Liquidation data intelligent analysis method and system based on market evaluation

Abstract

The present invention discloses a method and system for intelligent analysis of liquidation data based on market evaluation, and relates to the field of liquidation data technology. The present invention introduces an optimized threshold comprehensive score by combining an enhanced quality factor with an existing comprehensive score, and can dynamically adjust the selection criteria of a data set used for model optimization. Specifically, only when the first comprehensive score of a newly acquired data set is significantly greater than the optimized threshold comprehensive score will it be adopted for model optimization. This strict standard can avoid frequent retraining due to close data quality, thereby effectively reducing unnecessary computing resource consumption and time costs, and effectively reducing the risk of overfitting during model training. Therefore, on the premise of effectively improving the accuracy of the prediction results output by the optimized and trained machine learning model, more accurate liquidation decisions can be made, thereby effectively reducing the impact on the financial health of the enterprise.

Description

Technical Field

[0001] The present invention relates to the field of liquidation data technology, and in particular to a method and system for intelligent analysis of liquidation data based on market evaluation. Background Art

[0002] In modern financial markets and corporate asset management, intelligent analysis of liquidation data is crucial for accurately assessing asset values, optimizing liquidation processes, and maximizing recovery rates. Existing intelligent analysis methods for liquidation data typically rely on multiple data sources and technical means, aiming to provide reliable market assessment results through a systematic process. The following are the basic implementation steps of existing methods:

[0003] First, relevant liquidation data is obtained from multiple public and private data sources, such as financial statements, transaction records, social media data, and market assessment data such as market reports. The relevant liquidation data includes but is not limited to user comments, geographic location information, etc., and then the data in different formats are preliminarily integrated to form a unified data set. This step involves data format conversion, time series alignment, and consistency of data identifiers for subsequent processing. The integrated data set is then preprocessed, usually including noise removal, outlier detection, standardization, etc. Then, a suitable machine learning model is selected, usually a neural network model. At the same time, the data set is divided into a training set and a validation set, and it is necessary to ensure that the data distribution of each part is consistent. The selected machine learning model is then trained with the training set and its performance is evaluated on the validation set. The evaluation process adopts cross-validation, and in each cross-validation process, the precision and recall on each validation set are calculated separately, and the F1 score of each validation set is calculated by combining the precision and recall on each validation set. Finally, the F1 scores on all validation sets are averaged to obtain the performance of the current machine learning model, especially for the current training machine The final F1 score of the effectiveness of the entire data set when learning the model is then calculated by calculating the completeness score, distribution score, noise detection score and annotation quality score of the entire data set, and combining them with the final F1 score. During the combination process, normalization is ensured to be completed, so as to calculate the comprehensive score of the entire data set used to train the machine learning model, and use the comprehensive score to measure the quality of the current entire data set. If the comprehensive score of the current entire data set does not meet the preset threshold, a new batch of data will be obtained, and under the premise of reducing the F1 score, the completeness score, distribution score, noise detection score and annotation quality score of the newly obtained data set are calculated and combined to obtain the comprehensive score of the newly obtained data set. Then, it is measured whether the comprehensive score is greater than the comprehensive score. If it is greater, the newly obtained data set is used to optimize the training of the current pre-trained machine learning model. Otherwise, the current newly obtained data set is not adopted until the final optimized and trained machine learning model is obtained. Then, the final optimized and trained machine learning model can be put into use, that is, the trained model is used to predict the new clearing data and generate corresponding prediction results.

