Search

CN-121981816-A - Construction party credit ability quantitative assessment method based on multidimensional weighted modeling

CN121981816ACN 121981816 ACN121981816 ACN 121981816ACN-121981816-A

Abstract

The invention relates to a construction side credit ability quantization evaluation method based on multidimensional weighted modeling, belonging to the technical field of building engineering management, wherein the method comprises the steps of obtaining original behavior data, preprocessing and feature engineering processing to generate feature vectors; the method comprises the steps of constructing a derivative feature library, classifying features into corresponding evaluation dimensions, configuring feature weights, constructing a credit evaluation model, converting feature vectors into standardized credits, setting a configurable rule library through a rule dynamic engine, establishing a dynamic adjustment mechanism, adjusting dimension division standards, feature weights, rewarding and punishing rules and optimizing credit evaluation model parameters, mapping the standardized credits into credit grades and corresponding scores, and generating evaluation results after verifying rationality through a mapping verification mechanism. The method and the system improve evaluation accuracy and rationality, dynamically adapt to project and personnel changes, ensure evaluation timeliness and improve engineering management and control efficiency.

Inventors

  • KANG JIAN
  • CHEN YIN

Assignees

  • 上海紫鸾网络科技有限公司

Dates

Publication Date
20260505
Application Date
20260205

Claims (12)

  1. 1. A construction side credit ability quantitative assessment method based on multidimensional weighted modeling is characterized by comprising the following steps: s1, acquiring original behavior data related to credit of a constructor, preprocessing the original behavior data, performing feature engineering processing on the preprocessed original behavior data, and generating feature vectors; S2, constructing a derivative feature library, classifying features in the derivative feature library into corresponding evaluation dimensions based on preset dimension division standards, configuring feature weights for the features in the evaluation dimensions and the dimensions, and constructing a credit evaluation model by combining real-time event-driven reward and punishment rules in a manner of overlapping weighted dimension scores and real-time deductions based on basic scores, so as to convert the feature vectors into continuous standardized credit scores; s3, setting a configurable rule base through a rule dynamic engine, establishing a dynamic adjustment mechanism, adjusting the dimension division standard, the feature weight and the reward and punishment rule, and simultaneously optimizing the credit evaluation model parameters; And S4, mapping the standardized credit score into discrete credit grade and corresponding score, introducing a mapping verification mechanism to verify rationality, and generating an evaluation result.
  2. 2. The method of claim 1, wherein the specific process of preprocessing the original behavior data is to perform integrity check on the original behavior data, screen out data with missing information, supplement and perfect, identify abnormal values and reject, normalize and normalize the data after supplement and correct, and generate preprocessed original behavior data.
  3. 3. The method of claim 1, wherein the specific process of performing feature engineering processing on the preprocessed original behavior data is that basic features related to credit of a constructor are extracted from the preprocessed original behavior data, derivative features are built by combining with credit evaluation requirements of construction industry through feature fusion and feature conversion modes, then correlation screening is performed on the basic features and the derivative features, redundant features are removed, and standardized processing is performed on the screened features to generate the feature vector.
  4. 4. The method of claim 1, wherein the specific process of constructing the derivative feature library is to classify and comb the feature vectors, divide feature classes based on preset dimension division standards, label corresponding credit-related attributes and description information for features, establish a feature-in-warehouse checking mechanism, perform validity and suitability verification on the classified features, reject features which do not meet credit evaluation requirements, and orderly incorporate the checked features into a feature storage system to generate the derivative feature library.
  5. 5. The method according to claim 1, wherein the reward and punishment rules specifically comprise a positive incentive rule and a negative punishment rule, the positive incentive rule is set according to a behavior importance level and an influence radiation range based on positive behavior standards defined in a preset reward and punishment rule base, a corresponding incentive score adopts a accounting mode combining fixed score superposition and weighted score increment, the negative punishment rule is set according to a negative behavior standard defined in a preset reward and punishment rule base, a corresponding punishment score is set according to a behavior violation severity level and a hazard result level, and a accounting mode combining fixed score deduction and proportional deduction is adopted.
  6. 6. The method of claim 1, wherein the specific process of constructing the credit assessment model comprises the steps of firstly obtaining construction industry historical credit data, project performance data and the feature vector as training samples, dividing a training set and a verification set, selecting an algorithm framework based on assessment requirements, inputting the training set for training, marking a punishment rule by combining preset basic components and dimension weights, constructing a mapping relation between the feature vector and standardized credit, verifying by the verification set, optimizing a model structure and parameters for deviation data, and correcting mapping logic to generate the credit assessment model.
  7. 7. The method according to claim 1, wherein the specific process of converting the feature vector into the continuous standardized credit score is that feature information of a corresponding evaluation dimension in the feature vector is extracted, the configured evaluation dimension and feature weights under the dimension are combined, weighted scores of features under the dimension are calculated and summed to generate weighted dimension scores, the weighted dimension scores and the real-time deduction are overlapped based on the basic score, and meanwhile, a preliminary credit score is generated by combining with the reward and punishment rule, and the preliminary credit score is mapped to a preset continuous score interval through standardized processing to generate the standardized credit score.
  8. 8. The method of claim 1, wherein the specific process of setting the configurable rule base through the rule dynamic engine is that a standardized rule storage frame is built based on the rule dynamic engine, storage areas are divided according to evaluation dimension weights, credit level intervals and reward and punishment rules, the configured dimension weight standards, credit level intervals, reward and punishment situations and score standards are input according to classification, a mapping association of rule items and evaluation flows is established according to the classification input frame, operation interfaces for rule new addition, revision and deletion are reserved, logic consistency and data validity verification are carried out on the input rule items, conflict rules are removed, and parameters are calibrated.
  9. 9. The method of claim 1, wherein the dynamic adjustment mechanism comprises the specific working processes of obtaining construction project stage change information, industry credit management policy updating content and historical credit evaluation data feedback results, analyzing the evaluation dimension and the influence degree of features under the dimension on the evaluation results, adjusting the feature weight, synchronously tracking the same industry credit distribution trend and evaluation requirement, calibrating a credit level interval, revising the reward and punishment rules by combining construction side behavior feature evolution and project performance risk point migration, and synchronizing to the configurable rule base and the credit evaluation model in real time.
  10. 10. The method of claim 1, wherein the specific process of optimizing the credit assessment model parameters includes obtaining historical assessment results of the credit assessment model, actual performance feedback data and adjustment records of the configurable rule base, constructing an iteration sample set, eliminating abnormal data and performing data standardization preprocessing, analyzing deviation of a prediction result and actual performance on the basis of the iteration sample set, positioning parameters to be optimized, performing iterative calibration on the parameters by adopting a gradient descent algorithm, synchronizing embedding of the adjusted rule content, generating optimized model parameters, and substituting the optimized model parameters into the credit assessment model.
  11. 11. The method according to claim 1, wherein the specific process of mapping the standardized credit score into discrete credit grades and corresponding scores includes presetting a credit grade interval, defining corresponding credit grades and grade score ranges, substituting the standardized credit score into the preset credit grade interval for matching, analyzing the corresponding discrete credit grades according to the fallen interval, synchronously extracting the preset grade scores, and generating the corresponding score results.
  12. 12. The method of claim 1, wherein the mapping verification mechanism specifically comprises retrieving the discrete credit level and the corresponding scoring result, comparing the mapping record of the historical credit of the constructor with the historical record, verifying the consistency of the mapping result, comparing the mapping result with the credit level distribution and scoring interval of constructors of the same industry type, verifying the suitability of the level and the score and the intra-industry distinction, adjusting the scoring range of the industry boundary of the credit level interval based on the consistency and the rationality verification result, and re-executing the mapping and re-verifying.

