Construction method of intelligent evaluation model

Abstract

The invention provides an intelligent evaluation model construction method. The intelligent evaluation model construction method comprises the steps of 1, extracting scoring indexes from a bid inviting file and establishing a scoring tree-shaped model; 2, extracting index item contents from the bidding file by using a natural language processing technology; 3, matching the content of the index item with the scoring tree model, and establishing a fuzzy scoring model to calculate the score of a single item; step 4, establishing a Bayesian scoring model based on the scoring tree model to calculate index scores; 5, establishing a first weighted scoring model in combination with the scoring tree model and the index weight to calculate a factor score; step 6, establishing a second weighted scoring model by using the factor scores and the weights to calculate a comprehensive score; 7, summarizing the models to generate an intelligent evaluation model; according to the invention, the bid evaluation automation and intelligence are realized, the bid evaluation efficiency and fairness are improved, and the standardization and normalization of the bidding and tendering field are promoted.

Description

Technical Field

[0001] The present invention relates to the technical field of electronic bidding, and in particular to a method for constructing an intelligent evaluation model. Background Art

[0002] With the rapid development of information technology, electronic bidding systems have become an indispensable part of modern procurement activities; electronic bidding not only improves procurement efficiency, but also enhances transparency and fairness; however, in the process of electronic bidding, the automation and intelligent processing of the bid evaluation stage is still a problem that needs to be solved urgently;

[0003] In the prior art, manual judgment in the bid evaluation process is often affected by personal preferences and experience, resulting in subjectivity and inconsistency in the bid evaluation results; the bid documents usually contain a large amount of text descriptions and technical specifications, and manual reading and understanding of this information is time-consuming and prone to omissions or misunderstandings; at the same time, the bid evaluation involves multi-level scoring indicators, and it is a challenge to manage and coordinate the relationship between these indicators; if some information in the bid documents is not clear enough, the evaluators are required to make subjective judgments, which increases the uncertainty and difficulty of the bid evaluation; at the same time, in the absence of sufficient historical data or prior knowledge, it is difficult for the evaluators to make accurate judgments. Summary of the invention

[0004] The present invention provides a method for constructing an intelligent evaluation model to solve the problems of poor subjectivity, low processing efficiency and complex scoring logic existing in the prior art.

[0005] The present invention provides a method for constructing an intelligent evaluation model, comprising:

[0006] Step 1: extract the scoring index items, scoring indicators, scoring factors, and scoring items in the bidding documents, and establish a scoring tree model according to their relationships;

[0007] Step 2: extracting the index item content corresponding to the scoring index item from the bidding document based on the scoring tree model and natural language processing technology;

[0008] Step 3, matching and comparing the index item content of the bidding document with the scoring index items in the scoring tree model, and establishing a fuzzy scoring model based on the comparison results to calculate the scores of each scoring index item;

[0009] Step 4: According to the topological structure of the scoring tree model and various scoring index items, a Bayesian scoring model is established to calculate the scores of various scoring index items;

[0010] Step 5: Based on the topological structure of the scoring tree model, the scores of each scoring indicator and the preset indicator weights, a first weighted scoring model is established to calculate the scores of each scoring factor;

[0011] Step 6: Based on the topological structure of the scoring tree model, the scores of each scoring factor and the preset factor weights, a second weighted scoring model is established to calculate the sum of the scores of each scoring item, i.e., the comprehensive score;

[0012] Step seven, aggregating the scoring tree model, the fuzzy scoring model, the Bayesian model, the first weighted scoring model and the second weighted scoring model to generate an intelligent evaluation model.

[0013] According to a method for constructing an intelligent evaluation model provided by the present invention, in step 1, the relationship among the scoring index items, scoring indexes, scoring factors, and scoring items includes:

[0014] Scoring items are a classification of the main aspects or dimensions that need to be considered in the bid evaluation process, which include several scoring factors;

[0015] Scoring factors are specific criteria or dimensions for specific scoring items, which are used to refine the content of the scoring items and include several scoring indicators.

