Bid evaluation system and method based on machine learning

Abstract

The invention provides a bid evaluation system and method based on machine learning, and relates to the technical field of electronic bidding and tendering, and the system comprises a data acquisition module which collects historical bid evaluation data and carries out the data preprocessing, and obtains target bid evaluation data; the feature extraction module is used for extracting features in the target bid evaluation data and screening the features to obtain target features; the model construction module introduces a machine learning algorithm based on the target features to construct a target recommendation model; and the bid evaluation module is used for determining a bid evaluation standard and method most suitable for the current bid evaluation scene by using the target recommendation model, and generating a target report for output. Based on the target features obtained by feature extraction and screening after preprocessing of historical bid evaluation data, the machine learning algorithm is introduced to construct the target recommendation model so as to determine the bid evaluation standard and method most suitable for the current bid evaluation scene, and the target report is generated, so that the bid evaluation efficiency is effectively improved according to different bid evaluation environments. Personalized bid evaluation standards and method suggestions are provided, and the scientificity of bid evaluation is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of electronic bidding and tendering, and in particular to a bid evaluation system and method based on machine learning. Background Art

[0002] In recent years, with the rapid development of the Internet and information technology, electronic bid evaluation systems have gradually become popular, replacing traditional manual bid evaluation methods. This change not only simplifies the bid evaluation process, but also significantly improves the efficiency of bid evaluation work. However, the existing bid evaluation system has a single evaluation method and relatively fixed evaluation criteria, and cannot provide personalized evaluation criteria and method suggestions based on different bid evaluation environments.

[0003] Therefore, the present invention provides a bid evaluation system and method based on machine learning. Summary of the invention

[0004] The present invention provides a bid evaluation system and method based on machine learning, which is used to construct a target recommendation model by introducing a machine learning algorithm based on target features obtained by feature extraction and screening after preprocessing historical bid evaluation data, so as to determine the bid evaluation criteria and methods that are most suitable for the current bid evaluation scenario, and generate a target report, thereby effectively providing personalized bid evaluation criteria and method suggestions according to different bid evaluation environments, thereby improving the scientific nature of bid evaluation.

[0005] The present invention provides a bid evaluation system based on machine learning, comprising:

[0006] Data acquisition module: used to collect historical bid evaluation data and perform data preprocessing to obtain target bid evaluation data;

[0007] Feature extraction module: used to extract features from the target bid evaluation data and screen them to obtain target features;

[0008] Model building module: used to introduce machine learning algorithms to build a target recommendation model based on the target features;

[0009] Bid evaluation module: used to use the target recommendation model to determine the bid evaluation criteria and methods that are most suitable for the current bid evaluation scenario, and generate a target report output.

[0010] Preferably, data preprocessing refers to data cleaning, data conversion, and removal of duplicate and erroneous data items.

[0011] Preferably, the feature extraction module includes:

[0012] Feature extraction unit: divides the target bid evaluation data according to data types, and extracts features of the target bid evaluation data whose data type is text by using a set natural language processing technology to obtain a first feature;

[0013] The target bid evaluation data with numerical data type is normalized and output as the first feature;

[0014] For structured data, extract key fields as the first feature output;

[0015] For semi-structured data, regular expressions are introduced to extract features and output as the first feature;

[0016] Feature screening unit: used to screen target features from the first features based on set recommendation requirements.