[0004] Although this method continuously evaluates the quality of the newly acquired dataset and compares it with the quality of the previous dataset to ensure that the dataset used when the model is updated has a high quality, thereby helping to improve the overall performance of the model, the existing implementation method often ignores a more serious problem, namely the risk of overfitting during repeated optimization training of the model. The main reason for this risk is that the existing method only considers when the comprehensive score is greater than the comprehensive score, then immediately uses the newly acquired dataset to optimize the training of the prediction model, ignoring the situation when the comprehensive score is only slightly greater than the comprehensive score. This situation proves that the quality of the newly acquired dataset is close enough to the quality of the dataset used when the prediction model was initially trained. Therefore, in this case, when the newly acquired dataset is still used to optimize the training of the prediction model, the risk of model overfitting is very likely to occur. When this risk occurs, it is easy to cause the accuracy of the prediction results output by the trained machine learning model to decrease, which can easily lead to incorrect liquidation decisions, thereby affecting the financial health of the company.

[0005] Therefore, the existing technology urgently needs a technical solution for a liquidation data intelligent analysis method and system based on market evaluation. Summary of the Invention

[0006] In order to solve the above technical problems, the present invention provides a method for intelligent analysis of liquidation data based on market evaluation, which specifically includes the following steps:

[0007] Step S1: Obtain at least two new data sets, and calculate a first comprehensive score for each new data set;

[0008] Step S1a: Calculate the completeness score, distribution score, noise detection score, and annotation quality score of each new dataset and combine them to obtain a comprehensive score for each new dataset;

[0009] Step S1b, calculating the data consistency score of each new data set;

[0010] Step S1b1, calculating the mean of all data points in each set of new data sets to obtain the mean of each set of new data sets;

[0011] Step S1b2: Calculate the standard deviation of each new data set based on the mean of each new data set;

[0012] Step S1b3: Calculate the data consistency score of each new data set based on the mean and standard deviation of each new data set;

[0013] The calculation formula for obtaining the data consistency score of each new data set is:

[0014] ;

[0015] Where, represents the data consistency score of the i-th group of new data sets; Represents the total number of data points in the i-th group of new data sets; Represents the jth data point in the i-th group of new data sets; Represents the mean of the i-th group of new data sets; Represents the standard deviation of the i-th group of new data sets;

[0016] Step S1c: combining the comprehensive score of each new data set with the data consistency score of each new data set to obtain a first comprehensive score for each new data set;

[0017] The calculation formula for obtaining the first comprehensive score of each new data set is:

[0018] ;

[0019] Where, Represents the first comprehensive score of the i-th group of new data sets; Represents the comprehensive score of the i-th group of new data sets; represents the data consistency score of the i-th group of new data sets; Represents the first weight coefficient, which is used to balance the contribution of the comprehensive score and the data consistency score, and 0< <1;

[0020] Step S2: calling the comprehensive score, and calculating the optimized threshold comprehensive score based on the comprehensive score and the first comprehensive score;

[0021] Step S2a, calculating the difference between the first comprehensive score and the comprehensive score of each new data set;

[0022] Step S2b: Calculate an enhanced quality factor based on the difference between the first comprehensive score and the comprehensive score of each new data set;

[0023] Step S2b1: based on the calculation result of the difference between the first comprehensive score and the comprehensive score of each new data set, retain the differences with positive calculation results and eliminate the differences with negative calculation results;

[0024] Step S2b2, counting the number of differences whose calculation results are positive numbers;

[0025] Step S2b3: Based on the statistical results, obtain the maximum value of the differences whose calculated results are positive numbers;

[0026] Step S2b4: Calculate the enhancement quality factor based on the difference values ​​whose calculation results are positive and the maximum value of the difference values ​​whose calculation results are positive;

[0027] Among them, the calculation formula for the enhanced quality factor is:

[0028] ;

[0029] Where, represents the enhancement quality factor; Represents the number of differences that evaluate to positive numbers; represents the kth difference whose calculation result is a positive number; Represents the maximum value among the differences whose calculation results are positive numbers;

[0030] Step S2c: combining the enhanced quality factor with the comprehensive score to obtain an optimized threshold comprehensive score;

[0031] Among them, the calculation formula for the optimized threshold comprehensive score is:

[0032] ;

[0033] Where, represents the optimized threshold comprehensive score; represents the comprehensive score; represents the enhancement quality factor; Represents the second weight coefficient, which is used to balance the contribution of the comprehensive score and the enhanced quality factor, and 0< <1;

[0034] Step S3: Reacquire a new dataset, and calculate the first comprehensive score of the current new dataset according to the calculation principle of the first comprehensive score. At the same time, determine the first comprehensive score of the current new dataset according to the optimized threshold comprehensive score. If the first comprehensive score of the current new dataset is greater than the optimized threshold comprehensive score, retain the current new dataset; if the first comprehensive score of the current new dataset is less than or equal to the optimized threshold comprehensive score, remove it, and acquire another new dataset until it is retained.