[0016] Scoring indicators are further refined evaluation criteria based on scoring factors, which are used to quantify or describe the requirements or expected levels of scoring factors. They include several scoring indicator items.

[0017] The scoring index item is the specific measurement value or judgment condition under the scoring index, which is used to quantify the requirements or expected level of the scoring index and is the most basic evaluation unit in the bid evaluation.

[0018] According to a method for constructing an intelligent evaluation model provided by the present invention, in step 2, extracting the index item content corresponding to the scoring index item from the bidding document includes:

[0019] Extract and analyze keywords and semantics in bidding documents through natural language processing technology to obtain index item contents;

[0020] For numerical data, extract the numerical value and generate numerical fuzzy sets according to the scoring index items in the bidding documents;

[0021] For text data, TF-IDF technology is used to quantify the text and convert it into a vector value. At the same time, a text fuzzy set is generated according to the scoring index items in the bidding documents.

[0022] The extracted indicator items are manually identified to eliminate errors in natural language processing technology.

[0023] According to a method for constructing an intelligent evaluation model provided by the present invention, in step three, the fuzzy scoring model is divided into:

[0024]

[0025] Among them, μ(x) is the membership function; a is the left boundary of the fuzzy set, indicating that the fuzzy set has the minimum value of significant membership at the beginning; b is the vertex of the fuzzy set, indicating the most ideal state, that is, the membership reaches the maximum value; c is the right boundary of the fuzzy set, indicating that the fuzzy set ends with the maximum value of significant membership; among them, a, b, and c are all obtained from the scoring index items; x is the input value;

[0026] Substituting the numerical data in the obtained indicator item content as the input value x into the numerical scoring model, and determining and calculating the corresponding result based on the relationship between the input value x and a, b, and c, that is, the numerical score given in the scoring indicator item;

[0027] Substituting the vector value corresponding to the text data in the obtained index item content as the input value x into the numerical scoring model, and determining and calculating the corresponding result based on the relationship between the input value x and a, b, and c, that is, the text score given in the scoring index item;

[0028] The numerical score and the text score are added according to the preset weight to obtain the scoring index item score S corresponding to the scoring index item. V ;

[0029] Among them, the weight of the numerical score is greater than the weight of the text score.

[0030] According to a method for constructing an intelligent evaluation model provided by the present invention, in step 4, the Bayesian scoring model uses Beta distribution as a prior distribution, and the Beta distribution is a conjugate prior of the binomial distribution, specifically:

[0031] S41, setting initial prior parameters α and β; the Bayesian scoring model includes two prior parameters α and β, which are set based on the importance of the scoring index item relative to the scoring index; the importance is quantified as a decimal between 0 and 1, and the values ​​of the parameters α and β are adjusted by a preset coefficient, and the larger the quantized value corresponding to the importance, the larger the parameter α and the smaller the parameter β, and vice versa;

[0032] S42, for each scoring index item score S Vi Update the prior parameters α and β;

[0033] The updated final prior parameter α i =α+S Vi ;

[0034] The updated final prior parameter β i =β+(1-S Vi );

[0035] Among them, SVi is the score of the i-th scoring indicator item;

[0036] S43, calculating the expected value of the posterior distribution as the final score of the scoring indicator item;

[0037]

[0038] Among them, S Wi is the final score of the i-th scoring indicator item;

[0039] S44, setting weights based on the quantified values ​​of the importance of each scoring indicator item, and calculating the final score S of the scoring indicator. Fj ;

[0040]

[0041] Among them, S Fj is the final score of the jth scoring indicator, n is the number of scoring indicator items, φ i is the weight of the i-th scoring indicator item.

[0042] According to a method for constructing an intelligent evaluation model provided by the present invention, in step five, the first weighted scoring model is:

[0043]

[0044] Among them, S Gk is the final score of the kth scoring factor, m is the number of scoring factors, ω j is the weight of the jth scoring indicator.

