Financial decision support system and method based on big data

Abstract

The invention relates to the technical field of financial analysis, in particular to a financial decision support system and method based on big data. The method comprises the steps of obtaining historical enterprise data of a plurality of enterprises, preprocessing the historical enterprise data to obtain preprocessed enterprise data, then creating a decision support model, inputting the preprocessed enterprise data into the decision support model, and calculating a decision score corresponding to the enterprise by combining the historical enterprise data of each enterprise through the decision support model. The method comprises the steps of training a decision support model of a tested enterprise, inputting real-time enterprise data into the trained decision support model, obtaining a decision score, output by the trained decision support model, of the tested enterprise, and finally dividing a decision grade of the tested enterprise based on the decision score of the tested enterprise. And the decision score of the tested enterprise is obtained by combining the real-time decision score of the tested enterprise and the decision score of the approximate enterprise, so that the reliability of the decision score of the tested enterprise is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of financial analysis, and in particular to a financial decision support system and method based on big data. Background Art

[0002] With the rapid development of information technology and the advent of the big data era, enterprises are faced with massive data resources. These data contain rich information and are of great value to the financial decision-making of enterprises. However, how to effectively use these data and improve the scientificity and accuracy of financial decision-making has become a major challenge facing enterprises. Traditional financial decision-making methods often rely on experience and intuition, which are difficult to cope with the complex and changing market environment.

[0003] Therefore, it is of great practical significance and application value to study a financial decision support system and method based on big data, provide enterprises with comprehensive and accurate financial information through data collection, processing, analysis and mining, and assist decision makers in making more informed decisions.

[0004] A Chinese patent with announcement number CN109658217B discloses an intelligent financial decision-making big data analysis system, which implements overall financial planning by conducting risk assessment on corporate finances and providing reliable preventive measures, and provides managers with reliable decision-making analysis and guidance in financial management. However, the prior art does not explain how to make corresponding financial decisions based on the company's corporate data in real time, which makes it difficult to update financial decisions in a timely manner, causing financial decisions to lag behind corporate development. Summary of the invention

[0005] The purpose of the present invention is to propose a financial decision support system and method based on big data in response to the problems existing in the background technology.

[0006] The technical solution of the present invention: On the one hand, the present application provides a financial decision support method based on big data, comprising: Acquire historical enterprise data of multiple enterprises, and pre-process the historical enterprise data to obtain pre-processed enterprise data; Create a decision support model, input the pre-processed enterprise data into the decision support model, calculate the decision score corresponding to each enterprise through the decision support model combined with the historical enterprise data of each enterprise, and obtain the trained decision support model; Collecting real-time enterprise data of the enterprise under test, inputting the real-time enterprise data into the trained decision support model, obtaining the real-time decision score of the enterprise under test output by the trained decision support model, and calculating the decision score of the enterprise under test by combining the real-time decision score of the enterprise under test with the decision score of the approximate enterprise; The decision-making levels of the tested enterprises are divided based on their decision scores.

[0007] Preferably, historical enterprise data of multiple enterprises are obtained, and the historical enterprise data are preprocessed to obtain preprocessed enterprise data, including: Create enterprise data tables; Obtain historical enterprise data of multiple enterprises, and put all the historical enterprise data into an enterprise data table; the historical enterprise data includes multiple sub-data such as enterprise type, enterprise legal representative, enterprise shareholder, enterprise registered capital, and enterprise establishment time; Randomly select a sub-data from the enterprise data table; Select the maximum value and the minimum value contained in the sub-data, divide the sub-data evenly into multiple data segments based on the maximum value and the minimum value contained in the sub-data, and set a corresponding score for each data segment in order from small to large; Return to randomly select a sub-data from the enterprise data table until all types of sub-data in the enterprise data table have been selected, and obtain the score of each type of sub-data.

[0008] Preferably, a decision support model is created, the pre-processed enterprise data is input into the decision support model, and the decision support model calculates the decision score corresponding to each enterprise in combination with the historical enterprise data of each enterprise to obtain a trained decision support model, including: Create decision support models; Divide the preprocessed enterprise data into training and test sets in random proportions; Inputting the training set into the decision support model to train the decision support model, so that the decision support model calculates the decision score corresponding to each enterprise according to the enterprise data of each enterprise, and obtains a trained decision support model; The test set is input into the trained decision support model, the predicted decision score output by the trained decision support model is obtained, and it is verified whether the predicted decision score is correct.

