Listed company risk assessment method based on quantitative index logical operation

Abstract

The invention particularly relates to a listed company risk assessment method based on quantitative index logical operation, which comprises the following steps: S1, extracting key indexes: extracting the key indexes from temporary announcements or regular announcements disclosed by listed companies; s2, determining a key index and threshold comparison operation expression; s3, constructing a logical operation model, combining the plurality of expressions in the step S2 into the logical operation model, and constructing the logical operation model representing different meanings; s4, risk tags of listed companies in the same industry are calculated in batches, threshold comparison results are calculated in batches, and results of the basic logic model A and the derivative logic model B are calculated; and S5, backtracking historical data to determine an ideal threshold expression, calculating a previous historical data result, dynamically observing a calculation result, finally obtaining a credible model, and calculating potential risks of all companies in the same industry. The method has the advantages that dependence of the model on large-scale sample data is reduced, hardware computing resources are saved, and enterprise risk assessment efficiency and accuracy are improved.

Description

Technical Field

[0001] The present invention relates to the field of risk assessment and financial technology, and in particular to a risk assessment method for listed companies based on quantitative indicator logic operations. Background Art

[0002] The current global economic situation is complex, market competition is increasingly fierce, and the risks faced by enterprises are also increasing. Accurate assessment of potential risks of enterprises is crucial to ensure the stable development of enterprises and prevent future crises. Current risk assessment methods mainly focus on analyzing the political, economic, social, technological and environmental risk factors faced by enterprises. This process involves collecting massive amounts of data and using machine learning technology to train algorithm models to predict the possibility of risk occurrence. Subsequently, the model is continuously optimized by comparing the prediction results with the preset risk threshold. However, the performance of machine learning models is highly dependent on the scale and diversity of training data, and parameters need to be constantly adjusted to achieve optimal performance, which not only consumes a lot of computing resources and time, but also because machine learning models are usually regarded as "black boxes", their decision logic is difficult to intuitively understand. Therefore, when building a risk assessment system covering listed companies across the industry, the actual application effect of machine learning methods is not ideal.

[0003] Therefore, there is an urgent need to design a risk assessment method for listed companies based on quantitative indicator logical operations, reduce dependence on sample data volume, reduce the demand for computing hardware resources, and visualize the calculation process to avoid the unknowable situation of decision logic. Summary of the invention

[0004] The purpose of the present invention is to overcome the shortcomings of the prior art and provide a risk assessment method for listed companies based on quantitative indicator logical operations, so as to reduce the dependence of the risk assessment model on large-scale sample data, save hardware computing resources, and quickly build a reliable model by analyzing historical data and transparent calculation processes, thereby improving the accuracy and calculation efficiency of the model, improving the efficiency and accuracy of enterprise risk assessment, conducting risk assessments on multiple companies, and quickly attaching risk rating labels to all listed companies in the same industry, playing an important role in the supervision of listed companies.