[0045] According to a method for constructing an intelligent evaluation model provided by the present invention, in step six, the second weighted scoring model is:

[0046]

[0047] Among them, S Qz is the final score of the zth scoring item, p is the number of scoring items, ε k is the weight of the kth scoring factor.

[0048] According to a method for constructing an intelligent evaluation model provided by the present invention, in step six, it also includes:

[0049] After outputting the comprehensive score, the comprehensive scores of each bidding document are sorted.

[0050] Compared with the prior art, the present invention has the following beneficial effects:

[0051] Automatically process information in bidding documents, reducing the time spent on manual reading and understanding of bidding documents; reducing subjective factors in the bidding evaluation process, making the bidding evaluation results more objective and fair;

[0052] Through the scoring tree model and natural language processing technology, key information can be quickly located and extracted to speed up the bid evaluation process; the scoring tree model clearly organizes the relationship between scoring indicators and simplifies the bid evaluation process;

[0053] The use of fuzzy scoring model and Bayesian scoring model ensures the consistency and repeatability of scoring;

[0054] The use of the first weighted scoring model and the second weighted scoring model makes the calculation of the bid evaluation results simpler, and the second weighted scoring model makes the scoring of the scoring items more sensitive to the scoring boundaries of the scoring factors, thus avoiding the problem of similar scores.

[0055] To sum up, this solution solves the main problems in the traditional bid evaluation process through automation and intelligent means, improves the efficiency, fairness and accuracy of bid evaluation, and also provides support for standardization and regularization in the bidding field.

[0056] Other features and advantages of the present invention will be described in the following description, and partly become apparent from the description, or understood by practicing the present invention. The purpose and other advantages of the present invention can be realized and obtained by the structures particularly pointed out in the written description and the accompanying drawings.

[0057] The technical solution of the present invention is further described in detail below through the accompanying drawings and embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0058] The accompanying drawings are used to provide a further understanding of the present invention and constitute a part of the specification. Together with the embodiments of the present invention, they are used to explain the present invention and do not constitute a limitation of the present invention. In the accompanying drawings:

[0059] Figure 1 It is a flowchart of a method for constructing an intelligent evaluation model provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0060] The preferred embodiments of the present invention are described below in conjunction with the accompanying drawings. It should be understood that the preferred embodiments described herein are only used to illustrate and explain the present invention, and are not used to limit the present invention.

[0061] Embodiment 1:

[0062] The present invention provides a method for constructing an intelligent evaluation model. Figure 1 ,include:

[0063] Step 1: extract the scoring index items, scoring indicators, scoring factors, and scoring items in the bidding documents, and establish a scoring tree model according to their relationships;

[0064] Step 2: Based on the scoring tree model and natural language processing technology, extract the index item content corresponding to the scoring index item from the bidding document;

[0065] Step 3: Match and compare the index item content of the bidding document with the scoring index items in the scoring tree model, and establish a fuzzy scoring model based on the comparison results to calculate the scores of each scoring index item;

[0066] Step 4: According to the topological structure of the scoring tree model and various scoring indicators, a Bayesian scoring model is established to calculate the scores of various scoring indicators;

[0067] Step 5: Based on the topological structure of the scoring tree model, the scores of each scoring indicator and the preset indicator weights, a first weighted scoring model is established to calculate the scores of each scoring factor;

[0068] Step 6: Based on the topological structure of the scoring tree model, the scores of each scoring factor and the preset factor weights, a second weighted scoring model is established to calculate the sum of the scores of each scoring item, i.e., the comprehensive score;

[0069] Step 7, summarizing the scoring tree model, the fuzzy scoring model, the Bayesian model, the first weighted scoring model and the second weighted scoring model to generate an intelligent evaluation model;