[0009] Preferably, the training set is input into the decision support model to train the decision support model, so that the decision support model calculates the decision score corresponding to each enterprise according to the enterprise data of each enterprise, and obtains the trained decision support model, including: Select an enterprise and the historical enterprise data of the enterprise from the training set; Calculate the financial score of the enterprise according to Formula 1 based on the historical enterprise data of the enterprise; Formula 1; in, is the financial score of the enterprise, is the score of the i-th sub-data of the enterprise, is the weight corresponding to the i-th sub-data of the enterprise, is the total number of sub-data; Establishing the coupling relationship between enterprise-enterprise data and decision-making scores, and sequentially inputting the coupling relationship between enterprise-enterprise data and decision-making scores into the decision support model as a training sample, so that the decision support model continuously learns the relationship between enterprise data and decision-making scores; Return to select an enterprise and its historical enterprise data from the training set until all enterprises in the training set are selected, obtain the decision score of each enterprise, and input each enterprise and its decision score as a training sample into the decision support model to train the decision support model, and obtain the trained decision support model.

[0010] Preferably, the test set is input into the trained decision support model, the predicted decision score output by the trained decision support model is obtained, and the predicted decision score is verified to be correct, including: Randomly select a company and its historical enterprise data from the test set; The decision score of the enterprise is calculated by formula 1; Inputting the historical data of the enterprise into the trained decision support model to obtain the predicted decision score output by the trained decision support model; Set a deviation threshold to determine whether the deviation between the predicted decision score and the decision score is greater than or equal to the deviation threshold; If the deviation between the predicted decision score and the decision score is greater than or equal to the deviation threshold, the trained decision support model is unqualified and the pre-processed enterprise data is divided into a training set and a test set at a random ratio until the deviation between the predicted decision score and the decision score is less than the deviation threshold; If the deviation between the predicted decision score and the decision score is less than the deviation threshold, the trained decision support model is qualified and returns to randomly select an enterprise from the test set and the historical enterprise data of the enterprise until all enterprises in the test set are selected.

[0011] Preferably, real-time enterprise data of the enterprise under test is collected, the real-time enterprise data is input into the trained decision support model, the real-time decision score of the enterprise under test output by the trained decision support model is obtained, and the decision score of the enterprise under test is calculated by combining the real-time decision score of the enterprise under test with the decision score of the approximate enterprise, including: Collect real-time enterprise data; The real-time enterprise data is input into the trained decision support model, and the trained decision support model outputs the real-time decision score of the tested enterprise and the decision scores of multiple similar enterprises based on the real-time enterprise data; The decision score of the tested enterprise is calculated by combining the real-time decision score of the tested enterprise and the decision scores of multiple similar enterprises.

[0012] Preferably, the real-time enterprise data is input into a trained decision support model, and the trained decision support model outputs a decision score of the enterprise and decision scores of multiple similar enterprises based on the real-time enterprise data, including: Select K sub-data; record the selected K sub-data as initial target sub-data; Select an initial target sub-data, and randomly select M initial target sub-data from all enterprise data as the initial cluster centers; Calculate the distances between the remaining target sub-data and the initial cluster center respectively, and divide the enterprise data corresponding to the target sub-data to the closest initial cluster center according to the distances between the remaining target sub-data and the initial cluster center, forming M target clusters; Calculate the mean of the distances within each target cluster, and use the data point corresponding to the mean as the new cluster center; Divide enterprise data into clusters with the nearest cluster center according to the distance from the new cluster center; Set distance threshold; Calculate the distance between the new cluster center and the previous cluster center, and determine whether the distance between the new cluster center and the previous cluster center is less than or equal to the distance threshold; If the distance between the new cluster center and the previous cluster center is greater than the distance threshold, return to calculate the mean of the distances within each target cluster, and use the data point corresponding to the mean as the new cluster center until the distance between the new cluster center and the previous cluster center is less than or equal to the distance threshold; If the distance between the new cluster center and the previous cluster center is less than or equal to the distance threshold, the last clustering result is output; the enterprise corresponding to the enterprise data contained in the last clustering result is taken as an approximate enterprise; Return to select an initial target sub-data until all the initial target sub-data are selected, and get A similar enterprise.

[0013] Preferably, the decision score of the enterprise under test is calculated by combining the real-time decision score of the enterprise under test and the decision scores of a plurality of similar enterprises, including: Input the enterprise data of all similar enterprises into the decision support model in sequence to obtain the decision score of each similar enterprise; Input real-time enterprise data into the decision support model to obtain real-time enterprise decision scores; The decision score of the tested enterprise is calculated by formula 2 based on the enterprise's real-time decision score and the decision scores of all similar enterprises; Formula 2; in, is the decision score of the enterprise under test, It is the real-time decision-making score of the enterprise under test. is the decision score of the ith approximate enterprise, , M is the approximate total number of enterprises.