[0035] Step S4: Based on the judgment result, select the retained new data set to optimize the pre-trained machine learning model to obtain an optimized and trained machine learning model, and use the optimized and trained machine learning model to predict the input liquidation data, output the prediction result, and make a liquidation decision based on the prediction result.

[0036] The market evaluation-based liquidation data intelligent analysis system implements the above-mentioned market evaluation-based liquidation data intelligent analysis method and includes the following modules:

[0037] A first comprehensive score calculation module: configured to obtain at least two new data sets and calculate a first comprehensive score for each new data set;

[0038] Optimization threshold comprehensive score calculation module: connected to the first comprehensive score calculation module, used to call the comprehensive score, and calculate the optimization threshold comprehensive score based on the comprehensive score and the first comprehensive score;

[0039] New data set determination module: connected to the optimization threshold comprehensive score calculation module, used to re-acquire a new data set, and calculate the first comprehensive score of the current new data set according to the calculation principle of the first comprehensive score, and at the same time determine the first comprehensive score of the current new data set according to the optimization threshold comprehensive score. If the first comprehensive score of the current new data set is greater than the optimization threshold comprehensive score, the current new data set is retained; if the first comprehensive score of the current new data set is less than or equal to the optimization threshold comprehensive score, it is eliminated and a new data set is obtained again until it is retained;

[0040] Model optimization and decision-making module: connected to the new data set judgment module, used to select the retained new data set based on the judgment results to optimize the pre-trained machine learning model, obtain the optimized and trained machine learning model, and use the optimized and trained machine learning model to predict the input liquidation data, output the prediction results, and make liquidation decisions based on the prediction results.

[0041] The embodiments of the present invention have the following technical effects:

[0042] The present invention introduces an optimized threshold comprehensive score by combining the enhanced quality factor with the existing comprehensive score, which can dynamically adjust the selection criteria of the data set used for model optimization. Specifically, only when the first comprehensive score of the newly acquired data set is significantly greater than the optimized threshold comprehensive score will it be adopted for model optimization. This strict standard can avoid frequent retraining due to close data quality, thereby not only effectively reducing unnecessary computing resource consumption and time costs, but also effectively reducing the risk of overfitting during model training, so that more accurate liquidation decisions can be made on the premise of effectively improving the accuracy of the prediction results output by the optimized and trained machine learning model, thereby effectively reducing the impact on the financial health of the enterprise. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] In order to more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the specific embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0044] Figure 1 is a flow chart of a method for intelligent analysis of liquidation data based on market evaluation provided by an embodiment of the present invention;

[0045] Figure 2 This is a framework diagram of a liquidation data intelligent analysis system based on market evaluation provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0046] To make the objectives, technical solutions, and advantages of the present invention more clear, the technical solutions of the present invention are described clearly and completely below. Obviously, the embodiments described are only some of the embodiments of the present invention, not all of them. All other embodiments derived by persons of ordinary skill in the art based on the embodiments of the present invention without inventive effort are also within the scope of protection of the present invention.