[0017] Preferably, the feature screening unit comprises:

[0018] Preliminary screening block: used to screen out first candidate features from the first features according to set recommendation requirements;

[0019] The remaining first features except the first candidate feature are regarded as available features to be selected and are combined to obtain an available feature set;

[0020] Performing statistical correlation analysis on the available feature to be selected in the available feature set and the first candidate feature to obtain a first correlation coefficient;

[0021] Mark the available features to be selected whose first correlation coefficient is greater than the set correlation threshold as second candidate features;

[0022] Re-screening block: used to analyze the correlation between the first candidate feature, the second candidate feature and the bid evaluation result respectively, and obtain a first correlation coefficient;

[0023] Outputting the first candidate feature and the second candidate feature whose first correlation coefficient is greater than the set correlation threshold as target features;

[0024] The first candidate feature and the second candidate feature whose first correlation coefficient is not greater than the set correlation threshold are regarded as pending features;

[0025] Taking the pending feature as the screening condition, if the historical feature combination of the current pending feature is screened from the feature combination database, the current pending feature is marked as the first available feature, the corresponding historical feature combination is marked as the candidate feature combination, and the combined feature result of the historical feature combination is marked as the candidate combined feature;

[0026] Analyze whether the remaining features in the candidate feature combination except the available features are all pending features, and if so, calculate the first correlation coefficient between the corresponding candidate combination feature and the bid evaluation result;

[0027] Output the candidate combined features whose first correlation coefficient is greater than the set correlation threshold as the target feature;

[0028] Using the pending feature as a screening condition, if the historical feature derivation record of the current pending feature is screened from the feature derivation database, the current pending feature is marked as the second available feature;

[0029] According to the historical feature derivation record, a first derived feature is calculated based on the undetermined feature;

[0030] Calculating a first correlation coefficient between the first derived feature and the bid evaluation result;

[0031] The first derived feature whose first correlation coefficient is greater than the set association threshold is output as the target feature.

[0032] Preferably, the calculation formula of the first correlation coefficient is as follows:

[0033] In the formula, h represents the first correlation coefficient of the current candidate feature; a j It is represented as the observed value of the current candidate feature at the jth historical data point, where j = 1, 2, 3, ..., m, and m represents the total number of historical data points; It is expressed as the mean difference of the observations of the current candidate feature at all historical data points; b j It is represented as the observed value of the historical bid evaluation result at the jth historical data point; It is represented by the observed mean difference of the historical bid evaluation results at all historical data points; ω is represented by the important weight of the current candidate feature; x1 is represented by the number of occurrences of the current candidate feature in the historical bid evaluation data; x0 is represented by the number of occurrences of all candidate features in the historical bid evaluation data; ln is represented by the natural logarithm; e is represented by the base of the natural logarithm.

[0034] Preferably, the model building module includes:

[0035] Model building unit: used to label the target bid evaluation data and target features for bid evaluation scenarios and use them as training data to train the neural network to obtain an initial recommendation model;

[0036] Model optimization unit: used to evaluate the performance of the initial recommendation model using a test block; and optimize the initial recommendation model using an optimization block;

[0037] Test block: used to test the initial recommendation model using a predetermined test evaluation data set to obtain a test evaluation result;

[0038] According to the test evaluation result, evaluating the bidding performance of the current initial recommendation model based on the performance evaluation index, and obtaining a first performance index coefficient;

[0039] When there is a first performance indicator coefficient of the performance evaluation indicator that is not greater than a preset performance threshold, optimizing the initial recommendation model using an optimization block to obtain a target recommendation model;

[0040] When the first performance indicator coefficients of all performance evaluation indicators are greater than the preset performance threshold, the current initial recommendation model is used as the target recommendation model;

[0041] Optimization block: used to optimize the initial recommendation model to obtain the target recommendation model. The steps are as follows:

[0042] Step 11: Obtain the initial weight and initial threshold of the current initial recommendation model;

[0043] Step 12: Use the CIWOA algorithm to initialize the population, convert the initial weights and initial thresholds into the spatial position vectors of the individuals, and use the mean square error function of the initial recommendation model as the fitness function;

[0044] Step 13: Calculate the fitness values ​​of all individuals in the population, determine the individual with the current optimal fitness value, and record its position vector as the current optimal individual;

[0045] Step 14: Update the next generation position according to the target search strategy and calculate the current optimal individual fitness value;

[0046] Step 15: When the maximum number of iterations is reached, the optimization algorithm is stopped and the current optimal weight and optimal threshold are output;

[0047] Step 16: Assign the optimal weight and optimal threshold to the initial recommendation model to complete model optimization and generate a target recommendation model.