[0014] Preferably, the decision-making levels of the tested enterprises are divided based on the decision-making scores of the tested enterprises, including: Set the enterprise decision-making level and the decision-making score range corresponding to each enterprise decision-making level; Determine the decision score range of the tested enterprise, and set the tested enterprise to the enterprise decision level corresponding to the decision score range.

[0015] On the other hand, the present application also provides a financial decision support system based on big data, including: A data acquisition component, through which real-time enterprise data and historical enterprise data are acquired; A data processing component, through which the aforementioned financial decision support method based on big data is executed, the data acquisition component is communicatively connected with the data processing component, the data processing component includes a data processing unit and a data analysis unit, the data processing unit is used to process enterprise data to obtain data processing results, and the data analysis unit is used to output enterprise decisions in combination with the data processing results.

[0016] Compared with the prior art, the above technical solution of the present invention has the following beneficial technical effects: By acquiring historical enterprise data of multiple enterprises and preprocessing the historical enterprise data to obtain preprocessed enterprise data, a decision support model is created, the preprocessed enterprise data is input into the decision support model, and the decision support model is used to calculate the corresponding decision score of each enterprise in combination with the historical enterprise data of the enterprise. Then, the real-time enterprise data of the tested enterprise is collected, and the real-time enterprise data is input into the trained decision support model to obtain the decision score of the tested enterprise output by the trained decision support model. Finally, the decision level of the tested enterprise is divided based on the decision score of the tested enterprise. The present application calculates the real-time decision score of the tested enterprise by setting a corresponding score and weight for each sub-data in the historical enterprise data, and obtains the decision score of the tested enterprise by combining the real-time decision score of the tested enterprise with the decision score of an approximate enterprise, so as to improve the reliability of the decision score of the tested enterprise. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] Figure 1 A flowchart of a financial decision support method based on big data proposed by the present invention; Figure 2 This is a schematic diagram of the structure of a financial decision support system based on big data proposed by the present invention.

[0018] Reference numerals: 100, data acquisition component; 200, data processing component; 201, data processing unit; 202. Data analysis unit. DETAILED DESCRIPTION

[0019] Embodiment 1, as Figure 1 As shown, the present invention proposes a financial decision support method based on big data, comprising: S100, acquiring historical enterprise data of multiple enterprises, and preprocessing the historical enterprise data to obtain preprocessed enterprise data; S200, creating a decision support model, inputting the pre-processed enterprise data into the decision support model, calculating the decision score corresponding to each enterprise through the decision support model combined with the historical enterprise data of each enterprise, and obtaining a trained decision support model; S300, collecting real-time enterprise data of the enterprise under test, inputting the real-time enterprise data into the trained decision support model, obtaining the real-time decision score of the enterprise under test output by the trained decision support model, and calculating the decision score of the enterprise under test by combining the real-time decision score of the enterprise under test with the decision score of the approximate enterprise; S400, dividing the decision-making level of the tested enterprise based on the decision-making score of the tested enterprise.

[0020] In the present invention, historical enterprise data of multiple enterprises are obtained and preprocessed to obtain preprocessed enterprise data, and then a decision support model is created, and the preprocessed enterprise data is input into the decision support model, and the decision score corresponding to each enterprise is calculated by the decision support model in combination with the historical enterprise data of the enterprise, and then the real-time enterprise data of the tested enterprise is collected, and the real-time enterprise data is input into the trained decision support model to obtain the decision score of the tested enterprise output by the trained decision support model, and finally the decision level of the tested enterprise is divided based on the decision score of the tested enterprise. The present application calculates the real-time decision score of the tested enterprise by setting a corresponding score and weight for each sub-data in the historical enterprise data, and obtains the decision score of the tested enterprise by combining the real-time decision score of the tested enterprise with the decision score of an approximate enterprise, so as to improve the reliability of the decision score of the tested enterprise.

[0021] In an optional embodiment, the S100 includes: S110, create an enterprise data table; S120, obtaining historical enterprise data of multiple enterprises, and putting all the historical enterprise data into an enterprise data table; the historical enterprise data includes multiple sub-data such as enterprise type, enterprise legal representative, enterprise shareholder, enterprise registered capital, and enterprise establishment time; Specifically, both historical enterprise data and real-time enterprise data belong to enterprise data. The sub-data contained in historical enterprise data and real-time enterprise data are the same. Historical enterprise data is mainly used for training and verification of decision support models, while real-time enterprise data is used to calculate the decision score of the enterprise. S130, randomly selecting a sub-data from the enterprise data table; S140, selecting a maximum value and a minimum value included in the sub-data, dividing the sub-data evenly into a plurality of data segments based on the maximum value and the minimum value included in the sub-data, and setting a corresponding score for each data segment in order from small to large; S150, returning to step S130, until all types of sub-data in the enterprise data table are selected, and the score of each type of sub-data is obtained.

