A multi-dimensional analysis-based bidding risk early warning method and system

Abstract

The application discloses a kind of based on multi-dimension analysis's bidding risk early warning method and system, it is related to artificial intelligence risk assessment field.Multiple angles multi-source acquisition bidding related data and bidding enterprise credit data, after completing format alignment, it is stored in bidding database;Bidding data in bidding database is retrieved, generates text dataset, extracts trend characteristics based on text dataset analysis financial data, calculates financial ratio;Loop through text dataset, obtain the keyword and key sentence in text dataset by the way of text similarity and keyword feature extraction, and the correlation of bidding is higher than preset threshold value;Risk prediction model is constructed, and financial ratio, keyword and key sentence are input into risk prediction model, and output early warning value, determine current bidding risk level.The application is according to project and historical preset accurate threshold value and real-time update evaluation data, realize real-time accurate early warning and help relevant party reduce loss.

Description

[0001] This application is the following application.

[0002] The application number is: 202510494455.1

[0003] Application date: April 18, 2025

[0004] The application is titled: "A Divisional Application of a Method and System for Early Warning of Bidding Risks Based on Artificial Intelligence". Technical Field

[0005] This invention relates to the field of artificial intelligence risk assessment, and more specifically to a bidding risk early warning method and system based on multi-dimensional analysis. Background Technology

[0006] For a long time, risk warning in bidding and tendering has relied primarily on relevant personnel reviewing bidding documents and procedures based on their own experience. This approach not only requires a significant investment of manpower and time but is also highly subjective, as different reviewers have varying standards and sensitivities for risk assessment, easily leading to inconsistent and inaccurate warning results. Furthermore, traditional methods have limited data analysis capabilities, making it difficult to comprehensively and deeply analyze massive amounts of bidding and tendering data. This hinders the uncovering of hidden relationships and potential risks, making it difficult to build effective risk assessment models and thus preventing timely and accurate risk detection. In addition, traditional risk warning mechanisms cannot monitor bidding and tendering activities in real time, only reviewing and processing them afterward. Faced with sudden and time-sensitive risks, it is difficult to take effective countermeasures in a timely manner, resulting in serious losses.

[0007] In recent years, artificial intelligence (AI) technology has made groundbreaking progress. Machine learning, deep learning, and natural language processing technologies can quickly and accurately analyze massive amounts of bidding data, uncovering potential patterns and risk characteristics. By constructing intelligent risk early warning models, real-time monitoring and dynamic warnings of bidding activities can be achieved, enabling timely detection and prevention of various risks. Applying AI technology to the field of bidding risk early warning can not only significantly improve the accuracy and efficiency of risk warnings and reduce interference from human factors, but also provide regulatory authorities and enterprises with scientific decision-making basis, contributing to the healthy and orderly development of the bidding market.

[0008] Therefore, developing artificial intelligence-based risk warning methods and systems for bidding and tendering has extremely important practical significance and application value. It is expected to break through the limitations of traditional risk warning models and significantly improve the risk management level of bidding and tendering activities. Summary of the Invention

[0009] In view of this, the present invention provides a bidding risk early warning method and system based on multi-dimensional analysis to solve the problems existing in the background technology.

[0010] To achieve the above objectives, the present invention adopts the following technical solution: A bidding risk early warning method based on multi-dimensional analysis includes the following steps: We collect bidding-related data and bidding enterprise credit data from multiple sources and angles by connecting with bidding platforms, corporate websites and credit databases, and store them in the bidding database after format alignment. The bidding data in the bidding database is retrieved and divided into non-text data and text data. The non-text data is converted into text data to generate a text dataset. Based on the text dataset, the financial data is analyzed to extract trend features and calculate financial ratios. The text dataset is iterated through, and keywords and key sentences with bidding relevance higher than a preset threshold are obtained through text similarity and keyword feature extraction. The text similarity is determined as follows: each bidding data in the text dataset is iterated through, and each bidding data is compared with the remaining bidding data. When the similarity is greater than the preset threshold, a warning of possible collusion is issued. First, the bidding text data is converted into vector form through a word vector model. Then, the cosine value of the cosine angle between the two vectors is calculated. The closer this value is to 1, the higher the text similarity. Based on the nature and scale of the project and historical risk data, a pre-set early warning threshold is established. Data is collected and updated in real time to build a risk prediction model. Financial ratios, keywords, and key phrases are input into the risk prediction model, and early warning values ​​are output. The early warning values ​​are compared with the pre-set early warning thresholds to determine the current bidding risk level.