[0048] Preferably, the bid evaluation module includes:

[0049] Recommendation unit: used to obtain the characteristics of the current bid evaluation environment and input the target recommendation model, and output the target bid evaluation criteria and target bid evaluation method;

[0050] Report generation unit: used to generate target reports based on target evaluation criteria and target evaluation methods.

[0051] The present invention provides a bid evaluation method based on machine learning, characterized by comprising:

[0052] Step 1: Collect historical bid evaluation data and perform data preprocessing to obtain target bid evaluation data;

[0053] Step 2: extracting and screening the features in the target bid evaluation data to obtain target features;

[0054] Step 3: Based on the target features, introduce a machine learning algorithm to build a target recommendation model;

[0055] Step 4: Use the target recommendation model to determine the evaluation criteria and methods that are most suitable for the current evaluation scenario, and generate a target report output.

[0056] Compared with the prior art, the present invention has the following beneficial effects:

[0057] By extracting and screening the target features obtained by preprocessing the historical bid evaluation data, a machine learning algorithm is introduced to build a target recommendation model to determine the bid evaluation criteria and methods that are most suitable for the current bid evaluation scenario, and generate a target report. This effectively provides personalized bid evaluation criteria and method recommendations based on different bid evaluation environments, thereby improving the scientific nature of the bid evaluation.

[0058] 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.

[0059] The technical solution of the present invention is further described in detail below through the accompanying drawings and embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0060] 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:

[0061] Figure 1 is a structural diagram of a bid evaluation system based on machine learning in an embodiment of the present invention;

[0062] Figure 2 The present invention is a flowchart of a bid evaluation method based on machine learning in an embodiment of the present invention. DETAILED DESCRIPTION

[0063] 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.

[0064] The embodiment of the present invention provides a bid evaluation system based on machine learning, such as Figure 1 As shown, including:

[0065] Data acquisition module: used to collect historical bid evaluation data and perform data preprocessing to obtain target bid evaluation data;

[0066] Feature extraction module: used to extract features from the target bid evaluation data and screen them to obtain target features;

[0067] Model building module: used to introduce machine learning algorithms to build a target recommendation model based on the target features;

[0068] Bid evaluation module: used to use the target recommendation model to determine the bid evaluation criteria and methods that are most suitable for the current bid evaluation scenario, and generate a target report output.

[0069] In this embodiment, the historical bid evaluation data includes historical bid evaluation criteria, bid evaluation results, bid documents, and expert feedback, etc., wherein the historical bid evaluation criteria include the scoring items, scoring factors, and scoring indicators used in previous bid evaluations; the bid evaluation results include the results of previous bid evaluations, including the scores of the successful bidder and each bidder; the bid documents include part or all of the key information of the bid documents, such as technical solutions, prices, qualifications, etc.; expert feedback includes the feedback from the evaluation experts during the bid evaluation process, including satisfaction with the current bid evaluation criteria, improvement suggestions, etc.; data preprocessing refers to data cleaning, data conversion, and removal of duplicate and erroneous data items, the purpose of which is to ensure data quality and prepare for subsequent machine learning model training; the target bid evaluation data is the data obtained after data preprocessing of the historical bid evaluation data.

[0070] In this embodiment, the target features are features obtained by screening the features extracted from the target bid evaluation data, such as project type, technical solution, budget limit, etc.; machine learning algorithm refers to an important branch of artificial intelligence, which aims to enable machines to automatically acquire and organize knowledge by simulating the human learning process, such as linear regression; the target recommendation model is used to determine the bid evaluation criteria and bid evaluation methods that are most suitable for the current bid evaluation scenario; the target report is composed of the bid evaluation criteria, bid evaluation methods, detailed explanation of the scoring criteria, project overview, bid evaluation environment and report generation time that are most suitable for the current bid evaluation scenario determined by the target recommendation model; the bid evaluation environment includes technology procurement, service bidding, engineering construction, etc., and different bid evaluation environments have different bid evaluation criteria and methods.

