Financial risk assessment method and system fusing network information and macroscopic data

Abstract

The invention relates to the technical field of financial science and technology and big data analysis, in particular to a financial risk assessment method and system fusing network information and macroscopic data, and the method comprises the steps: obtaining heterogeneous data of a target cooperative country; performing quantification processing on the social network information text data to generate at least one social network information risk index; performing standardization processing on the macroscopic index data; fusing the social network information risk indexes and the standardized macroscopic index data, and constructing a weighted decision matrix; and calculating and outputting a financial cooperation risk index for the target cooperation country by adopting a comprehensive evaluation model based on the weighted decision matrix. According to the method, network information data are required to be from different types of media information sources, are quantified as independent risk dimensions, and are fused with traditional macroscopic data, so that objective economic basic risks and subjective social dynamic risks can be captured at the same time, a final risk assessment result is more comprehensive and stereoscopic, and the risk assessment efficiency is improved. And the accuracy and depth of evaluation are effectively improved.

Description

Technical Field

[0001] This application relates to the fields of financial technology and big data analysis technology, and in particular to a financial risk assessment method and system that integrates network information and macro data. Background Technology

[0002] Against the backdrop of deepening global economic integration and regional cooperation, transnational financial cooperation has become a key force driving economic development and optimizing resource allocation. However, such cooperation is often accompanied by complex financial risks. Therefore, establishing a scientific, comprehensive, and dynamic financial risk assessment system is of vital technical value and practical significance for ensuring the security and sustainability of financial cooperation and improving decision-making efficiency.

[0003] To address the aforementioned issues, existing technologies have attempted to incorporate social network information data into risk assessment models. However, the applicant's research revealed significant technical shortcomings in practical applications. First, existing technologies typically treat network information data as a homogeneous pool of information, failing to effectively differentiate the authority and influence of various sources. Second, existing technologies employ a limited analytical dimension for network information, largely relying on simple judgments of "positive" or "negative" sentiment, neglecting the more complex dynamic risks within the network information field. Therefore, how to differentiate and weight heterogeneous network information sources, extract deeper risk indicators beyond single emotional polarities from network information data, and then systematically integrate these indicators with traditional macroeconomic indicators to construct a more comprehensive and accurate financial risk assessment model is a pressing technical problem that needs to be solved in this field. Summary of the Invention

[0004] The objective of this application is to provide a financial risk assessment method that integrates online information and macroeconomic data, comprising: acquiring heterogeneous data from a target partner country, wherein the heterogeneous data includes structured macroeconomic indicator data and unstructured social network information text data, wherein the social network information text data originates from at least two different types of media sources; quantifying the social network information text data to generate at least one social network information risk indicator; standardizing the macroeconomic indicator data; integrating the social network information risk indicator with the standardized macroeconomic indicator data to construct a weighted decision matrix; and calculating and outputting a financial cooperation risk index for the target partner country based on the weighted decision matrix using a comprehensive evaluation model.

[0005] By adopting the above technical solutions, it is clearly required that network information data come from different types of media sources and be quantified as an independent risk dimension. Then, it is integrated with traditional macroeconomic data, thereby simultaneously capturing objective economic fundamental risks and subjective social dynamic risks. This makes the final risk assessment results more comprehensive and three-dimensional, effectively improving the accuracy and depth of the assessment.

[0006] Optionally, the step of quantifying the social network information text data specifically includes: classifying the media sources into official or mainstream media as the first type of media sources and social media or online forums as the second type of media sources; calculating a first sentiment index for the network information text data of the first type of media sources; calculating a second sentiment index for the network information text data of the second type of media sources; and calculating an information differentiation risk index based on the difference between the first sentiment index and the second sentiment index, as one of the social network information risk indicators.

[0007] By adopting the above technical solution, and by distinguishing between mainstream media and non-mainstream media and calculating the difference in sentiment index between the two, the important social risk of information differentiation is quantified. The level of the information differentiation risk index directly reflects the degree of consistency between official information and public sentiment, and can reveal potential social tensions or policy communication barriers that traditional sentiment analysis cannot detect, providing a more penetrating perspective for risk warning.