[0022] It should be noted that the historical enterprise data collected in this application does not refer to the enterprise's previous enterprise data. The historical enterprise data here also refers to the enterprise's latest enterprise data. The enterprise's historical data is only used to distinguish it from real-time enterprise data. It is enterprise data used to train decision support models. The scoring of sub-data is not only used to calculate the decision score in formula 1, but also to calculate the real-time decision score in formula 2.

[0023] After collecting historical enterprise data, this application sets corresponding scores for each type of sub-data in the historical enterprise data. For example, for the sub-data of enterprise establishment time, the longer the enterprise is established, the higher the score. When setting the specific score, the score of enterprises with an establishment time of 0-5 years can be set to 0.5, and the score of enterprises with an establishment time of 6-10 years can be set to 0.8. The reason why the corresponding score of the enterprise increases with the increase of the establishment time is that the longer the establishment time is, the more it can indicate to a certain extent that the enterprise is a more trustworthy enterprise, so improving the score of the enterprise will increase the decision-making score of the enterprise and enable the enterprise to receive a better financial decision.

[0024] For corporate data such as corporate shareholders that cannot be quantified, big data can be used to obtain other shareholding companies of the corporate shareholders, and corresponding scores can be set for the corporate shareholders based on the other shareholding companies of the corporate shareholders. The better the operating data of the other shareholding companies of the corporate shareholders, the higher the score of the corporate shareholders will be. When obtaining other shareholding companies of the corporate shareholders through big data, other companies held by the largest shareholder of the company will be given priority.

[0025] In an optional embodiment, the S200 includes: S210, create a decision support model; S220, dividing the pre-processed enterprise data into a training set and a test set in a random ratio; Specifically, the ratio of the training set should be higher than that of the test set, which can be 7:3; S230, inputting the training set into the decision support model to train the decision support model, so that the decision support model calculates the decision score corresponding to each enterprise according to the enterprise data of each enterprise, and obtains a trained decision support model; S240, input the test set into the trained decision support model, obtain the predicted decision score output by the trained decision support model, and verify whether the predicted decision score is correct.

[0026] It should be noted that by training the decision support model with historical enterprise data, the decision support model continuously learns the coupling relationship between the enterprise, enterprise data and decision scores, so that the trained decision support model can directly output the decision score of the enterprise based on the real-time enterprise data of the enterprise.

[0027] In an optional embodiment, the S230 includes: S231, selecting an enterprise and the historical enterprise data of the enterprise from the training set; S232, calculating the financial score of the enterprise according to Formula 1 based on the historical enterprise data of the enterprise; Formula 1; in, is the financial score of the enterprise, is the i-th sub-data of the enterprise, is the weight corresponding to the i-th sub-data of the enterprise, is the total number of sub-data; S233, establishing a coupling relationship between enterprise-enterprise data and decision scores, and sequentially inputting the coupling relationship between enterprise-enterprise data and decision scores into a decision support model as a training sample, so that the decision support model continuously learns the relationship between enterprise data and decision scores. S234, returns to select an enterprise and the historical enterprise data of the enterprise from the training set, until all the enterprises in the training set are selected, obtains the decision score of each enterprise, and inputs each enterprise and its decision score as a training sample into the decision support model to train the decision support model, and obtains the trained decision support model.

[0028] It should be noted that since the enterprise data contains multiple sub-data, each sub-data has different degrees of influence on the enterprise's decision-making score. Therefore, by setting corresponding weights for each sub-data, the influence of different sub-data in the decision-making score can be reflected. Since the weight can be adjusted freely, the proportion of the target sub-data in the decision-making score can be increased by increasing the weight.

[0029] In the prior art, the operating data of an enterprise is often given more importance, but the weight of the legal representative and shareholders of the enterprise in the decision-making score of the enterprise is easily overlooked. Therefore, in Formula 1, the weight corresponding to the legal person who meets the standards and the weight corresponding to the shareholders of the enterprise can be increased, thereby improving the influence of the legal representative and shareholders of the enterprise in the decision-making score.