[0047] Example 1: Figure 1 As shown, the present invention provides a method for intelligent analysis of liquidation data based on market evaluation, comprising the following steps:

[0048] Step S1: Obtain at least two new data sets, and calculate a first comprehensive score for each new data set;

[0049] Step S1a: Calculate the completeness score, distribution score, noise detection score, and annotation quality score of each new dataset and combine them to obtain a comprehensive score for each new dataset;

[0050] It is worth noting that when calculating the completeness score of each new data set, we usually first obtain the number of missing values ​​in each new data set and count the total number of data points in each new data set, and then use the formula , to calculate the completeness score of each new data set, where, represents the completeness score of the i-th group of new data sets; Represents the number of missing values ​​in the i-th group of new data sets; Represents the total number of data points in the i-th group of new data sets. It is worth noting that the actual meaning of this formula is to measure the proportion of missing values ​​in the data set. The closer it is to 1, the more complete the data is.

[0051] When calculating the distribution score of each new data set, the test statistic is usually calculated first, that is, the maximum difference between the data set and the standard distribution. The calculation formula is , where Represents the test statistic, which is the maximum difference between the data set and the standard distribution; Represents the empirical distribution function of the i-th group of new data sets; Represents the total number of data points in the i-th group of new data sets; Represents the standard distribution function; x represents the value of the data point in the new data set, that is, the specific value or observation value; then the formula is used , to calculate the distribution score of each new data set, where Represents the distribution score of the i-th group of new data sets; Represents the test statistic, that is, the maximum difference between the data set and the standard distribution. It is worth noting that the practical significance of this formula is to measure the similarity between the empirical distribution of the data set and the standard distribution. The closer it is to 1, the more uniform the distribution.

[0052] When calculating the noise detection score of each new data set, the number of abnormal data and the total number of data in each new data set are counted, and then the formula is used. , to calculate the noise detection score of each new data set, where, represents the noise detection score of the i-th new data set; Represents the number of abnormal data in the i-th group of new data sets; Represents the total number of data points in the i-th group of new data sets. It is worth noting that the actual meaning of this formula is to measure the proportion of outliers in the data set. The closer it is to 1, the cleaner the data.

[0053] When calculating the annotation quality score of each new data set, the number of correct annotations and the total number of annotations in each new data set are counted, and then the formula is used. , to calculate the annotation quality score of each new dataset, where Represents the annotation quality score of the i-th group of new datasets; Represents the number of correct annotations in the i-th group of new data sets; Represents the total number of annotations in the i-th new dataset. It is worth noting that the actual significance of this formula is to measure the accuracy of the annotations in the dataset. The closer it is to 1, the more accurate the annotations.

[0054] Then, by combining the above completeness score, distribution score, noise detection score, and annotation quality score, the comprehensive score of each new dataset is calculated. The specific calculation formula is as follows: ;

[0055] Where, Represents the comprehensive score of the i-th group of new data sets; represents the completeness score of the i-th group of new data sets; Represents the distribution score of the i-th group of new data sets; represents the noise detection score of the i-th new data set; Represents the annotation quality score of the i-th group of new datasets; 、 、 and Represent the weight coefficients of the completeness score, distribution score, noise detection score, and annotation quality score, respectively. It is worth noting that the practical significance of this formula is that it provides a comprehensive dataset quality assessment indicator by weighted averaging the scores of multiple dimensions (completeness, distribution, noise, and annotation quality). This helps ensure that the dataset used for model training has a high basic quality.

[0056] It is worth further explaining that the independent calculation of each scoring dimension helps to accurately evaluate different aspects of the dataset.

[0057] Step S1b, calculating the data consistency score of each new data set;

[0058] It is worth noting that the data consistency score effectively supplements the internal data consistency and stability that are ignored in the comprehensive score.