[0008] Optionally, the calculation steps of the first sentiment index and the second sentiment index include: establishing a media authority weighting system, classifying the media sources into different authority levels and assigning corresponding weight coefficients; performing sentiment scoring on the online information text data of each media source to obtain a basic sentiment score; and weighting and aggregating the basic sentiment score with the corresponding weight coefficient to calculate the first sentiment index and the second sentiment index respectively.

[0009] By adopting the above technical solution and introducing a media authority weighting system, the shortcomings of existing technologies in processing homogeneous network information sources are solved. This weighting system allows emotional signals from media with different influences to be differentiated and included in the final emotional index, making the quantitative results more consistent with the power structure and information influence distribution of the network information field in the real world. This refined processing method greatly improves the real-world fit of the emotional index and the reliability of the evaluation results.

[0010] Optionally, the step of performing sentiment scoring on the online information text data of various media sources includes: processing sentences in the online information text data based on a preset sentiment dictionary, degree adverb dictionary, and negation word dictionary; identifying positive sentiment words, negative sentiment words, degree modifiers, and negation words in the sentences; and calculating the basic sentiment score of the sentences according to preset scoring rules and in combination with the identified words.

[0011] By adopting the above technical solution, a specific, reproducible, and interpretable sentiment scoring technical path is provided. Compared with complex deep learning black box models, the dictionary-based method has a clear and transparent computational logic and is easy to customize and optimize by supplementing professional vocabulary in the field of financial cooperation, ensuring the professionalism and accuracy of sentiment analysis results in specific application scenarios.

[0012] Optionally, before calculating the first sentiment index and the second sentiment index, the following steps are also included: collecting the online information text data from the Internet using web crawler technology; constructing a keyword library containing national identifiers and bilateral financial cooperation business terms, as well as an exclusion keyword library; and using the keyword library to filter the collected online information text data to extract text related to the financial cooperation theme.

[0013] By adopting the above technical solutions, the data quality of the input network information analysis model is ensured, the data acquisition is achieved efficiently through automated crawling technology, and the precise screening through a carefully constructed keyword library can filter out a large amount of irrelevant noise from massive amounts of Internet information, greatly improving the relevance and purity of the corpus.

[0014] Optionally, the macroeconomic indicator data includes at least one of the following categories: governance stability indicators, including the degree of governance stability and regulatory quality; social stability indicators, including the rule of law and the global peace index; domestic economic stability indicators, including debt repayment capacity, per capita GDP and GDP growth rate; and trade and investment stability indicators, including net inflows of foreign direct investment and the proportion of bilateral trade to GDP.

[0015] By adopting the above technical solutions, the macroeconomic data dimensions that form the basis of risk assessment are clarified. By covering multiple levels such as governance, society, economy, trade and investment, the comprehensiveness of the assessment model is ensured, so that the final risk index can be based on a systematic examination of the macroeconomic fundamentals of the target country, avoiding assessment bias that may be caused by single-dimensional data.

[0016] Optionally, the step of quantifying the social network information text data further includes: obtaining policy implementation indicator data reflecting the strength of policy implementation in the target cooperative country; standardizing the policy implementation indicator data and the social network information sentiment index from non-mainstream media; calculating the absolute difference between the standardized policy implementation indicator and the social network information sentiment index to generate a policy expectation deviation index, which serves as one of the social network information risk indicators.

[0017] By adopting the above technical solutions, the policy expectation deviation index can effectively identify the risk of unmet public expectations caused by information asymmetry, insufficient policy publicity, or poor implementation by quantifying the gap between public sentiment and actual policy outcomes. It can provide early warning of potential trust crises before negative events actually occur, providing policymakers with a valuable window of opportunity for risk intervention.

[0018] Optionally, the policy implementation indicator data includes at least one of the following: the number of bilateral investment projects with the target partner country; the turnover of contracted projects completed in the target partner country; and the proportion of bilateral aid to the recipient country's GDP.