[0070] The principles and beneficial effects of the above embodiments are: extracting information such as scoring index items from the bidding documents in an automated manner and establishing a scoring tree model, which greatly reduces the time for manual sorting and understanding; using natural language processing technology to automatically extract relevant information from the bidding documents, further improving the bid evaluation efficiency;

[0071] Using fuzzy scoring model to process the matching and comparison between indicator content and scoring indicator can effectively handle fuzzy and ambiguous information; Bayesian scoring model takes into account prior probability and actual data, and can give more accurate scoring results;

[0072] Based on a clear scoring tree model and automated scoring calculation, the influence of human subjective factors is reduced, making the bid evaluation process more fair and transparent; the hierarchical relationship between scoring indicators is clearly defined through the scoring tree model, making the entire bid evaluation process more concise and clear;

[0073] Through standardized scoring processes and models, the standardization and regularization of bid evaluation work is promoted, which helps to unify the bid evaluation criteria for different projects.

[0074] In order to further optimize the above embodiment, in step 1, the relationship among the scoring index items, the scoring index, the scoring factors, and the scoring items includes:

[0075] Scoring items are a classification of the main aspects or dimensions that need to be considered in the bid evaluation process, which include several scoring factors;

[0076] Scoring factors are specific criteria or dimensions for specific scoring items, which are used to refine the content of the scoring items and include several scoring indicators.

[0077] Scoring indicators are further refined evaluation criteria based on scoring factors, which are used to quantify or describe the requirements or expected levels of scoring factors. They include several scoring indicator items.

[0078] Scoring indicator items are specific measurement values ​​or judgment conditions under scoring indicators, which are used to quantify the requirements or expected levels of scoring indicators and are the most basic evaluation units in bid evaluation.

[0079] It should be noted that scoring items are the classification of the main aspects or dimensions that need to be considered in the bid evaluation process; they are usually relatively macro concepts, such as "technical solution", "project management capability", etc.

[0080] Scoring factors are specific evaluation criteria or dimensions under a specific scoring item, which are used to refine the content of the scoring item. For example, under the scoring item "technical solution", there may be scoring factors such as "technical innovation" and "feasibility of implementation plan";

[0081] Scoring indicators are further refined evaluation criteria based on scoring factors, which are used to quantify or describe the requirements or expected levels of scoring factors. Continuing with the above example, the scoring factor "technological innovation" may include scoring indicators such as "number of patents" and "degree of technological advancement";

[0082] Scoring indicator items are specific measurement values ​​or judgment conditions under scoring indicators, which are used to quantify the requirements or expected levels of scoring indicators. Scoring indicator items are the most basic evaluation units. For example, the scoring indicator "number of patents" can be subdivided into scoring indicator items such as "the number of patents owned is greater than 5";

[0083] For example:

[0084] Scoring item: "Technical solution"

[0085] Scoring factor: “Technical innovation”

[0086] Scoring indicator: “Number of patents”

[0087] Scoring indicator: "The number of patents owned is greater than 5"

[0088] Scoring indicator: "The number of patents owned is between 2 and 5"

[0089] Scoring indicator: "The number of patents owned is less than 2"

[0090] This hierarchical structure helps evaluators to systematically evaluate bid documents and ensure the comprehensiveness and objectivity of the evaluation; each layer provides a more specific guidance framework for the next layer, thereby ensuring that the various attributes of the bid documents can be accurately measured during the bid evaluation process.

[0091] In order to further optimize the above embodiment, in step 2, the index items corresponding to the scoring index items are extracted from the bidding document, including:

[0092] Extract and analyze keywords and semantics in bidding documents through natural language processing technology to obtain index item contents;

[0093] For numerical data, extract the numerical value and generate numerical fuzzy sets according to the scoring index items in the bidding documents;

[0094] For text data, TF-IDF technology is used to quantify the text and convert it into a vector value. At the same time, a text fuzzy set is generated according to the scoring index items in the bidding documents.