[0059] Step S1b1, calculating the mean of all data points in each set of new data sets to obtain the mean of each set of new data sets;

[0060] Step S1b2: Calculate the standard deviation of each new data set based on the mean of each new data set;

[0061] Step S1b3: Calculate the data consistency score of each new data set based on the mean and standard deviation of each new data set;

[0062] The calculation formula for obtaining the data consistency score of each new data set is:

[0063] ;

[0064] Where, represents the data consistency score of the i-th group of new data sets; Represents the total number of data points in the i-th group of new data sets; Represents the jth data point in the i-th group of new data sets; Represents the mean of the i-th group of new data sets; Represents the standard deviation of the i-th group of new data sets;

[0065] It is worth noting that by calculating the data consistency score of each new data set, the quality of the new data set can be further scientifically measured on the existing basis. Specifically, the data consistency score can effectively measure the consistency and stability within the new data set, and by calculating the data consistency score, the statistical consistency of each data point in the new data set can be ensured, thereby improving the stability and accuracy of subsequent model optimization training. It is worth further explaining that the practical significance of this formula is to measure the consistency and stability within the data set. The closer it is to 1, the more consistent the data.

[0066] Step S1c: combining the comprehensive score of each new data set with the data consistency score of each new data set to obtain a first comprehensive score for each new data set;

[0067] The calculation formula for obtaining the first comprehensive score of each new data set is:

[0068] ;

[0069] Where, Represents the first comprehensive score of the i-th group of new data sets; Represents the comprehensive score of the i-th group of new data sets; represents the data consistency score of the i-th group of new data sets; Represents the first weight coefficient, which is used to balance the contribution of the comprehensive score and the data consistency score, and 0< <1;

[0070] It is worth noting that by combining the comprehensive score with the data consistency score, the quality of the new dataset can be evaluated more comprehensively. The data consistency score supplements the internal data consistency and stability ignored in the comprehensive score, thereby improving the accuracy and reliability of the overall evaluation. It is worth noting that the setting of the first weight coefficient is usually determined through experiments to determine the optimal value, for example, by cross-validation method, trying different value and select the one that optimizes the performance of the subsequent machine learning model value.

[0071] Step S2: calling the comprehensive score, and calculating the optimized threshold comprehensive score based on the comprehensive score and the first comprehensive score;

[0072] It is worth noting that by introducing enhanced quality factors to dynamically adjust and optimize the threshold comprehensive score, it can effectively adapt to the ever-changing data environment.

[0073] Step S2a, calculating the difference between the first comprehensive score and the comprehensive score of each new data set;

[0074] Step S2b: Calculate an enhanced quality factor based on the difference between the first comprehensive score and the comprehensive score of each new data set;

[0075] Step S2b1: based on the calculation result of the difference between the first comprehensive score and the comprehensive score of each new data set, retain the differences with positive calculation results and eliminate the differences with negative calculation results;

[0076] It is worth noting that retaining the differences with positive calculation results helps to screen out new data sets with higher quality than the benchmark data sets. The benchmark data sets refer to the data sets used to train machine learning models in the existing technology. Eliminating negative differences can effectively avoid introducing new data sets of poor quality, thereby improving the stability and generalization ability of subsequent model optimization training.

[0077] Step S2b2, counting the number of differences whose calculation results are positive numbers;

[0078] Step S2b3: Based on the statistical results, obtain the maximum value of the differences whose calculated results are positive numbers;

[0079] Step S2b4: Calculate the enhancement quality factor based on the difference values ​​whose calculation results are positive and the maximum value of the difference values ​​whose calculation results are positive;

[0080] It is worth noting that the enhanced quality factor can quantify the degree of quality improvement of the new dataset relative to the benchmark dataset, thereby helping to better select high-quality new datasets.

[0081] Among them, the calculation formula for the enhanced quality factor is:

[0082] ;

[0083] Where, represents the enhancement quality factor; Represents the number of differences that evaluate to positive numbers; represents the kth difference whose calculation result is a positive number; Represents the maximum value among the differences whose calculation results are positive numbers;

[0084] It is worth further explaining that the practical significance of this formula is to quantify the degree of quality improvement of the new dataset relative to the benchmark dataset.

