training set to train the model. The multivariate regression analysis model is used for the comprehensive cost estimation of the entire project cycle; (7) Use the principal component data of the test set to evaluate the effectiveness of the multivariate regression analysis model.

2. The method for constructing an engineering investment budget full-cycle model according to claim 1, wherein The method further includes: (8) if the result of the effect evaluation of the multivariate regression analysis model is that the model effect is good, then the multivariate regression analysis model is applied to the comprehensive cost estimation of the entire cycle of the engineering project; if the result of the effect evaluation of the multivariate regression analysis model is that the model effect is not good, then re-execute steps (4) to (7).

3. A method for constructing a full-cycle model of engineering investment budget according to claim 1, characterized in that, (4) Conduct a full-cycle analysis of project investment factors based on the random forest regression model to explore factors affecting project investment, including: The random forest algorithm is used to analyze, mine and learn the backup data set, and the influencing factors of engineering investment corresponding to the key stages of the entire investment budget cycle are obtained according to the importance ranking of the indicators of the random forest; The analyzed influencing factor data are divided into training set and test set.

4. A method for constructing a full-cycle model of engineering investment budget according to claim 1, characterized in that, (6) Construct a multivariate regression analysis model and use the principal component data of the training set to train the model. The multivariate regression analysis model is used for the comprehensive cost estimation of the entire project cycle, including: According to the principal component data of the training set, a comprehensive cost estimation model for the entire project cycle is constructed through a multivariate regression analysis algorithm, and the weights of various indicators of the principal components are obtained by training the model.

5. A method for constructing an all-round life cycle model of engineering investment budget according to claim 2, characterized in that (9) If the result of the performance evaluation of the multivariate regression analysis model is that the model effect is not good, return to step (4) and reselect the indicators of the random forest to sort the importance, and obtain the project investment influencing factors corresponding to the key stages of the entire investment budget cycle.

6. The construction method of an engineering investment budget full-cycle model according to claim 2, characterized in that, (10) If the result of the performance evaluation of the multivariate regression analysis model is that the model effect is good, the least squares polynomial regression method is used to fit the cost proportion of each period of the project over time and the cumulative cost proportion over time to show the time trend of the project cost and construct the financial target curve of the project.

Images