[0030] For example, existing companies A and B are small and medium-sized enterprises with similar operating data, but the legal person of company A is A, and the legal person of company B is B. After big data screening, the operating conditions of other related companies of company A are better than those of other related companies of company B. Therefore, when calculating the decision scores of companies A and B, the score of company A in the sub-data of "legal representative of the enterprise" should be higher than the score of company B in the sub-data of "legal representative of the enterprise". Then, by increasing the weight of the sub-data of "legal representative of the enterprise", the influence of different legal representatives on the decision scores of the enterprises can be reflected in the decision scores of companies A and B.

[0031] In an optional embodiment, the S240 includes: S241, randomly select an enterprise and the historical enterprise data of the enterprise from the test set; S242, calculating the decision score of the enterprise by formula 1; S243, inputting the historical data of the enterprise into the trained decision support model to obtain the predicted decision score output by the trained decision support model; S244, setting a deviation threshold, and determining whether the deviation between the predicted decision score and the decision score is greater than or equal to the deviation threshold; S245, if the deviation between the predicted decision score and the decision score is greater than or equal to the deviation threshold, the trained decision support model is unqualified, and the pre-processed enterprise data is divided into a training set and a test set in a random proportion until the deviation between the predicted decision score and the decision score is less than the deviation threshold; S246, if the deviation between the predicted decision score and the decision score is less than the deviation threshold, the trained decision support model is qualified, and the process returns to step S241 until all enterprises in the test set are selected.

[0032] It should be noted that the test set is used to verify whether the trained decision support model has the ability to automatically output the decision score of the enterprise based on the input enterprise data, thereby determining whether the decision support model is trained.

[0033] When verifying, the larger the deviation threshold is set, the looser the review of the decision support model. Similarly, an overall pass rate threshold can also be set. For example, if the overall pass rate threshold is 80%, it only needs to be that the decision scores of 80% or more of the companies in the test set and the predicted decision scores output by the trained decision support model are less than the deviation threshold.

[0034] In an optional embodiment, the S300 includes: S310, collects real-time enterprise data; Specifically, the sub-data included in the real-time enterprise data and the historical enterprise data are the same, including enterprise type, enterprise legal representative, enterprise shareholders, enterprise registered capital, enterprise establishment time and other sub-data; S320, inputting the real-time enterprise data into the trained decision support model, and the trained decision support model outputs the real-time decision score of the tested enterprise and the decision scores of multiple similar enterprises based on the real-time enterprise data; S330, calculating a decision score of the enterprise under test by combining the real-time decision score of the enterprise under test and the decision scores of a plurality of similar enterprises.

[0035] It should be noted that, since the trained decision support model has been obtained in the aforementioned embodiment, it is only necessary to input the real-time enterprise information of the tested enterprise into the trained decision support model to obtain the real-time decision score of the tested enterprise.

[0036] In order to further improve the reliability of the decision-making scores of the tested enterprises, on the basis of real-time decision-making scores, it is necessary to additionally integrate the decision-making scores of similar enterprises that are close to the tested enterprise, so as to obtain the final decision-making score of the tested enterprise.

[0037] Since the real-time data of an enterprise can only reflect the operating status of the enterprise up to the time point being tested, and financial decisions have a far-reaching impact on the subsequent operations of the tested enterprise, similar enterprises are introduced to give a more reliable decision score to the tested enterprise by observing the subsequent operating conditions of similar enterprises.

[0038] In an optional embodiment, the S320 includes: S321, select K sub-data; record the selected K sub-data as initial target sub-data; S322, select an initial target sub-data, and randomly select M initial target sub-data from all enterprise data as initial cluster centers; S323, respectively calculating the distances between the remaining target sub-data and the initial cluster center, and dividing the enterprise data corresponding to the target sub-data to the closest initial cluster center according to the distances between the remaining target sub-data and the initial cluster center, to form M target clusters; Specifically, when calculating the distance between the target sub-data and the initial cluster center, what is actually calculated is the data corresponding to the target sub-data. That is, if the selected target sub-data is the establishment time of the enterprise, then it is only necessary to calculate the degree of proximity between the establishment time of other enterprises and the establishment time of the enterprise under test; S324, calculating the mean of the distances within each target cluster, and taking the data point corresponding to the mean as the new cluster center; S325, dividing the enterprise data into the cluster where the nearest cluster center is located according to the distance from the new cluster center; S326, setting a distance threshold; S327, calculating the distance between the new cluster center and the previous cluster center, and determining whether the distance between the new cluster center and the previous cluster center is less than or equal to a distance threshold; S328, if the distance between the new cluster center and the previous cluster center is greater than the distance threshold, return to calculate the mean of the distances within each target cluster, and use the data point corresponding to the mean as the new cluster center until the distance between the new cluster center and the previous cluster center is less than or equal to the distance threshold; S329, if the distance between the new cluster center and the previous cluster center is less than or equal to the distance threshold, output the last clustering result; and take the enterprise corresponding to the enterprise data included in the last clustering result as an approximate enterprise; S329-1, return to select an initial target sub-data, until all the initial target sub-data are selected, and obtain A similar enterprise.