[0095] Manually identify the extracted indicator items to eliminate errors in natural language processing technology;

[0096] It should be noted that in step 2, the index item content corresponding to the scoring index item is extracted by natural language processing (NLP) technology, which can be performed according to the following process:

[0097] 1. Keyword and semantic extraction analysis:

[0098] Use natural language processing technology to analyze bid documents and identify key words and phrases related to scoring criteria;

[0099] Analyze the context in the bidding documents and understand the meaning of key words and phrases and the relationship between them to obtain more accurate information;

[0100] 2. Numerical data processing:

[0101] Extract numerical data from the bidding documents and correspond them to the scoring index items in the bidding documents;

[0102] For numerical data, create a numerical fuzzy set, that is, use fuzzy logic to express the uncertainty range of the numerical value; this helps to consider the actual meaning of the numerical value rather than just the absolute value during the bid evaluation process;

[0103] 3. Text data quantification:

[0104] For non-numerical text data, TF-IDF (Term Frequency Inverse Document Frequency) technology is used to quantify it;

[0105] The text data is converted into vector form for comparison and analysis. Similarly, the corresponding text fuzzy sets are generated according to the scoring index items in the bidding documents to evaluate the relevance and importance of the text content.

[0106] 4. Manual verification:

[0107] After automatic extraction and processing, experienced bid evaluation personnel are required to manually review the extracted indicator items;

[0108] This step is to ensure the accuracy of the extracted information and eliminate the errors that may be caused by natural language processing technology;

[0109] Assume that one of the scoring indicators in the tender document is "Experience of the Project Team", and this indicator item requires the assessment of the relevant work experience of the project team members in the past five years;

[0110] Keyword and semantic extraction analysis:

[0111] Natural language processing tools can identify sentences related to “project team” and “work experience” from bid documents;

[0112] Analyze the semantics of these sentences to determine which information is directly related to the scoring indicators;

[0113] Numerical data processing:

[0114] If the bidding document mentions that a team member has "7 years of relevant work experience", this value can be extracted and a fuzzy set can be generated based on the requirements of the scoring indicator items;

[0115] Fuzzy sets can be expressed as: "very rich (>10 years)", "rich (610 years)", "average (25 years)", etc.

[0116] Text data quantization:

[0117] For a paragraph describing team members’ qualifications and experience, the TF-IDF algorithm can be used to calculate the weights of key terms and convert this information into a vector;

[0118] The generated vectors can be used to compare the descriptions of project team experience in different bidding documents and generate text fuzzy sets accordingly;

[0119] Manual verification:

[0120] The bid evaluation team needs to review the data extracted through natural language processing technology and confirm its accuracy and applicability;

[0121] For example, check that all relevant work experience descriptions are correctly identified and ensure that no important points of information are omitted;

[0122] Through the above steps, the index items related to the scoring index items can be effectively extracted from the bidding documents, and provide strong support for the subsequent bid evaluation work.

[0123] In order to further optimize the above embodiment, in step three, the fuzzy scoring model is divided into:

[0124]

[0125] Among them, μ(x) is the membership function; a is the left boundary of the fuzzy set, indicating that the fuzzy set has the minimum value of significant membership at the beginning; b is the vertex of the fuzzy set, indicating the most ideal state, that is, the membership reaches the maximum value; c is the right boundary of the fuzzy set, indicating that the fuzzy set ends with the maximum value of significant membership; among them, a, b, and c are all obtained from the scoring index items; x is the input value;

[0126] Substitute the numerical data in the obtained indicator item content as the input value x into the numerical scoring model, and determine and calculate the corresponding result based on the relationship between the input value x and a, b, and c, that is, the numerical score given in the scoring indicator item;

[0127] Substitute the vector value corresponding to the text data in the obtained index item content as the input value x into the numerical scoring model, and determine and calculate the corresponding result based on the relationship between the input value x and a, b, and c, that is, the text score given in the scoring index item;