[0085] Step S2c: combining the enhanced quality factor with the comprehensive score to obtain an optimized threshold comprehensive score;

[0086] It is worth noting that combining the enhanced quality factor with the comprehensive score can effectively introduce a strict selection standard for the data set for model optimization, which is the optimization threshold comprehensive score, so that the new data set can be significantly better than the benchmark data set in quality, thereby effectively preventing the overfitting problem; further, by introducing the enhanced quality factor, the optimization threshold comprehensive score can be dynamically adjusted so that it can adapt to the ever-changing data environment. Specifically, each time the model is optimized and trained, it is often necessary to recalculate the F1 score and recalculate the comprehensive score of the entire data set used for optimizing and training the machine learning model based on the F1 score combined with the integrity score, distribution score, noise detection score and annotation quality score. This comprehensive score will only continue to improve as the model is continuously optimized and trained. To this end, the enhanced quality factor is introduced and combined with the comprehensive score. During the process of continuous optimization and training of the model, its optimization threshold comprehensive score can be continuously updated to adapt to the process and requirements of continuous optimization training of the model, thereby meeting the high scientificity and rigor of the optimization threshold comprehensive score.

[0087] Among them, the calculation formula for the optimized threshold comprehensive score is:

[0088] ;

[0089] Where, represents the optimized threshold comprehensive score; represents the comprehensive score; represents the enhancement quality factor; Represents the second weight coefficient, which is used to balance the contribution of the comprehensive score and the enhanced quality factor, and 0< <1;

[0090] It's worth noting that the practical significance of this formula lies in its combination of a comprehensive score and an enhanced quality factor, providing a rigorous screening criterion that ensures that the quality of new datasets is significantly superior to that of the baseline dataset. This approach helps dynamically adjust the optimization threshold to adapt to changing data environments, thereby improving the model's stability and generalization capabilities.

[0091] Step S3: Reacquire a new dataset, and calculate the first comprehensive score of the current new dataset according to the calculation principle of the first comprehensive score. At the same time, determine the first comprehensive score of the current new dataset according to the optimized threshold comprehensive score. If the first comprehensive score of the current new dataset is greater than the optimized threshold comprehensive score, retain the current new dataset; if the first comprehensive score of the current new dataset is less than or equal to the optimized threshold comprehensive score, remove it, and acquire another new dataset until it is retained.

[0092] It is worth noting that a strict screening mechanism can effectively ensure that the quality of the new dataset is significantly better than the benchmark dataset, thereby effectively ensuring the accuracy of the pre-trained machine learning model after subsequent optimization.

[0093] Step S4: Based on the judgment result, a new data set is selected to be retained for optimization training of the pre-trained machine learning model to obtain an optimized and trained machine learning model. The optimized and trained machine learning model is used to predict the input liquidation data, output the prediction result, and make a liquidation decision based on the prediction result.

[0094] It is worth noting that the use of the retained new data set to further train the model is mainly to further improve the model performance; the output prediction results include time series prediction results, such as the prediction of key time nodes in the liquidation process, specifically the expected time to complete the liquidation, the completion time of tasks in each stage, etc. Specifically, for example, it is predicted that a certain company is expected to complete all liquidation work within 3 months, of which the first two months are mainly for dealing with fixed assets, and the last month is for dealing with current assets; based on the time series prediction results, the liquidation decisions made include optimizing the liquidation process, ensuring that the liquidation work is completed on time, etc. For example, when it is predicted that a certain company is expected to complete all liquidation work within 3 months, of which the first two months are mainly for dealing with fixed assets, and the last month is for dealing with current assets, a detailed liquidation plan can be formulated based on this to ensure that all work is carried out as planned.