[0011] Optionally, non-text data can be converted into text data. Specifically, the non-text data is sequentially traversed and retrieved to extract structured data, and the structured data is preprocessed and standardized. Based on the standardized structured data, image text is generated, and optical character recognition technology is used to obtain the text data in the image text, thus completing the data conversion.

[0012] Optionally, financial data can be analyzed based on text datasets to extract trend features and calculate financial ratios. Specifically, natural language processing tools can be used to extract key financial performance indicators (KPIs), and trend features can be extracted by analyzing the time series characteristics of the KPIs. Financial ratios can then be calculated based on the development capability ratio.

[0013] Optionally, the text dataset is iterated through, and keywords and key sentences with bidding relevance higher than a preset threshold are obtained from the text dataset through text similarity and keyword feature extraction. This includes the following steps: The system iterates through each bidding data entry in the text dataset, compares each bidding data entry with the remaining bidding data, and issues a warning about the possibility of bid rigging when the similarity exceeds a preset threshold. If the similarity is less than the preset threshold, the keywords of each bidding data are extracted using keyword feature extraction, and the key sentences containing the keywords are located to obtain pre-screened keywords and key sentences; Calculate the first correlation value between the keyword and the current bidding information. If the first correlation value is greater than the first threshold, the keyword and the key sentence containing the keyword are determined to be bidding information. If the correlation value is less than the first threshold, calculate the second correlation value between the key sentence containing the keyword and the current bidding information. If the second correlation value is greater than the second threshold, the current key sentence is determined to be bidding information.

[0014] Optionally, keyword feature extraction includes the following steps: Initial keywords are extracted from the data text using clustering algorithms. The initial keywords are analyzed based on the preference of bidding statements. The weight of words is determined by calculating the word frequency and inverse document frequency in the text. The initial keywords are represented by principal component analysis algorithm, and the keyword features are extracted by weighted average, feature-level fusion, and decision-level fusion.

[0015] Optionally, when a warning of potential bid-rigging is issued, it can be sent to the relevant parties to remove the current bidding data.

[0016] Optionally, a risk prediction model can be constructed using a BP neural network, and the risk prediction model can be optimized using a loss function.

[0017] A bidding risk early warning system based on multi-dimensional analysis includes: Bidding data collection and storage module: It is used to connect with bidding platforms, corporate websites and credit databases to collect bidding-related data and bidding corporate credit data from multiple angles and sources, and store them in the bidding database after format alignment. Bidding and tendering data analysis module: used to retrieve bidding and tendering data from the bidding and tendering database, which is divided into non-text data and text data. It converts non-text data into text data, generates a text dataset, analyzes financial data based on the text dataset to extract trend features, and calculates financial ratios. Keyword and Key Sentence Acquisition Module: This module iterates through the text dataset and extracts keywords and key sentences from the text dataset whose relevance to bidding and tendering exceeds a preset threshold using text similarity and keyword feature extraction. The text similarity is determined as follows: Each bidding and tendering record in the text dataset is iterated through, and each record is compared with the remaining records. When the similarity exceeds a preset threshold, a warning is issued regarding the possibility of bid rigging. First, the bidding and tendering text data is converted into vector form using a word vector model. Then, the cosine of the angle between the two vectors is calculated; the closer this value is to 1, the higher the text similarity. The bidding risk level determination module is used to preset early warning thresholds based on the nature and scale of the project and historical risk data, collect and update data in real time, build a risk prediction model, input financial ratios, keywords and key sentences into the risk prediction model, output early warning values, compare the early warning values ​​with the preset early warning thresholds, and determine the current bidding risk level.

[0018] As can be seen from the above technical solution, compared with the prior art, the present invention provides a bidding risk early warning method and system based on multi-dimensional analysis, which has the following beneficial effects: 1. By connecting with bidding platforms, corporate websites, and credit databases, data is collected from multiple angles and sources, comprehensively covering various information related to bidding, including project information and corporate credit data. This avoids the incomplete information problem that may exist with a single data source, making risk assessment more accurate and comprehensive.