[0039] It should be noted that in this embodiment, cluster analysis is performed on the tested enterprises to obtain the enterprises that are close to the tested enterprises in terms of K target sub-data. Approximate enterprises, each target sub-data is to find However, the number of similar enterprises gathered by each target sub-data can be different. The number of clustered approximate firms is , sub-goal The number of similar companies clustered is ,and , then the total number of approximate enterprises is .

[0040] If there is a tested enterprise C, which is a small and medium-sized enterprise established 8 years ago and whose shareholders include C, D and E, and the target sub-data of the tested enterprise are enterprise type, enterprise shareholders and enterprise establishment time, then the value obtained by taking "enterprise type" as the target sub-data is The types of similar enterprises are all small and medium-sized enterprises. Taking "corporate shareholders" as the target sub-data, we get The shareholders of similar enterprises include at least one or more of C, D and E. The target sub-data is “Enterprise Establishment Time”. The establishment time of the similar enterprises is 8-10 years. Because the similar enterprises are very similar to the tested enterprise in one or more sub-data, the decision scores of these similar enterprises can be used as a reference for the tested enterprise.

[0041] In an optional embodiment, the S330 includes: S331, inputting the enterprise data of all similar enterprises into the decision support model in sequence to obtain the decision score of each similar enterprise; S332, inputting real-time enterprise data into a decision support model to obtain a real-time decision score for the enterprise; S333, calculating the decision score of the tested enterprise by formula 2 based on the real-time decision score of the enterprise and the decision scores of all similar enterprises; Formula 2; in, is the decision score of the enterprise under test, It is the real-time decision-making score of the enterprise under test. is the decision score of the ith approximate enterprise, , M is the total number of approximate enterprises; Optionally, the decision score of the tested enterprise can also be calculated by Formula 3; Formula 3; in, is the decision score of the enterprise under test, is the score of the i-th sub-data of the tested enterprise, is the weight corresponding to the i-th sub-data of the tested enterprise, The target subdata The ratings of the companies tested, The target subdata No. Ratings of similar companies, The target subdata The corresponding weights; Since K sub-data are selected as target sub-data, the remaining unselected sub-data in the real-time enterprise data of the tested enterprise in Formula 3 are still calculated according to the original scores and weights. Only the scores of the target sub-data of the approximate enterprises are integrated into the scores of the target sub-data of the tested enterprise, thereby reducing the influence of useless sub-data in the approximate enterprises and making the calculation of the decision score of the tested enterprise more accurate.

[0042] It should be noted that similar enterprises to the enterprise under test are found through the aforementioned embodiment. In order to increase the influence of the decision scores of the similar enterprises on the score of the enterprise under test, the decision scores of the similar enterprises are integrated into the decision scores of the enterprise under test through Formula 2, thereby making the final decision scores of the enterprise under test more reliable.

[0043] In an optional embodiment, the S400 includes: S410, setting the enterprise decision level and the decision scoring range corresponding to each enterprise decision level; S420, determining the decision scoring range of the tested enterprise, and setting the tested enterprise to the enterprise decision level corresponding to the decision scoring range.

[0044] It should be noted that different decision scores correspond to different financial decisions. Generally speaking, the higher the decision score of the tested enterprise, the higher its corresponding financial decision level.

[0045] Each level of corporate decision-making level corresponds to different financial decisions. The higher the corporate decision-making level, the more preferential its financial decisions will be. For example, companies corresponding to high-level corporate decision-making levels can enjoy lower interest rate loan policies than companies corresponding to low-level corporate decision-making levels.

[0046] Since this application obtains the enterprise decision-making level by collecting the real-time enterprise data of the enterprise and using the real-time enterprise data of the enterprise as the basis, the enterprise decision-making level of the enterprise can be continuously adjusted as the real-time enterprise data changes, ensuring that the financial decisions implemented by the enterprise are updated synchronously with the real-time enterprise data of the enterprise, thereby improving the accuracy and timeliness of financial decisions and avoiding lagging financial decisions.

[0047] like Figure 2 As shown, the present application also provides a financial decision support system based on big data, including a data acquisition component 100 and a data processing component 200.