[0071] The beneficial effect of the above technical solution is: by introducing a machine learning algorithm to construct a target recommendation model based on the target features obtained by feature extraction and screening after preprocessing the historical bid evaluation data, the bid evaluation criteria and methods that are most suitable for the current bid evaluation scenario are determined, and a target report is generated, which effectively provides personalized bid evaluation criteria and method recommendations based on different bid evaluation environments, thereby improving the scientific nature of the bid evaluation.

[0072] The embodiment of the present invention provides a bid evaluation system based on machine learning, wherein the feature extraction module comprises:

[0073] Feature extraction unit: divides the target bid evaluation data according to data types, and extracts features of the target bid evaluation data whose data type is text by using a set natural language processing technology to obtain a first feature;

[0074] The target bid evaluation data with numerical data type is normalized and output as the first feature;

[0075] For structured data, extract key fields as the first feature output;

[0076] For semi-structured data, regular expressions are introduced to extract features and output as the first feature;

[0077] Feature screening unit: used to screen target features from the first features based on set recommendation requirements.

[0078] In this embodiment, the target bid evaluation data is the data obtained after data preprocessing of the historical bid evaluation data, wherein the historical bid evaluation data includes historical bid evaluation criteria, bid evaluation results, bid documents, expert feedback, etc.; data preprocessing refers to data cleaning, data conversion, and removal of duplicate and erroneous data items, the purpose of which is to ensure data quality and prepare for subsequent machine learning model training; data types include text, numerical, and structured, etc.; the first feature refers to the feature extracted from the text type target bid evaluation data using the TF-IDF technology, the value after normalization of the numerical type target bid evaluation data, the key field extracted from the structured type target bid evaluation data, and the introduction of regular expressions to extract the features of the semi-structured type target bid evaluation data; setting natural language processing technology refers to TF-IDF technology; regular expression refers to a method of describing text patterns, which uses a specific syntax to define a string search pattern for extracting features of semi-structured data; setting recommended requirements refers to bid evaluation scenarios, bid evaluation criteria, and bid evaluation methods; the target feature is a feature screened from the first feature based on the set recommended requirements.

[0079] The beneficial effect of the above technical solution is that by extracting and screening features of target bid evaluation data, the accuracy of feature selection can be effectively improved, providing a solid foundation for subsequent model training and decision support.

[0080] An embodiment of the present invention provides a bid evaluation system based on machine learning, wherein the feature screening unit comprises:

[0081] Preliminary screening block: used to screen out first candidate features from the first features according to set recommendation requirements;

[0082] The remaining first features except the first candidate feature are regarded as available features to be selected and are combined to obtain an available feature set;

[0083] Performing statistical correlation analysis on the available feature to be selected in the available feature set and the first candidate feature to obtain a first correlation coefficient;

[0084] Mark the available features to be selected whose first correlation coefficient is greater than the set correlation threshold as second candidate features;

[0085] Re-screening block: used to analyze the correlation between the first candidate feature, the second candidate feature and the bid evaluation result respectively, and obtain a first correlation coefficient;

[0086] Outputting the first candidate feature and the second candidate feature whose first correlation coefficient is greater than the set correlation threshold as target features;

[0087] The first candidate feature and the second candidate feature whose first correlation coefficient is not greater than the set correlation threshold are regarded as pending features;

[0088] Taking the pending feature as the screening condition, if the historical feature combination of the current pending feature is screened from the feature combination database, the current pending feature is marked as the first available feature, the corresponding historical feature combination is marked as the candidate feature combination, and the combined feature result of the historical feature combination is marked as the candidate combined feature;

[0089] Analyze whether the remaining features in the candidate feature combination except the available features are all pending features, and if so, calculate the first correlation coefficient between the corresponding candidate combination feature and the bid evaluation result;