[0019] 2. By analyzing financial data based on text datasets to extract trend features and calculate financial ratios, we can gain a deeper understanding of a company's operating conditions and potential risks from a financial perspective, providing quantitative indicators for risk assessment. Simultaneously, by extracting keywords and key phrases with high relevance to bidding and tendering through text similarity and keyword feature extraction, we can capture key information related to bidding and tendering risks in the text data, uncovering potential risks from a semantic level.

[0020] Real-time data collection and updating, along with the construction of risk prediction models, enables timely reflection of dynamic changes in the market and projects, ensuring the timeliness of risk assessments. By inputting financial ratios, keywords, and key phrases into the risk prediction model, early warning values ​​are output and compared with preset early warning thresholds to determine the risk level. This achieves real-time monitoring and accurate assessment of bidding risks, helping stakeholders to take timely measures to address risks and reduce losses. Attached Figure Description

[0021] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort.

[0022] Figure 1 This is a schematic diagram of the method flow of the present invention; Figure 2 This is a schematic diagram of the system structure of the present invention. Detailed Implementation

[0023] 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.

[0024] This invention discloses a method and system for early warning of bidding risks based on multi-dimensional analysis, such as... Figure 1 As shown, it includes the following steps: Step 1: Connect with the bidding platform, the company's official website and the credit database to collect bidding-related data and the credit data of bidding companies from multiple angles and sources, and store them in the bidding database after aligning the formats. Step 2: Retrieve bidding data from the bidding database, which is divided into non-text data and text data. Convert the non-text data into text data to generate a text dataset. Analyze the financial data based on the text dataset to extract trend features and calculate financial ratios. Step 3: Iterate through the text dataset and extract keywords and key sentences in the text dataset that have a bidding relevance higher than a preset threshold by means of text similarity and keyword feature extraction; Step 4: Based on the nature and scale of the project and historical risk data, preset early warning thresholds, collect and update data in real time, build a risk prediction model, input financial ratios, keywords and key phrases into the risk prediction model, output early warning values, compare the early warning values ​​with the preset early warning thresholds, and determine the current bidding risk level.

[0025] Furthermore, in step two, the non-text data is converted into text data. Specifically, the non-text data is sequentially traversed and retrieved to extract structured data, and the structured data is preprocessed and standardized. Based on the standardized structured data, image text is generated, and optical character recognition technology is used to obtain the text data in the image text, thus completing the data conversion.

[0026] Furthermore, step two, which involves analyzing financial data based on text datasets to extract trend features and calculate financial ratios, specifically involves: using natural language processing tools to extract key financial performance indicators (KPIs); analyzing the time series characteristics of these KPIs to extract trend features; and calculating financial ratios based on the development capability ratio.

[0027] Furthermore, in step three, the text dataset is iterated through repeatedly, and keywords and key sentences with bidding-related relevance higher than a preset threshold are obtained from the text dataset through text similarity and keyword feature extraction. Specifically, this includes the following steps: Step 3.1: Iterate through each bidding data entry in the text dataset, comparing its similarity with the remaining bidding data. If the similarity exceeds a preset threshold, issue a warning about the possibility of collusion. First, convert the bidding text data into vector form using a word vector model. Then, calculate the cosine of the angle between two vectors. The closer this value is to 1, the higher the text similarity. This measures the minimum number of single-character editing operations (insertion, deletion, replacement) required to convert one string to another. The smaller the edit distance, the higher the text similarity.

[0028] Step 3.2: If the similarity is less than the preset threshold, use keyword feature extraction to extract the keywords of each bidding data and locate the key sentence where the keywords are located to obtain the pre-screened keywords and key sentences; Step 3.3: Calculate the first relevance value between the keyword and the current bidding information. When the first relevance value is greater than the first threshold, determine that the keyword and the key sentence containing the keyword are bidding information. When the relevance value is less than the first threshold, calculate the second relevance value between the key sentence containing the keyword and the current bidding information again. When the second relevance value is greater than the second threshold, determine that the current key sentence is bidding information.

[0029] Furthermore, in step 3.1, a nested loop structure is used to process each bidding data entry in the dataset. The outer loop selects a benchmark bidding data entry, while the inner loop compares this benchmark entry with each of the remaining bidding data entries.