[0048] Real-time enterprise data and historical enterprise data are acquired through the data acquisition component 100, and the financial decision support method based on big data as described in Example 1 is executed through the data processing component 200. The data acquisition component 100 is communicated with the data processing component 200, and the data processing component 200 includes a data processing unit 201 and a data analysis unit 202. The enterprise data is processed by the data processing unit 201 to obtain a data processing result, and the enterprise decision is output by the data analysis unit 202 in combination with the data processing result.

[0049] It should be noted that the enterprise data is acquired through the data acquisition component 100, and the enterprise data includes historical enterprise data and real-time enterprise data, wherein the historical enterprise data is used for training the decision support model, and the real-time enterprise data is used to calculate the decision score of the tested enterprise, so that the data processing component 200 can implement financial decisions for the tested enterprise according to the decision score of the tested enterprise.

[0050] The embodiments of the present invention are described in detail above with reference to the accompanying drawings, but the present invention is not limited thereto, and various changes can be made within the knowledge scope of technicians in the relevant technical field without departing from the purpose of the present invention.

Claims

1. A financial decision support method based on big data, characterized in that: include: Acquire historical enterprise data of multiple enterprises, and pre-process the historical enterprise data to obtain pre-processed enterprise data; Create a decision support model, input the pre-processed enterprise data into the decision support model, calculate the decision score corresponding to each enterprise through the decision support model combined with the historical enterprise data of each enterprise, and obtain the trained decision support model; Collecting real-time enterprise data of the enterprise under test, inputting the real-time enterprise data into the trained decision support model, obtaining the real-time decision score of the enterprise under test output by the trained decision support model, and calculating the decision score of the enterprise under test by combining the real-time decision score of the enterprise under test with the decision score of the approximate enterprise; The decision-making levels of the tested enterprises are divided based on their decision scores.

2. The financial decision support method based on big data according to claim 1, characterized in that: The historical enterprise data of multiple enterprises are obtained, and the historical enterprise data are preprocessed to obtain preprocessed enterprise data, including: Create enterprise data tables; Obtain historical enterprise data of multiple enterprises, and put all the historical enterprise data into an enterprise data table; the historical enterprise data includes multiple sub-data such as enterprise type, enterprise legal representative, enterprise shareholder, enterprise registered capital, and enterprise establishment time; Randomly select a sub-data from the enterprise data table; Select the maximum value and the minimum value contained in the sub-data, divide the sub-data evenly into multiple data segments based on the maximum value and the minimum value contained in the sub-data, and set a corresponding score for each data segment in order from small to large; Return to randomly select a sub-data from the enterprise data table until all types of sub-data in the enterprise data table have been selected, and obtain the score of each type of sub-data.

3. A financial decision support method based on big data according to claim 2, characterized in that: Create a decision support model, input the pre-processed enterprise data into the decision support model, calculate the decision score corresponding to each enterprise through the decision support model calculation combined with the historical enterprise data of each enterprise, and obtain the trained decision support model, including: Create decision support models; Divide the preprocessed enterprise data into training and test sets in random proportions; Inputting the training set into the decision support model to train the decision support model, so that the decision support model calculates the decision score corresponding to each enterprise according to the enterprise data of each enterprise, and obtains a trained decision support model; The test set is input into the trained decision support model, the predicted decision score output by the trained decision support model is obtained, and it is verified whether the predicted decision score is correct.

4. The financial decision support method based on big data according to claim 3 is characterized in that: The training set is input into the decision support model to train the decision support model, so that the decision support model calculates the decision score corresponding to each enterprise according to the enterprise data of each enterprise, and obtains the trained decision support model, including: Select an enterprise and the historical enterprise data of the enterprise from the training set; Calculate the financial score of the enterprise according to Formula 1 based on the historical enterprise data of the enterprise; Formula 1; in, is the financial score of the enterprise, is the score of the i-th sub-data of the enterprise, is the weight corresponding to the i-th sub-data of the enterprise, is the total number of sub-data; Establishing the coupling relationship between enterprise-enterprise data and decision-making scores, and sequentially inputting the coupling relationship between enterprise-enterprise data and decision-making scores into the decision support model as a training sample, so that the decision support model continuously learns the relationship between enterprise data and decision-making scores; Return to select an enterprise and its historical enterprise data from the training set until all enterprises in the training set are selected, obtain the decision score of each enterprise, and input each enterprise and its decision score as a training sample into the decision support model to train the decision support model, and obtain the trained decision support model.