[0090] Output the candidate combined features whose first correlation coefficient is greater than the set correlation threshold as the target feature;

[0091] Using the pending feature as a screening condition, if the historical feature derivation record of the current pending feature is screened from the feature derivation database, the current pending feature is marked as the second available feature;

[0092] According to the historical feature derivation record, a first derived feature is calculated based on the undetermined feature;

[0093] Calculating a first correlation coefficient between the first derived feature and the bid evaluation result;

[0094] The first derived feature whose first correlation coefficient is greater than the set association threshold is output as the target feature.

[0095] In this embodiment, setting recommended requirements refers to evaluation scenarios, evaluation criteria and evaluation methods; the first candidate feature refers to a feature directly related to the set recommended requirements; the available feature to be selected refers to the first feature that is not the first candidate feature; the available feature set is a feature set obtained by combining the available features to be selected; the first correlation coefficient is used to characterize the degree of correlation between the available features to be selected and the first candidate features, and is obtained by calculating the Pearson correlation coefficient; the set correlation threshold is pre-set, generally 0.7; the second candidate feature refers to the available feature to be selected whose first correlation coefficient is greater than the set correlation threshold.

[0096] In this embodiment, the first correlation coefficient is used to characterize the degree of correlation between the feature and the bid evaluation result; the setting of the correlation threshold is predetermined, generally 0.7; the pending feature refers to the first candidate feature or the second candidate feature whose first correlation coefficient is not greater than the set correlation threshold; the feature combination database is composed of historical feature combinations and combined features obtained by using feature combinations in historical feature combinations; the first available feature refers to the pending feature corresponding to the historical feature combination in the feature combination database; the candidate feature combination refers to the historical feature combination of the pending feature existing in the feature combination database; the candidate combination feature refers to the feature obtained by combining all the features in the candidate feature combination, such as combining the bidding time and the project start time to generate a time feature.

[0097] In this embodiment, the feature derivation database is composed of feature derivation records; the historical feature derivation records are composed of features, historical feature derivation calculation results and derivation time; the second available feature refers to the pending feature of the historical feature derivation record in the feature derivation database; the first derived feature refers to the new statistics and ratios calculated based on the pending feature according to the historical feature derivation record, such as the discount rate of the bid quotation and the deviation from the average quotation.

[0098] The beneficial effect of the above technical solution is that by performing two screenings from the acquired first features to obtain the target features, the accuracy of feature selection can be improved, which in turn helps to improve the accuracy of subsequent model building.

[0099] The embodiment of the present invention provides a bid evaluation system based on machine learning, and the calculation formula of the first correlation coefficient is as follows:

[0100] In the formula, h represents the first correlation coefficient of the current candidate feature; a j It is represented as the observed value of the current candidate feature at the jth historical data point, where j = 1, 2, 3, ..., m, and m represents the total number of historical data points; It is expressed as the mean difference of the observations of the current candidate feature at all historical data points; b j It is represented as the observed value of the historical bid evaluation result at the jth historical data point; It is represented by the observed mean difference of the historical bid evaluation results at all historical data points; ω is represented by the important weight of the current candidate feature; x1 is represented by the number of occurrences of the current candidate feature in the historical bid evaluation data; x0 is represented by the number of occurrences of all candidate features in the historical bid evaluation data; ln is represented by the natural logarithm; e is represented by the base of the natural logarithm.

[0101] The beneficial effect of the above technical solution is that by calculating the first correlation coefficient, an effective data basis can be provided for screening features, thereby improving the accuracy of feature selection, and thus helping to ensure the accuracy of subsequent model establishment.