[0128] Add the numerical score and the text score according to the preset weight to obtain the scoring index item score S corresponding to the scoring index item V ;

[0129] Among them, the weight of the numerical score is greater than the weight of the text score;

[0130] It should be noted that the fuzzy scoring model can be used to quantify fuzzy concepts to make them easier to understand and handle; in this model, we define a membership function, which represents the degree of membership of the input value corresponding to a specific fuzzy set;

[0131] For example:

[0132] Assume that a scoring indicator item is "project team experience" and the parameters of the fuzzy set are known to be as follows:

[0133] A = 2 years (left boundary);

[0134] b = 6 years (peak);

[0135] c = 10 years (right boundary);

[0136] If the average experience of the project team in a bidding document is 5 years, then the input value x = 5;

[0137] The numerical score calculated according to the fuzzy scoring model is 0.375;

[0138] Assume that the weight of the numerical score is 0.7;

[0139] Assume that the team experience represented by the vector value is described as "experienced and professional" and is converted into a vector value through TF-IDF. Assume that the value of the vector value x after fuzzification is 7.5;

[0140] The text score calculated according to the fuzzy scoring model is:

[0141] u(x)=1

[0142] Assume that the weight of the text score is 0.3;

[0143] Total score calculation:

[0144] S V =(0.375*0.7)+(1*0.3)=0.2625+0.3=0.5625

[0145] Therefore, the final score S of this scoring indicator item V is 0.5625;

[0146] In actual operation, the values ​​of a, b, and c need to be set according to the specific scoring indicators and the provisions in the bidding documents;

[0147] The weighting of numerical and textual data also needs to be determined according to the guidelines in the tender documents.

[0148] In order to further optimize the above embodiment, in step 4, the Bayesian scoring model uses Beta distribution as the prior distribution, and the Beta distribution is the conjugate prior of the binomial distribution, specifically:

[0149] S41, setting initial prior parameters α and β; the Bayesian scoring model includes two prior parameters α and β, which are set based on the importance of the scoring index item relative to the scoring index; the importance is quantified as a decimal between 0 and 1, and the values ​​of the parameters α and β are adjusted by the preset coefficients, and the larger the quantized value corresponding to the importance, the larger the parameter α and the smaller the parameter β, and vice versa;

[0150] S42, for each scoring index item score S Vi Update the prior parameters α and β;

[0151] The updated final prior parameter α i =α+S Vi ;

[0152] The updated final prior parameter β i =β+(1-S Vi );

[0153] Among them, S Vi is the score of the i-th scoring indicator item;

[0154] S43, calculating the expected value of the posterior distribution as the final score of the scoring indicator item;

[0155]

[0156] Among them, S Wi is the final score of the i-th scoring indicator item;

[0157] S44, setting weights based on the quantified values ​​of the importance of each scoring indicator item, and calculating the final score S of the scoring indicator. Fj ;

[0158]

[0159] Among them, S Fj is the final score of the jth scoring indicator, n is the number of scoring indicator items, φ i is the weight of the i-th scoring index item;

[0160] It should be noted that the Bayesian scoring model is a statistical method for updating our beliefs given prior knowledge; here, we use the Beta distribution as the prior distribution because it is the conjugate prior of the binomial distribution and is very suitable for handling probability estimates of binary outcomes (such as success or failure);

[0161] For example:

[0162] Suppose we are evaluating a scoring indicator that contains three scoring indicator items, namely A, B and C; suppose the initial prior parameters are α=1 and β=1;

[0163] 1. Set initial prior parameters:

[0164] Assume that the importance of scoring index item A is 0.8, the importance of scoring index item B is 0.5, and the importance of scoring index item C is 0.3;

[0165] Therefore, we can set α = 1 + 0.8*k, β = 1-0.8*k, where k is a preset coefficient used to adjust the impact of importance;

[0166] 2. Update prior parameters:

[0167] Assume that the score of rating indicator item A is 0.9, the score of rating indicator item B is 0.6, and the score of rating indicator item C is 0.4;