Description

Construction party credit ability quantitative assessment method based on multidimensional weighted modeling Technical Field The invention belongs to the technical field of building engineering management, and particularly relates to a construction side credit ability quantitative assessment method based on multidimensional weighted modeling. Background At present, the credit evaluation method for construction side in the building industry gradually changes from traditional qualitative evaluation vectorization evaluation, but the prior art still has a plurality of limitations, and is difficult to meet the requirements of fine management of the industry. The traditional evaluation method relies on manual auditing and experience judgment, the evaluation index is single, only basic information such as qualification grade, past performance record and the like of a constructor is often focused, and the integrated analysis of multidimensional behavior data in the whole construction process is lacking, so that the evaluation result is one-sided and high in subjectivity, and the real credit ability of the constructor cannot be comprehensively reflected. With the popularization of big data technology in the construction industry, part of evaluation schemes begin to attempt to introduce data-driven quantization modes, but still have obvious defects. On one hand, the existing assessment model is imperfect in characteristic system, the excavation depth of relevant data of construction side credit is insufficient, the construction side credit is mostly stayed on a basic data layer, derivative characteristic construction aiming at characteristics of construction industry is lacked, the correlation degree analysis of characteristics and credit capacity is insufficient, redundant characteristics are more, assessment precision is affected, on the other hand, assessment dimension division is unreasonable, weight configuration is lacked in flexibility, a fixed weight mode is mostly adopted, dynamic changes of different project stages and different personnel groups cannot be adapted, and timely rewarding and punishing are difficult to be carried out on the construction side behaviors by combining real-time events, so that adaptability and timeliness of the assessment model are poor. In addition, the dynamic adjustment mechanism of the existing evaluation method is absent, and after most models are constructed, parameter optimization and rule adjustment are difficult to carry out according to industry policy updating, project propulsion conditions and construction side behavior evolution, and stability and rationality of an evaluation result are difficult to guarantee. Meanwhile, the mapping of the credit and the credit grade lacks a scientific verification mechanism, the problems of fuzzy grade division and insufficient industry distinction are easy to occur, and accurate and reliable basis cannot be provided for decisions such as project bid, performance supervision, risk prevention and control and the like. In view of the above technical drawbacks, there is a need for a construction side credit assessment method capable of integrating multidimensional data, precisely quantifying credit capability, and dynamically adapting to scene changes. Through constructing a perfect characteristic system, a flexible weight configuration mechanism, real-time linked reward and punishment rules and dynamic optimization models, full-period and fine quantitative evaluation of credit ability of construction parties is realized, the problems of strong subjectivity, single dimension, poor dynamic property, insufficient precision and the like of the existing method are solved, technical support is provided for the construction of a credit system of a building market, and high-quality development of the building industry is promoted. Disclosure of Invention In order to solve the problems in the prior art, the invention provides a construction side credit ability quantitative assessment method based on multidimensional weighted modeling. The invention aims at realizing the technical scheme that the construction party credit ability quantitative assessment method based on multidimensional weighted modeling comprises the following steps: s1, acquiring original behavior data related to credit of a constructor, preprocessing the original behavior data, performing feature engineering processing on the preprocessed original behavior data, and generating feature vectors; S2, constructing a derivative feature library, classifying features in the derivative feature library into corresponding evaluation dimensions based on preset dimension division standards, configuring feature weights for the features in the evaluation dimensions and the dimensions, and constructing a credit evaluation model by combining real-time event-driven reward and punishment rules in a manner of overlapping weighted dimension scores and real-time deductions based on basic scores, so