5. A financial decision support method based on big data according to claim 4, characterized in that: Input the test set into the trained decision support model, obtain the predicted decision score output by the trained decision support model, and verify whether the predicted decision score is correct, including: Randomly select a company and its historical enterprise data from the test set; The decision score of the enterprise is calculated by formula 1; Inputting the historical data of the enterprise into the trained decision support model to obtain the predicted decision score output by the trained decision support model; Set a deviation threshold to determine whether the deviation between the predicted decision score and the decision score is greater than or equal to the deviation threshold; If the deviation between the predicted decision score and the decision score is greater than or equal to the deviation threshold, the trained decision support model is unqualified and the pre-processed enterprise data is divided into a training set and a test set at a random ratio until the deviation between the predicted decision score and the decision score is less than the deviation threshold; If the deviation between the predicted decision score and the decision score is less than the deviation threshold, the trained decision support model is qualified and returns to randomly select an enterprise from the test set and the historical enterprise data of the enterprise until all enterprises in the test set are selected.

6. A financial decision support method based on big data according to claim 5, characterized in that: Collect the real-time enterprise data of the tested enterprise, input the real-time enterprise data into the trained decision support model, obtain the real-time decision score of the tested enterprise output by the trained decision support model, and calculate the decision score of the tested enterprise by combining the real-time decision score of the tested enterprise with the decision score of the approximate enterprise, including: Collect real-time enterprise data; The real-time enterprise data is input into the trained decision support model, and the trained decision support model outputs the real-time decision score of the tested enterprise and the decision scores of multiple similar enterprises based on the real-time enterprise data; The decision score of the tested enterprise is calculated by combining the real-time decision score of the tested enterprise and the decision scores of multiple similar enterprises.

7. A financial decision support method based on big data according to claim 6, characterized in that: The real-time enterprise data is input into the trained decision support model, and the trained decision support model outputs the decision score of the enterprise and the decision scores of multiple similar enterprises based on the real-time enterprise data, including: Select K sub-data; record the selected K sub-data as initial target sub-data; Select an initial target sub-data, and randomly select M initial target sub-data from all enterprise data as the initial cluster centers; Calculate the distances between the remaining target sub-data and the initial cluster center respectively, and divide the enterprise data corresponding to the target sub-data to the closest initial cluster center according to the distances between the remaining target sub-data and the initial cluster center, forming M target clusters; Calculate the mean of the distances within each target cluster, and use the data point corresponding to the mean as the new cluster center; Divide enterprise data into clusters with the nearest cluster center according to the distance from the new cluster center; Set distance threshold; Calculate the distance between the new cluster center and the previous cluster center, and determine whether the distance between the new cluster center and the previous cluster center is less than or equal to the distance threshold; If the distance between the new cluster center and the previous cluster center is greater than the distance threshold, return to calculate the mean of the distances within each target cluster, and use the data point corresponding to the mean as the new cluster center until the distance between the new cluster center and the previous cluster center is less than or equal to the distance threshold; If the distance between the new cluster center and the previous cluster center is less than or equal to the distance threshold, the last clustering result is output; the enterprise corresponding to the enterprise data contained in the last clustering result is taken as an approximate enterprise; Return to select an initial target sub-data until all the initial target sub-data are selected, and get A similar enterprise.

8. A financial decision support method based on big data according to claim 7, characterized in that: The decision score of the tested enterprise is calculated by combining the real-time decision score of the tested enterprise and the decision scores of multiple similar enterprises, including: Input the enterprise data of all similar enterprises into the decision support model in sequence to obtain the decision score of each similar enterprise; Input real-time enterprise data into the decision support model to obtain real-time enterprise decision scores; The decision score of the tested enterprise is calculated by formula 2 based on the enterprise's real-time decision score and the decision scores of all similar enterprises; Formula 2: in, is the decision score of the enterprise under test, It is the real-time decision-making score of the enterprise under test. is the decision score of the ith approximate enterprise, , M is the approximate total number of enterprises.

9. A financial decision support method based on big data according to claim 8, characterized in that: The decision-making levels of the tested enterprises are divided based on their decision-making scores, including: Set the enterprise decision-making level and the decision-making score range corresponding to each enterprise decision-making level; Determine the decision score range of the tested enterprise, and set the tested enterprise to the enterprise decision level corresponding to the decision score range.

10. A financial decision support system based on big data, characterized in that: include: A data acquisition component, through which real-time enterprise data and historical enterprise data are acquired; A data processing component, through which the financial decision support method based on big data as described in any one of claims 1 to 9 is executed, the data acquisition component is communicatively connected with the data processing component, the data processing component includes a data processing unit and a data analysis unit, the data processing unit is used to process enterprise data to obtain data processing results, and the data analysis unit is used to output enterprise decisions in combination with the data processing results.

Images