[0005] In order to achieve the above purpose, a risk assessment method for listed companies based on quantitative indicator logic operation is designed, the method comprising: S1. Extracting key indicators, extracting key indicators from temporary announcements or regular announcements publicly disclosed by listed companies; S2. Determining key indicator and threshold comparison operation expressions, including: S21. The threshold is set as a constant, expression 1: the first key indicator (I2) ≤ the first threshold (T2), expression 2: the second key indicator (I4) ≤ the second threshold (T4); S22. The threshold is set as an interval, expression 3: the third threshold (T11) ≤ the third key indicator (I1) ≤ the fourth threshold value (T12); S23. The threshold value is set as the algorithm, Expression 4: The fourth key indicator (I3) = the result calculated by the fifth threshold algorithm (T3); S3. Construct a logical operation model, combine the multiple expressions in step S2 into a logical operation model, and construct logical operation models representing different meanings according to the different industries in which different listed companies are located, including: S31. The logical operation model includes a basic logical model A and a derived logical model B. The operation formula of the basic logical model A is: Expression 1 and Expression 4 are "or" operated, and the operation formula of the derived logical model B is: Expression 3, basic Logical model A and expression 2 are operated with "AND"; S4. Batch calculate the risk labels of listed companies in the same industry. According to the industry classification of the China Securities Regulatory Commission, first, the quantitative indicator values ​​of N listed companies in the industry are batch extracted according to the method of step S1. Secondly, the threshold comparison results of these listed companies are batch calculated according to the expression of step S2 to obtain the list results. Finally, the results of basic logical model A and derived logical model B are calculated according to the logical operation model of step S3; S5. Go back to historical data to determine the ideal threshold expression, and use the set logical operation model to calculate the historical data results of N listed companies in the industry in the past few years. If the listed company hit by the calculation results of basic logical model A and derived logical model B has experienced actual risks in history and the calculation results are consistent with the actual risks, the model is considered to be credible. If the calculation results are inconsistent with the historical actual risks, by tracing the calculation results of each layer, combined with the horizontal comparison method between listed companies and multiple copies of historical data of listed companies, the key threshold that causes the distortion of the results is found and adjusted. After adjustment, the calculation is repeated, and the calculation results are dynamically observed. Finally, a credible model is obtained, and the potential risks of all listed companies in the same industry are calculated based on the credible model.

[0006] The method of the present invention further includes: in the step S2, the actual value Ix of each key indicator is compared with a preset threshold value Tx. If it exceeds or falls below a certain threshold value, it is considered that a risk condition is triggered.

[0007] The method of the present invention also includes: the step S3 constructs a logical operation model, combines the key indicators and threshold comparison operation expressions to construct a logical operation model, the model includes an input layer, a logical operation layer and an output layer, the input layer is the indicator and threshold comparison operation expression, the logical operation layer is the "and" logical operation and "or" logical operation between multiple input layers, and the output layer is the result of the operation. Different operation results can be obtained by adjusting the indicator threshold in the expression and the logical operator between the expressions, and a logical operation model representing different risk types is constructed in combination with the industry characteristics of the listed company.

[0008] Compared with the prior art, the present invention has the following advantages: 1. Comprehensive risk assessment: The present invention can process any combination of multiple quantitative indicators, including risk factors in different fields such as legal, financial and operational, extracted from the announcements of listed companies, making the risk assessment results more comprehensive.

[0009] 2. Saving of data resources and hardware computing power: The present invention directly extracts quantitative indicator data from company announcements, does not rely on large-scale diversified data for model training, and can accurately set key thresholds, thereby saving computing resources and time.

[0010] 3. Efficient model training: The present invention adopts hierarchical calculation factor design and traceable calculation process, which makes the decision logic easy to understand and facilitates the adjustment of operation factors, such as expression thresholds and logical operators, to quickly obtain the optimal model.

[0011] 4. Improve accuracy and credibility: This invention optimizes the key calculation factors of the logic model through historical data backtracking, enhances the accuracy of the model, and ensures that all companies in the industry are assessed according to unified risk assessment standards, thereby improving the credibility of risk assessment results. BRIEF DESCRIPTION OF THE DRAWINGS

[0012] Figure 1 , is the overall framework diagram of the present invention; Figure 2 , is a schematic diagram of a logic operation model of the present invention; Figure 3 , is a schematic diagram of tracing the logical operation results of the present invention. DETAILED DESCRIPTION

[0013] In order to make the purpose, principle and structure of the present invention more clear, it is further described below in conjunction with the drawings and specific embodiments.