[0102] The embodiment of the present invention provides a bid evaluation system based on machine learning, wherein the model building module comprises:

[0103] Model building unit: used to label the target bid evaluation data and target features for bid evaluation scenarios and use them as training data to train the neural network to obtain an initial recommendation model;

[0104] Model optimization unit: used to evaluate the performance of the initial recommendation model using a test block; and optimize the initial recommendation model using an optimization block;

[0105] Test block: used to test the initial recommendation model using a predetermined test evaluation data set to obtain a test evaluation result;

[0106] According to the test evaluation result, evaluating the bidding performance of the current initial recommendation model based on the performance evaluation index, and obtaining a first performance index coefficient;

[0107] When there is a first performance indicator coefficient of the performance evaluation indicator that is not greater than a preset performance threshold, optimizing the initial recommendation model using an optimization block to obtain a target recommendation model;

[0108] When the first performance indicator coefficients of all performance evaluation indicators are greater than the preset performance threshold, the current initial recommendation model is used as the target recommendation model;

[0109] Optimization block: used to optimize the initial recommendation model to obtain the target recommendation model. The steps are as follows:

[0110] Step 11: Obtain the initial weight and initial threshold of the current initial recommendation model;

[0111] Step 12: Use the CIWOA algorithm to initialize the population, convert the initial weights and initial thresholds into the spatial position vectors of the individuals, and use the mean square error function of the initial recommendation model as the fitness function;

[0112] Step 13: Calculate the fitness values ​​of all individuals in the population, determine the individual with the current optimal fitness value, and record its position vector as the current optimal individual;

[0113] Step 14: Update the next generation position according to the target search strategy and calculate the current optimal individual fitness value;

[0114] Step 15: When the maximum number of iterations is reached, the optimization algorithm is stopped and the current optimal weight and optimal threshold are output;

[0115] Step 16: Assign the optimal weight and optimal threshold to the initial recommendation model to complete model optimization and generate a target recommendation model.

[0116] In this embodiment, the target feature is a feature obtained by screening the first feature, wherein the first feature refers to the feature extracted from the text type target bid evaluation data using the TF-IDF technology, the value after normalization of the numerical type target bid evaluation data, the key field extracted from the structured type target bid evaluation data, and the feature extracted from the semi-structured type target bid evaluation data by introducing regular expressions; the initial recommendation model is a model obtained by training the neural network using the target bid evaluation data and the target feature annotated with the bid evaluation scenario as training data; the test bid evaluation data set refers to a predetermined data set that does not overlap with the training data; the test evaluation result refers to the result output after the test bid evaluation data set is input into the initial recommendation model.

[0117] In this embodiment, the performance evaluation index refers to accuracy, recall rate and F1 score; the first performance index coefficient refers to the value obtained by evaluating the test results using the performance evaluation index; the preset performance threshold is predetermined; the target recommendation model is used to determine the evaluation criteria and evaluation method that are most suitable for the current evaluation scenario; the initialization population refers to the initialization population size and the maximum number of iterations; the target search strategy refers to the search strategy of the CIWOA algorithm.

[0118] The beneficial effect of the above technical solution is: by performing scene annotation on the target evaluation data and target features, and using them as training data to train the neural network to obtain a model for performance testing and optimization to obtain a target recommendation model, the performance of the model can be effectively improved, and the accuracy of the evaluation criteria and methods recommended according to different evaluation environments can be ensured.

[0119] An embodiment of the present invention provides a bid evaluation system based on machine learning, wherein the bid evaluation module comprises:

[0120] Recommendation unit: used to obtain the characteristics of the current bid evaluation environment and input the target recommendation model, and output the target bid evaluation criteria and target bid evaluation method;

[0121] Report generation unit: used to generate target reports based on target evaluation criteria and target evaluation methods.

[0122] In this embodiment, the characteristics of the bid evaluation environment include the nature, scale, technical requirements, budget constraints and market conditions of the project; the target report is composed of the bid evaluation criteria, bid evaluation methods, detailed explanation of the scoring criteria, project overview, bid evaluation environment and report generation time that are most suitable for the current bid evaluation scenario determined by the target recommendation model.