[0030] Furthermore, keyword feature extraction includes the following steps: Step 3.2.1: Use clustering algorithms to extract initial keywords from the data text, analyze the current initial keywords based on the bidding statement preferences, and determine the weight of words by calculating the word frequency and inverse document frequency in the text; Step 3.2.2: Based on the principal component analysis algorithm, the initial keywords are represented by features, and the keyword features are extracted by weighted average, feature-level fusion, and decision-level fusion.

[0031] Furthermore, when a warning of potential bid-rigging is issued, it is sent to the relevant parties, and the current bidding data is removed.

[0032] Furthermore, in step four, a risk prediction model is constructed using a BP neural network, and the model is optimized using a loss function. The preset warning threshold specifically includes the following two steps: Based on historical data statistics: Statistical analysis is performed on historical risk data to calculate the average, standard deviation, and other statistical measures for different types of projects under different risk indicators. Based on these statistical measures and the acceptable risk level of the project, reasonable early warning thresholds are set. For example, for the cost overrun risk of a certain type of project, if historical data shows an average overrun rate of 10% and a standard deviation of 3%, a mild early warning threshold can be set when the cost overrun rate reaches 13% (average + standard deviation), and a severe early warning threshold can be set when it reaches 16% (average + twice the standard deviation).

[0033] Consider the project's unique characteristics: When setting thresholds, fully consider the specific characteristics of the current project, such as its urgency, technical difficulty, and market competition. For projects with high technical difficulty, it may be necessary to appropriately lower the risk warning threshold to identify and address potential risks earlier.

[0034] Furthermore, for optimizing risk prediction models, new data can be continuously added to the training dataset as real-time data is collected, and the model can be retrained periodically to adapt to changes in the market environment and project characteristics. The model's performance should be evaluated regularly to check if its predictive effectiveness on new data remains good. If a performance decline is detected, the cause should be analyzed promptly, and adjustments made to the model, such as updating model parameters, adding new features, or selecting a different model.

[0035] and Figure 1 Corresponding to the method shown, this invention also discloses an artificial intelligence-based bidding risk early warning system for... Figure 1 The implementation of the method, specifically its structure, is as follows: Figure 2 As shown, it includes: Bidding data collection and storage module: It is used to connect with bidding platforms, corporate websites and credit databases to collect bidding-related data and bidding corporate credit data from multiple angles and sources, and store them in the bidding database after format alignment. Bidding and tendering data analysis module: used to retrieve bidding and tendering data from the bidding and tendering database, which is divided into non-text data and text data. It converts non-text data into text data, generates a text dataset, analyzes financial data based on the text dataset to extract trend features, and calculates financial ratios. Keyword and Key Sentence Acquisition Module: This module iterates through the text dataset and extracts keywords and key sentences from the text dataset that have a bidding relevance higher than a preset threshold by using text similarity and keyword feature extraction. The bidding risk level determination module is used to preset early warning thresholds based on the nature and scale of the project and historical risk data, collect and update data in real time, build a risk prediction model, input financial ratios, keywords and key sentences into the risk prediction model, output early warning values, compare the early warning values ​​with the preset early warning thresholds, and determine the current bidding risk level.

[0036] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on its differences from other embodiments. Similar or identical parts between embodiments can be referred to interchangeably. For the apparatus disclosed in the embodiments, since they correspond to the methods disclosed in the embodiments, the description is relatively simple; relevant parts can be referred to the method section.

[0037] The above description of the disclosed embodiments enables those skilled in the art to make or use the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A bidding risk early warning method based on multi-dimensional analysis, characterized in that, Includes the following steps: We collect bidding-related data and bidding enterprise credit data from multiple sources and angles by connecting with bidding platforms, corporate websites and credit databases, and store them in the bidding database after format alignment. The bidding data in the bidding database is retrieved and divided into non-text data and text data. The non-text data is converted into text data to generate a text dataset. Based on the text dataset, the financial data is analyzed to extract trend features and calculate financial ratios. The text dataset is iterated through, and keywords and key sentences with bidding relevance higher than a preset threshold are obtained through text similarity and keyword feature extraction. The text similarity is determined as follows: each bidding data in the text dataset is iterated through, and each bidding data is compared with the remaining bidding data. When the similarity is greater than the preset threshold, a warning of possible collusion is issued. First, the bidding text data is converted into vector form through a word vector model. Then, the cosine value of the cosine angle between the two vectors is calculated. The closer this value is to 1, the higher the text similarity. Based on the nature and scale of the project and historical risk data, a pre-set early warning threshold is established. Data is collected and updated in real time to build a risk prediction model. Financial ratios, keywords, and key phrases are input into the risk prediction model, and early warning values ​​are output. The early warning values ​​are compared with the pre-set early warning thresholds to determine the current bidding risk level.