[0014] The present invention provides a method for automatically extracting key quantitative indicators from the disclosure announcements of listed companies, and then constructing a logical operation model of multiple risks based on multiple indicators and threshold comparison operation expressions. The overall framework is as follows: Figure 1As shown. This method does not need to rely on a large amount of sample data, saves hardware computing resources, and can quickly obtain a credible model by tracing back historical data and traceable calculation processes. The present invention aims to improve the efficiency and accuracy of corporate risk assessment, and can batch label all listed companies in the same industry with risk assessment labels. It also facilitates horizontal comparative analysis between listed companies in the same industry, and can be well used in the supervision of listed companies.

[0015] Embodiment 1: In this embodiment, the method described uses information such as temporary announcements or regular announcements publicly disclosed by listed companies to extract the required indicators and set threshold comparison expressions. After obtaining multiple expression results, the risk assessment results of multiple listed companies in the same industry are calculated in batches by building a logical operation model between the expressions. The detailed steps are as follows:

[0016] Step S1: Extract key indicators. Taking the secondary industry classification of the China Securities Regulatory Commission, "Manufacturing-Food Manufacturing", as an example, assume that the key indicators of financial fraud risk of listed companies in this industry are main operating income (I1), sales expenses (I2), sales expense growth rate (I3), and non-net profit (I4). The above indicators can be extracted from the semi-annual and annual reports publicly disclosed by listed companies in the "Periodic Report" column of the official website of the Shanghai Stock Exchange.

[0017] Specifically: 1. In the "Operating Income and Operating Costs" table under the "Notes to Consolidated Financial Statements" in the annual report announcement, the "Current Period Amount - Income" indicator value of the standard item "Main Business" of the listed company can be extracted, which is the main business income (I1). 2. Sales expenses (I2) can be extracted from the "Consolidated Income Statement". 3. Sales expense growth rate (I3) = (sales expenses - sales expenses in the same period of the previous year) / sales expenses in the same period of the previous year. 4. In the "Main Accounting Data" table in the "Company Profile and Main Financial Indicators" chapter, extract "Net profit attributable to shareholders of listed companies after deducting non-recurring gains and losses", which is the net profit after deducting non-recurring gains and losses (I4).

[0018] Step S2: Determine the indicator and threshold comparison operation expression. Compare the actual value Ix of each indicator with the preset threshold Tx. If it exceeds or falls below a certain threshold, it is considered that the risk condition is triggered and the risk assessment result is recorded as "true", otherwise it is recorded as "false".

[0019] The S2.1 threshold is set as a constant, assuming that expression 1: sales expenses (I2) ≤ industry median (T2), expression 2: non-net profit (I4) ≤ industry mean (T4).

[0020] S2.2 The threshold is set as an interval, assuming Expression 3: The minimum value of the confidence interval (T11) ≤ main operating income (I1) ≤ the maximum value of the confidence interval (T12).

[0021] The S2.3 threshold is set as an algorithm, assuming Expression 4: Sales expense growth rate (I3) = the top 5 in the secondary classification of the industry of the China Securities Regulatory Commission where the company is located (T3).

[0022] Step 3: Construct a logical operation model. Figure 2 As shown, multiple expressions in step 2 are combined into a logical operation model according to certain rules. According to the industry characteristics of the listed company, a logical operation model representing different meanings can be constructed.

[0023] S3.1 Assume that the logical operation model for financial fraud risk assessment in the "manufacturing-food manufacturing" industry to be constructed includes "basic logical model A" and "derivative logical model B". 1. The operation formula of the basic logical model A is: "Expression 1" and "Expression 4" are "OR" operations. 2. The operation formula of the derived logical model B is: "Expression 3" "Basic logical model A" and "Expression 2" are "AND" operations.

[0024] S3.2 The calculation result of "Basic Logic Model A" is defined as the evaluation result of the listed company's business risk label "Abnormal Growth of Sales Expenses", and the calculation result of "Derived Logic Model B" is defined as the evaluation result of the listed company's risk level label "High Risk of Financial Fraud".