[0123] The beneficial effect of the above technical solution is: by inputting the characteristics of the current bid evaluation environment into the target recommendation model, outputting the target bid evaluation criteria and target bid evaluation method and generating a report, it is effectively realized according to different bid evaluation environments, providing personalized bid evaluation criteria and method suggestions, and improving the scientific nature of the bid evaluation.

[0124] The embodiment of the present invention provides a bid evaluation method based on machine learning, characterized by comprising:

[0125] Step 1: Collect historical bid evaluation data and perform data preprocessing to obtain target bid evaluation data;

[0126] Step 2: extracting and screening the features in the target bid evaluation data to obtain target features;

[0127] Step 3: Based on the target features, introduce a machine learning algorithm to build a target recommendation model;

[0128] Step 4: Use the target recommendation model to determine the evaluation criteria and methods that are most suitable for the current evaluation scenario, and generate a target report output.

[0129] The beneficial effect of the above technical solution is: by introducing a machine learning algorithm to construct a target recommendation model based on the target features obtained by feature extraction and screening after preprocessing the historical bid evaluation data, the bid evaluation criteria and methods that are most suitable for the current bid evaluation scenario are determined, and a target report is generated, which effectively provides personalized bid evaluation criteria and method recommendations based on different bid evaluation environments, thereby improving the scientific nature of the bid evaluation.

[0130] Obviously, those skilled in the art can make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalents, the present invention is also intended to include these modifications and variations.

Claims

1. A bid evaluation system based on machine learning, characterized in that: include: Data acquisition module: used to collect historical bid evaluation data and perform data preprocessing to obtain target bid evaluation data; Feature extraction module: used to extract features from the target bid evaluation data and screen them to obtain target features; Model building module: used to introduce machine learning algorithms to build a target recommendation model based on the target features; Bid evaluation module: used to use the target recommendation model to determine the bid evaluation criteria and methods that are most suitable for the current bid evaluation scenario, and generate a target report output.

2. The bid evaluation system based on machine learning according to claim 1, characterized in that: Data preprocessing refers to data cleaning, data conversion, and removal of duplicate and erroneous data items.

3. The bid evaluation system based on machine learning according to claim 1, characterized in that: The feature extraction module comprises: Feature extraction unit: dividing the target bid evaluation data according to data types, and extracting features of the target bid evaluation data whose data type is text by using a set natural language processing technology to obtain a first feature; The target bid evaluation data with numerical data type is normalized and output as the first feature; For structured data, extract key fields as the first feature output; For semi-structured data, regular expressions are introduced to extract features and output as the first feature; Feature screening unit: used to screen target features from the first features based on set recommendation requirements.

4. The bid evaluation system based on machine learning according to claim 3, characterized in that: The feature screening unit comprises: Preliminary screening block: used to screen out first candidate features from the first features according to set recommendation requirements; The remaining first features except the first candidate feature are regarded as available features to be selected and are combined to obtain an available feature set; Performing statistical correlation analysis on the available feature to be selected in the available feature set and the first candidate feature to obtain a first correlation coefficient; Mark the available features to be selected whose first correlation coefficient is greater than the set correlation threshold as second candidate features; Re-screening block: used to analyze the correlation between the first candidate feature, the second candidate feature and the bid evaluation result respectively, and obtain a first correlation coefficient; Outputting the first candidate feature and the second candidate feature whose first correlation coefficient is greater than the set correlation threshold as target features; The first candidate feature and the second candidate feature whose first correlation coefficient is not greater than the set correlation threshold are regarded as pending features; Taking the pending feature as the screening condition, if the historical feature combination of the current pending feature is screened from the feature combination database, the current pending feature is marked as the first available feature, the corresponding historical feature combination is marked as the candidate feature combination, and the combined feature result of the historical feature combination is marked as the candidate combined feature; Analyze whether the remaining features in the candidate feature combination except the available features are all pending features, and if so, calculate the first correlation coefficient between the corresponding candidate combination feature and the bid evaluation result; Output the candidate combined features whose first correlation coefficient is greater than the set correlation threshold as the target feature; Using the pending feature as a screening condition, if the historical feature derivation record of the current pending feature is screened from the feature derivation database, the current pending feature is marked as the second available feature; According to the historical feature derivation record, a first derived feature is calculated based on the undetermined feature; Calculating a first correlation coefficient between the first derived feature and the bid evaluation result; The first derived feature whose first correlation coefficient is greater than the set association threshold is output as the target feature.