2. The bidding risk early warning method based on multi-dimensional analysis according to claim 1, characterized in that, The process of converting non-text data into text data involves: sequentially traversing and retrieving the non-text data to extract structured data, and preprocessing and standardizing the structured data; generating image text based on the standardized structured data, and using optical character recognition technology to obtain the text data in the image text, thus completing the data conversion.

3. The bidding risk early warning method based on multi-dimensional analysis according to claim 1, characterized in that, Based on the analysis of text datasets, financial data is analyzed to extract trend features and calculate financial ratios. Specifically, natural language processing tools are used to extract key financial performance indicators, and trend features are extracted by analyzing the time series characteristics of the key financial performance indicators. Financial ratios are calculated based on the ratio of development capacity.

4. The bidding risk early warning method based on multi-dimensional analysis according to claim 1, characterized in that, The text dataset is iterated through repeatedly, and keywords and key sentences with bidding-related relevance exceeding a preset threshold are extracted using text similarity and keyword feature extraction methods. This process includes the following steps: The system iterates through each bidding data entry in the text dataset, compares each bidding data entry with the remaining bidding data, and issues a warning about the possibility of bid rigging when the similarity exceeds a preset threshold. If the similarity is less than the preset threshold, the keywords of each bidding data are extracted using keyword feature extraction, and the key sentences containing the keywords are located to obtain pre-screened keywords and key sentences; Calculate the first correlation value between the keyword and the current bidding information. If the first correlation value is greater than the first threshold, the keyword and the key sentence containing the keyword are determined to be bidding information. If the correlation value is less than the first threshold, calculate the second correlation value between the key sentence containing the keyword and the current bidding information. If the second correlation value is greater than the second threshold, the current key sentence is determined to be bidding information.

5. The bidding risk early warning method based on multi-dimensional analysis according to claim 3, characterized in that, Keyword feature extraction includes the following steps: Initial keywords are extracted from the data text using clustering algorithms. The initial keywords are analyzed based on the preference of bidding statements. The weight of words is determined by calculating the word frequency and inverse document frequency in the text. The initial keywords are represented by principal component analysis algorithm, and the keyword features are extracted by weighted average, feature-level fusion, and decision-level fusion.

6. The bidding risk early warning method based on multi-dimensional analysis according to claim 3, characterized in that, When a warning of potential bid-rigging is detected, it is sent to the relevant parties, and the current bidding data is removed.

7. The bidding risk early warning method based on multi-dimensional analysis according to claim 1, characterized in that, A risk prediction model is constructed using a backpropagation neural network, and the model is optimized using a loss function.

8. A bidding risk early warning system based on multi-dimensional analysis, characterized in that, include: Bidding data collection and storage module: It is used to connect with bidding platforms, corporate websites and credit databases to collect bidding-related data and bidding corporate credit data from multiple angles and sources, and store them in the bidding database after format alignment. Bidding and tendering data analysis module: used to retrieve bidding and tendering data from the bidding and tendering database, which is divided into non-text data and text data. It converts non-text data into text data, generates a text dataset, analyzes financial data based on the text dataset to extract trend features, and calculates financial ratios. Keyword and Key Sentence Acquisition Module: This module iterates through the text dataset and extracts keywords and key sentences from the text dataset whose relevance to bidding and tendering exceeds a preset threshold using text similarity and keyword feature extraction. The text similarity is determined as follows: Each bidding and tendering record in the text dataset is iterated through, and each record is compared with the remaining records. When the similarity exceeds a preset threshold, a warning is issued regarding the possibility of bid rigging. First, the bidding and tendering text data is converted into vector form using a word vector model. Then, the cosine of the angle between the two vectors is calculated; the closer this value is to 1, the higher the text similarity. The bidding risk level determination module is used to preset early warning thresholds based on the nature and scale of the project and historical risk data, collect and update data in real time, build a risk prediction model, input financial ratios, keywords and key sentences into the risk prediction model, output early warning values, compare the early warning values ​​with the preset early warning thresholds, and determine the current bidding risk level.

Images