[0168] Updated prior parameter α i =1+0.8*k+0.9,β i =1-0.8*k+(1-0.9);

[0169] The scores of other indicators are similar;

[0170] 3. Calculate the expected value of the posterior distribution and the final score

[0171] The setting of initial a priori parameters needs to be determined according to specific circumstances, including the importance of the scoring indicators and the actual needs of the bid evaluation;

[0172] The weight of the scoring indicator items also needs to be set according to the importance of the scoring indicator items;

[0173] Through the above steps, we can use the Bayesian scoring model to calculate the final score of the scoring index items, and further calculate the final score of the scoring index, so as to provide scientific decision-making support for the bid evaluation process.

[0174] In order to further optimize the above embodiment, in step 5, the first weighted scoring model is:

[0175]

[0176] Among them, S Gk is the final score of the kth scoring factor, m is the number of scoring factors, ω j is the weight of the jth scoring indicator;

[0177] In step six, the second weighted scoring model is:

[0178]

[0179] Among them, S Qz is the final score of the zth scoring item, p is the number of scoring items, ε k is the weight of the kth scoring factor;

[0180] It should be noted that the first weighted scoring model means that the score differences will be directly accumulated, and the overall scoring result will be more averaged, that is, the extreme values ​​of individual scoring factors will not excessively affect the overall scoring result, that is, the score of the scoring factor;

[0181] The second weighted scoring model uses a multiplication algorithm, which means that if one scoring factor scores very low, it will have a very large negative impact on the final comprehensive score, that is, it will cause the score of the scoring item to drop significantly;

[0182] The above method can avoid multiple bidders having similar scores, making them easier to distinguish.

[0183] In order to further optimize the above embodiment, step six further includes:

[0184] After outputting the comprehensive scores, the comprehensive scores of the bidding documents are ranked;

[0185] It should be noted that ranking can help the bid evaluation team quickly determine the order of bid candidates and thus decide the winner. If there are bid documents with the same comprehensive score, additional criteria (such as price, innovation of technical solutions, etc.) can be considered for differentiation. In some cases, manual review of the ranking results may be required to ensure the accuracy and rationality of the ranking.

[0186] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for constructing an intelligent evaluation model, characterized in that: include: Step 1: extract the scoring index items, scoring indicators, scoring factors, and scoring items in the bidding documents, and establish a scoring tree model according to their relationships; Step 2: extracting the index item content corresponding to the scoring index item from the bidding document based on the scoring tree model and natural language processing technology; Step 3, matching and comparing the index item content of the bidding document with the scoring index items in the scoring tree model, and establishing a fuzzy scoring model based on the comparison results to calculate the scores of each scoring index item; Step 4: According to the topological structure of the scoring tree model and various scoring index items, a Bayesian scoring model is established to calculate the scores of various scoring index items; Step 5: Based on the topological structure of the scoring tree model, the scores of each scoring indicator and the preset indicator weights, a first weighted scoring model is established to calculate the scores of each scoring factor; Step 6: Based on the topological structure of the scoring tree model, the scores of each scoring factor and the preset factor weights, a second weighted scoring model is established to calculate the sum of the scores of each scoring item, i.e., the comprehensive score; Step seven, aggregating the scoring tree model, the fuzzy scoring model, the Bayesian model, the first weighted scoring model and the second weighted scoring model to generate an intelligent evaluation model.

2. The method for constructing an intelligent evaluation model according to claim 1, characterized in that: In step 1, the relationship among the scoring index items, scoring indexes, scoring factors, and scoring items includes: Scoring items are a classification of the main aspects or dimensions that need to be considered in the bid evaluation process, which include several scoring factors; Scoring factors are specific criteria or dimensions for specific scoring items, which are used to refine the content of the scoring items and include several scoring indicators. Scoring indicators are further refined evaluation criteria based on scoring factors, which are used to quantify or describe the requirements or expected levels of scoring factors. They include several scoring indicator items. The scoring index item is the specific measurement value or judgment condition under the scoring index, which is used to quantify the requirements or expected level of the scoring index and is the most basic evaluation unit in the bid evaluation.