[0095] Example 2: Figure 2 As shown, the present invention also proposes a liquidation data intelligent analysis system based on market evaluation, which executes the liquidation data intelligent analysis method based on market evaluation as described above, including the following modules:

[0096] A first comprehensive score calculation module: configured to obtain at least two new data sets and calculate a first comprehensive score for each new data set;

[0097] Optimization threshold comprehensive score calculation module: connected to the first comprehensive score calculation module, used to call the comprehensive score, and calculate the optimization threshold comprehensive score based on the comprehensive score and the first comprehensive score;

[0098] New data set determination module: connected to the optimization threshold comprehensive score calculation module, used to re-acquire a new data set, and calculate the first comprehensive score of the current new data set according to the calculation principle of the first comprehensive score, and at the same time determine the first comprehensive score of the current new data set according to the optimization threshold comprehensive score. If the first comprehensive score of the current new data set is greater than the optimization threshold comprehensive score, the current new data set is retained; if the first comprehensive score of the current new data set is less than or equal to the optimization threshold comprehensive score, it is eliminated and a new data set is obtained again until it is retained;

[0099] Model optimization and decision-making module: connected to the new data set judgment module, used to select the retained new data set based on the judgment results to optimize the pre-trained machine learning model, obtain the optimized and trained machine learning model, and use the optimized and trained machine learning model to predict the input liquidation data, output the prediction results, and make liquidation decisions based on the prediction results.

[0100] It should be noted that the terms used in the present invention are only for describing specific embodiments and are not intended to limit the scope of this application. As shown in the present specification, unless the context clearly indicates an exception, the words "one", "a", "a kind of" and / or "the" do not specifically refer to the singular and may also include the plural. The terms "comprise", "include" or any other variants thereof are intended to cover non-exclusive inclusion, so that the process, method or device comprising a series of elements includes not only those elements, but also includes other elements not explicitly listed, or also includes elements inherent to such process, method or device. In the absence of further restrictions, the elements defined by the sentence "comprise a..." do not exclude the presence of other identical elements in the process, method or device comprising the elements.

[0101] It should also be noted that the terms "center", "up", "down", "left", "right", "vertical", "horizontal", "inside", "outside", etc., indicating orientations or positional relationships, are based on the orientations or positional relationships shown in the accompanying drawings. They are only for the convenience of describing the present invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation. Therefore, they cannot be understood as limitations on the present invention. Unless otherwise clearly specified and limited, the terms "installed", "connected", "connected", etc. should be understood in a broad sense. For example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be a direct connection, or an indirect connection through an intermediate medium, or it can be a communication between the internal parts of two elements. For those of ordinary skill in the art, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.

[0102] 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 above embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the above embodiments, or replace some or all of the technical features therein with equivalents. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the technical solutions of the embodiments of the present invention.

Claims

1. The method for intelligent analysis of liquidation data based on market evaluation is characterized by: The following steps are involved: Step S1: Obtain at least two new data sets, and calculate a first comprehensive score for each new data set; Step S2: calling the comprehensive score, and calculating the optimized threshold comprehensive score based on the comprehensive score and the first comprehensive score; Step S2a, calculating the difference between the first comprehensive score and the comprehensive score of each new data set; Step S2b: Calculate an enhanced quality factor based on the difference between the first comprehensive score and the comprehensive score of each new data set; Step S2b1: based on the calculation result of the difference between the first comprehensive score and the comprehensive score of each new data set, retain the differences with positive calculation results and eliminate the differences with negative calculation results; Step S2b2, counting the number of differences whose calculation results are positive numbers; Step S2b3: Based on the statistical results, obtain the maximum value of the differences whose calculated results are positive numbers; Step S2b4: Calculate the enhancement quality factor based on the difference values ​​whose calculation results are positive and the maximum value of the difference values ​​whose calculation results are positive; ； Where, represents the enhancement quality factor; Represents the number of differences that evaluate to positive numbers; represents the kth difference whose calculation result is a positive number; Represents the maximum value among the differences whose calculation results are positive numbers; Step S2c: combining the enhanced quality factor with the comprehensive score to obtain an optimized threshold comprehensive score; Step S3: Reacquire a new dataset, and calculate the first comprehensive score of the current new dataset according to the calculation principle of the first comprehensive score. At the same time, determine the first comprehensive score of the current new dataset according to the optimized threshold comprehensive score. If the first comprehensive score of the current new dataset is greater than the optimized threshold comprehensive score, retain the current new dataset; if the first comprehensive score of the current new dataset is less than or equal to the optimized threshold comprehensive score, remove it, and acquire another new dataset until it is retained. Step S4: Based on the judgment result, a new data set is selected to be retained for optimization training of the pre-trained machine learning model to obtain an optimized and trained machine learning model. The optimized and trained machine learning model is used to predict the input liquidation data, output the prediction result, and make a liquidation decision based on the prediction result. Combining the enhanced quality factor with the comprehensive score effectively introduces a strict selection criterion for the dataset used for model optimization, namely the optimized threshold comprehensive score, so that the new dataset can significantly outperform the benchmark dataset in quality.