5. The bid evaluation system based on machine learning according to claim 4, characterized in that: The calculation formula of the first correlation coefficient is as follows: In the formula, h represents the first correlation coefficient of the current candidate feature; a j It is represented as the observed value of the current candidate feature at the jth historical data point, where j = 1, 2, 3, ..., m, and m represents the total number of historical data points; It is expressed as the mean difference of the observations of the current candidate feature at all historical data points; b j It is represented as the observed value of the historical bid evaluation result at the jth historical data point; It is represented by the observed mean difference of the historical bid evaluation results at all historical data points; ω is represented by the important weight of the current candidate feature; x1 is represented by the number of occurrences of the current candidate feature in the historical bid evaluation data; x0 is represented by the number of occurrences of all candidate features in the historical bid evaluation data; ln is represented by the natural logarithm; e is represented by the base of the natural logarithm.

6. The bid evaluation system based on machine learning according to claim 1, characterized in that: The model building module includes: Model building unit: used to label the target bid evaluation data and target features for bid evaluation scenarios and use them as training data to train the neural network to obtain an initial recommendation model; Model optimization unit: used to evaluate the performance of the initial recommendation model using a test block; and optimize the initial recommendation model using an optimization block; Test block: used to test the initial recommendation model using a predetermined test evaluation data set to obtain a test evaluation result; According to the test evaluation result, evaluating the bidding performance of the current initial recommendation model based on the performance evaluation index, and obtaining a first performance index coefficient; When there is a first performance indicator coefficient of the performance evaluation indicator that is not greater than a preset performance threshold, optimizing the initial recommendation model using an optimization block to obtain a target recommendation model; When the first performance indicator coefficients of all performance evaluation indicators are greater than the preset performance threshold, the current initial recommendation model is used as the target recommendation model; Optimization block: used to optimize the initial recommendation model to obtain the target recommendation model. The steps are as follows: Step 11: Obtain the initial weight and initial threshold of the current initial recommendation model; Step 12: Use the CIWOA algorithm to initialize the population, convert the initial weights and initial thresholds into the spatial position vectors of the individuals, and use the mean square error function of the initial recommendation model as the fitness function; Step 13: Calculate the fitness values ​​of all individuals in the population, determine the individual with the current optimal fitness value, and record its position vector as the current optimal individual; Step 14: Update the next generation position according to the target search strategy and calculate the current optimal individual fitness value; Step 15: When the maximum number of iterations is reached, the optimization algorithm is stopped and the current optimal weight and optimal threshold are output; Step 16: Assign the optimal weight and optimal threshold to the initial recommendation model to complete model optimization and generate a target recommendation model.

7. The bid evaluation system based on machine learning according to claim 1, characterized in that: The bid evaluation module includes: Recommendation unit: used to obtain the characteristics of the current bid evaluation environment and input the target recommendation model, and output the target bid evaluation criteria and target bid evaluation method; Report generation unit: used to generate target reports based on target evaluation criteria and target evaluation methods.

8. A bid evaluation method based on machine learning, characterized in that: include: Step 1: Collect historical bid evaluation data and perform data preprocessing to obtain target bid evaluation data; Step 2: extracting and screening the features in the target bid evaluation data to obtain target features; Step 3: Based on the target features, introduce a machine learning algorithm to build a target recommendation model; Step 4: Use the target recommendation model to determine the evaluation criteria and methods that are most suitable for the current evaluation scenario, and generate a target report output.

Images