3. The method for constructing an intelligent evaluation model according to claim 1, characterized in that: In step 2, extracting the index item content corresponding to the scoring index item from the bidding document includes: Extract and analyze keywords and semantics in bidding documents through natural language processing technology to obtain index item contents; For numerical data, extract the numerical value and generate numerical fuzzy sets according to the scoring index items in the bidding documents; For text data, TF-IDF technology is used to quantify the text and convert it into a vector value. At the same time, a text fuzzy set is generated according to the scoring index items in the bidding documents. The extracted indicator items are manually identified to eliminate errors in natural language processing technology.

4. The method for constructing an intelligent evaluation model according to claim 3, characterized in that: In step three, the fuzzy scoring model is divided into: Among them, μ(x) is the membership function; a is the left boundary of the fuzzy set, indicating that the fuzzy set has the minimum value of significant membership at the beginning; b is the vertex of the fuzzy set, indicating the most ideal state, that is, the membership reaches the maximum value; c is the right boundary of the fuzzy set, indicating that the fuzzy set ends with the maximum value of significant membership; among them, a, b, and c are all obtained from the scoring index items; x is the input value; Substituting the numerical data in the obtained indicator item content as the input value x into the numerical scoring model, and determining and calculating the corresponding result based on the relationship between the input value x and a, b, and c, that is, the numerical score given in the scoring indicator item; Substituting the vector value corresponding to the text data in the obtained index item content as the input value x into the numerical scoring model, and determining and calculating the corresponding result based on the relationship between the input value x and a, b, and c, that is, the text score given in the scoring index item; The numerical score and the text score are added according to the preset weight to obtain the scoring index item score S corresponding to the scoring index item. V ; Among them, the weight of the numerical score is greater than the weight of the text score.

5. The method for constructing an intelligent evaluation model according to claim 4, characterized in that: In step 4, the Bayesian scoring model uses Beta distribution as the prior distribution, which is the conjugate prior of the binomial distribution, specifically: S41, setting initial prior parameters α and β; the Bayesian scoring model includes two prior parameters α and β, which are set based on the importance of the scoring index item relative to the scoring index; the importance is quantified as a decimal between 0 and 1, and the values ​​of the parameters α and β are adjusted by a preset coefficient, and the larger the quantized value corresponding to the importance, the larger the parameter α and the smaller the parameter β, and vice versa; S42, for each scoring index item score S Vi Update the prior parameters α and β; The updated final prior parameter α i =α+S Vi ; The updated final prior parameter β i =β+(1-S Vi ), where S Vi is the score of the i-th scoring indicator item; S43, calculating the expected value of the posterior distribution as the final score of the scoring indicator item; Among them, S Wi is the final score of the i-th scoring indicator item; S44, setting weights based on the quantified values ​​of the importance of each scoring indicator item, and calculating the final score S of the scoring indicator. Fj ; Among them, S Fj is the final score of the jth scoring indicator, n is the number of scoring indicator items, is the weight of the i-th scoring indicator item.

6. The method for constructing an intelligent evaluation model according to claim 5, characterized in that: In step 5, the first weighted scoring model is: Among them, S Gk is the final score of the kth scoring factor, m is the number of scoring factors, ω j is the weight of the j-th scoring indicator.

7. The method for constructing an intelligent evaluation model according to claim 6, characterized in that: In step six, the second weighted scoring model is: Among them, S Qz is the final score of the zth scoring item, p is the number of scoring items, ε k is the weight of the kth scoring factor.

8. The method for constructing an intelligent evaluation model according to claim 7, characterized in that: In step six, it also includes: After outputting the comprehensive score, the comprehensive scores of each bidding document are sorted.

Images