[0025] Step S4: Batch calculate the risk labels of listed companies in the same industry. According to the industry classification of the China Securities Regulatory Commission, assuming that there are N listed companies in the "manufacturing-food manufacturing" industry, first, the quantitative index values ​​of the N listed companies are batch extracted according to the method of step 1, and then the threshold comparison results of these listed companies are batch calculated according to the expression of step 2. The obtained list results include company names, expression names, and expression calculation results. Finally, the results of "basic logic model A" and "derivative logic model B" are calculated according to the logical operation model of step 3.

[0026] S4.1 first batch extracts the main operating income (I1), sales expenses (I2), sales expense growth rate (I3), and non-net profit (I4) index values ​​of N listed companies. Then, the index values ​​are substituted into multiple expressions in step 2 to calculate the expression results of each listed company.

[0027] S4.2 then brings the expression result into the “Basic Logical Model A” to obtain the risk assessment result of the business risk label “Abnormal Growth of Sales Expenses” of N listed companies.

[0028] S4.3 After calculating the business risk label results, bring the results into the "derivative logic model B" to complete the risk assessment of the risk level label "high risk of financial fraud".

[0029] Step S5: Review historical data to determine the ideal threshold expression. Figure 3 As shown in the figure, the historical data results of N listed companies in the industry in the past few years are calculated using the set logical model. If the listed companies hit by the calculation results of "Basic Logical Model A" and "Derived Logical Model B" have experienced factual risks in history and the calculation results are consistent with the factual risks, the model is considered credible. If the calculation results do not match the historical factual risks, the key thresholds that cause the results to be distorted are found and adjusted. After the adjustment, the calculation is repeated, and the calculation results are dynamically observed to finally obtain a credible model. Specifically, Figure 3 As shown in the derivative logic model B, the output of the derivative logic model B is false. By tracing back the calculation results of each layer and combining the horizontal comparison method between listed companies and multiple historical data of listed companies, it is determined that the output of expression 2 is false, thereby finding the key threshold T4 that causes the result distortion, adjusting T4 from 0.7 to 0.8, and then calculating again. At this time, the output of the derivative logic model B is true, and a credible model is obtained. The potential risks of all listed companies in the same industry are calculated based on the credible model. It is worth noting that the factual risk refers to an event that actually occurred in the history of a listed company, and the event is sufficient to trigger a risk condition.

[0030] See also Figure 3 As one specific embodiment, the definitions of indicators, thresholds, expressions, basic logic model A, and derived logic model B can be found in steps S1 to S3 of the aforementioned implementation case.

[0031] Before threshold adjustment: Enter the values ​​of indicators I1, I2, I3, I4 and the values ​​of thresholds T1, T11, T12, T3, T4 for calculation. First, we get: the result of expression 1 is "false", the result of expression 4 is "true", the result of expression 2 is "false", and the result of expression 3 is "true". Secondly, the basic logic model A is obtained by performing an "or" operation between "expression 1" and "expression 4", and the result is "true". Finally, the derived logic model B is obtained by performing an "and" operation between "basic logic model A", "expression 2" and "expression 3", and the result is "false". The result of "false" indicates that there is no risk for the listed company.

[0032] Combining the horizontal comparison method between listed companies and between multiple historical data of listed companies, it is found that the result of "there is no risk for listed companies" calculated by this model is inconsistent with the actual situation of risks that have occurred in the history of listed companies. Risk events have actually occurred in listed companies, but through the calculation results of this model, the result is not "true", that is, the result of "there is risk for listed companies" is not obtained. This shows that the derivative logic model B before the threshold adjustment is not credible.

[0033] Find the critical threshold T4 that causes the result distortion, adjust T4 from 0.7 to 0.8, and then calculate again.

[0034] After the threshold is adjusted, the result of "Expression 2" changes from "false" to "true", and the result of derived logic model B is "true". The result of "true" indicates that the listed company has risks, which is consistent with the actual risk events that have occurred. This shows that the model after the threshold adjustment is a credible model.