2. The method for intelligent analysis of liquidation data based on market evaluation according to claim 1, characterized in that: Calculating a first comprehensive score for each new data set includes: Step S1a: Calculate the completeness score, distribution score, noise detection score, and annotation quality score of each new dataset and combine them to obtain a comprehensive score for each new dataset; Step S1b, calculating the data consistency score of each new data set; Step S1c: combining the comprehensive score of each new data set with the data consistency score of each new data set to obtain a first comprehensive score for each new data set; The calculation formula for obtaining the first comprehensive score of each new data set is: ； Where, Represents the first comprehensive score of the i-th group of new data sets; Represents the comprehensive score of the i-th group of new data sets; represents the data consistency score of the i-th group of new data sets; Represents the first weight coefficient, which is used to balance the contribution of the comprehensive score and the data consistency score, and 0< <1.

3. The method for intelligent analysis of liquidation data based on market evaluation according to claim 2, characterized in that: The calculation of the data consistency score of each new data set includes: Step S1b1, calculating the mean of all data points in each set of new data sets to obtain the mean of each set of new data sets; Step S1b2: Calculate the standard deviation of each new data set based on the mean of each new data set; Step S1b3: Calculate the data consistency score of each new data set based on the mean and standard deviation of each new data set; The calculation formula for obtaining the data consistency score of each new data set is: ； Where, represents the data consistency score of the i-th group of new data sets; Represents the total number of data points in the i-th group of new data sets; Represents the jth data point in the i-th group of new data sets; Represents the mean of the i-th group of new data sets; Represents the standard deviation of the i-th group of new data sets.

4. The method for intelligent analysis of liquidation data based on market evaluation according to claim 1, characterized in that: The calculation formula for the optimized threshold comprehensive score is: ； Where, represents the optimized threshold comprehensive score; represents the comprehensive score; represents the enhancement quality factor; Represents the second weight coefficient, which is used to balance the contribution of the comprehensive score and the enhanced quality factor, and 0< <1.

5. A liquidation data intelligent analysis system based on market evaluation, applied to the liquidation data intelligent analysis method based on market evaluation according to any one of claims 1 to 4, characterized in that: The system comprises: A first comprehensive score calculation module: configured to obtain at least two new data sets and calculate a first comprehensive score for each new data set; Optimization threshold comprehensive score calculation module: connected to the first comprehensive score calculation module, used to call the comprehensive score, and calculate the optimization threshold comprehensive score based on the comprehensive score and the first comprehensive score; New data set determination module: connected to the optimization threshold comprehensive score calculation module, used to re-acquire a new data set, and calculate the first comprehensive score of the current new data set according to the calculation principle of the first comprehensive score, and at the same time determine the first comprehensive score of the current new data set according to the optimization threshold comprehensive score. If the first comprehensive score of the current new data set is greater than the optimization threshold comprehensive score, the current new data set is retained; if the first comprehensive score of the current new data set is less than or equal to the optimization threshold comprehensive score, it is eliminated and a new data set is obtained again until it is retained; Model optimization and decision-making module: connected to the new data set judgment module, used to select the retained new data set based on the judgment results to optimize the pre-trained machine learning model, obtain the optimized and trained machine learning model, and use the optimized and trained machine learning model to predict the input liquidation data, output the prediction results, and make liquidation decisions based on the prediction results.

Images