[0035] It is worth noting that "result hit" means that the result calculated by the derived logic model B is "true". Combined with step S4 of "Implementation Example 1", the basic logic model A is the risk of "abnormal growth of sales expenses", and the derived logic model B is the risk of "high risk of financial fraud". If the value obtained by calculating the data of a certain year in the history of a listed company is "true", and it is found through public regulatory cases that the company has indeed encountered the corresponding risk in this year, it is determined to be a result hit.

[0036] The above description is only a specific implementation mode of the present invention, but the protection scope of the present invention is not limited thereto. Any technician familiar with the technical field can make equivalent substitutions or changes according to the technical solutions and novel concepts of the present invention within the technical scope disclosed by the present invention, which should be covered by the protection scope of the present invention.

Claims

1. A risk assessment method for listed companies based on quantitative index logic operation, characterized in that: The method comprises: S1. Extract key indicators from temporary announcements or regular announcements publicly disclosed by listed companies; S2. Determine the key indicators and threshold comparison operation expressions, including: S21. The threshold is set to a constant, Expression 1: the first key indicator (I2) ≤ the first threshold (T2), Expression 2: the second key indicator (I4) ≤ the second threshold (T4); S22. The threshold is set as an interval, Expression 3: the third threshold (T11) ≤ the third key indicator (I1) ≤ the fourth threshold (T12); S23. The threshold is set as an algorithm, Expression 4: the fourth key indicator (I3) = the result of the fifth threshold algorithm calculation (T3); S3. Construct a logical operation model, combine the multiple expressions in step S2 into a logical operation model, and construct logical operation models representing different meanings according to the different industries in which different listed companies are located, including: S31. The logic operation model includes a basic logic model A and a derived logic model B. The operation formula of the basic logic model A is: Expression 1 and Expression 4 are "or" operations, and the operation formula of the derived logic model B is: Expression 3, the basic logic model A and Expression 2 are "and" operations; S4. Batch calculate the risk labels of listed companies in the same industry. According to the industry classification of the China Securities Regulatory Commission, first, batch extract the quantitative index values ​​of N listed companies in the industry according to the method of step S1, then batch calculate the threshold comparison results of these listed companies according to the expression of step S2 to obtain the list results, and finally calculate the results of the basic logic model A and the derived logic model B according to the logical operation model of step S3; S5. Look back at historical data to determine the ideal threshold expression, and use the set logical operation model to calculate the historical data results of N listed companies in the industry in the past few years. If the listed companies hit by the calculation results of the basic logical model A and the derived logical model B have experienced actual risks in history and the calculation results are consistent with the actual risks, the model is considered credible. If the calculation results do not match the historical actual risks, by tracing the calculation results of each layer and combining the horizontal comparison method between listed companies and between multiple copies of historical data of listed companies, find the key threshold that causes distorted results and adjust it. After adjustment, calculate again and dynamically observe the calculation results to finally obtain a credible model. Based on the credible model, calculate the potential risks of all listed companies in the same industry.

2. A risk assessment method for listed companies based on quantitative index logic operation as claimed in claim 1, characterized in that: In step S2, the actual value Ix of each key indicator is compared with a preset threshold value Tx. If it exceeds or falls below a certain threshold value, it is considered that a risk condition is triggered.

3. A risk assessment method for listed companies based on quantitative index logic operation as claimed in claim 1, characterized in that: The step S3 constructs a logical operation model, and combines the key indicators and threshold comparison operation expressions to construct a logical operation model. The model includes an input layer, a logical operation layer and an output layer. The input layer is the indicator and threshold comparison operation expression, the logical operation layer is the "AND" logical operation and "OR" logical operation between multiple input layers, and the output layer is the risk name of the result obtained by the operation. Different operation results can be obtained by adjusting the indicator threshold in the expression and the logical operator between the expressions. Combined with the industry characteristics of the listed company, a logical operation model representing different risk types is constructed.

Images