[0019] By adopting the above technical solution, specific and measurable data inputs are provided for the calculation of the policy expectation deviation index. By clarifying these quantitative indicators that can directly reflect the strength and effectiveness of cooperation, the originally abstract concept of "policy implementation" becomes operational, enhancing the calculability and practical significance of the risk index.

[0020] Optionally, the step of calculating and outputting the financial cooperation risk index for the target cooperative countries using a comprehensive evaluation model specifically involves using the entropy weight-optimal solution distance method, including: calculating the weights of the social network information risk index and the macroeconomic indicator data using the entropy weight method, and constructing a weighted decision matrix based on the weights; determining the optimal and worst solutions based on the weighted decision matrix; calculating the Euclidean distance between the evaluation vector of each target cooperative country and the optimal and worst solutions; calculating the relative proximity based on the Euclidean distance, and using the relative proximity as the final financial cooperation risk index.

[0021] By adopting the above technical solutions, a scientific data fusion and comprehensive evaluation method is provided. The entropy weight method can objectively determine the importance of each indicator in the evaluation system based on the information content of each indicator data, avoiding the bias caused by subjective weighting. The superior-inferior solution distance method can rank and evaluate the superiority and inferiority of multiple evaluation objects by calculating the distance with the ideal solution. The combination of the two ensures the objectivity and fairness of the final evaluation results.

[0022] The second objective of this application is to provide a financial risk assessment system that integrates network information and macroeconomic data, comprising: a data acquisition module for acquiring heterogeneous data from a target partner country, the heterogeneous data including structured macroeconomic indicator data and unstructured social network information text data; a social network information processing module for quantifying the social network information text data to generate at least one social network information risk indicator, wherein the social network information risk indicator includes an information differentiation risk index obtained by calculating the sentiment differences between different information sources, and a policy expectation deviation index obtained by calculating the expectation difference between public sentiment and policy implementation; a macroeconomic data processing module for standardizing the macroeconomic indicator data; and a risk assessment module configured to integrate the social network information risk indicator generated by the social network information processing module and the macroeconomic indicator data processed by the macroeconomic data processing module, and to calculate and output a financial cooperation risk index for the target partner country using a comprehensive evaluation model. Attached Figure Description

[0023] Figure 1 This is a schematic diagram of the financial risk assessment method that integrates network information and macroeconomic data in this application.

[0024] Figure 2 This is a block diagram of the financial risk assessment system that integrates network information and macroeconomic data, as described in this application. Detailed Implementation

[0025] To make the objectives, technical solutions, and advantages of this application clearer, the technical solutions of this application will be clearly and completely described below in conjunction with specific embodiments. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of them. Based on the embodiments in this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0026] like Figure 1 As shown in the figure, this application provides a financial risk assessment method that integrates network information and macroeconomic data, including the following steps.

[0027] S01: Obtain heterogeneous data from the target partner countries. The heterogeneous data includes structured macroeconomic indicator data and unstructured social network information text data, wherein the social network information text data comes from at least two different types of media sources.

[0028] Understandably, obtaining structured macroeconomic indicator data can be achieved by calling a pre-defined application programming interface (API). Specifically, the system initiates requests following the Hypertext Transfer Protocol (HTTP) to the official data release platforms of the World Bank, the International Monetary Fund, and target partner countries (such as a national statistics bureau). For example, to obtain a country's GDP growth rate, the system constructs a request pointing to a specific address of the World Bank's data interface. The parameters of this request specify the country code to be obtained, the indicator to be the GDP market price fixed-base growth rate, and specify that the returned data is in a generic data exchange format with a time range of 2010 to 2023. The server executes such tasks periodically (e.g., quarterly). The obtained data is parsed and stored in a table in a relational database management system used to store macroeconomic data. The table structure is designed to include fields such as country code, indicator name, year, value, and unit. For example, a record might contain the following information: country code representing a country, indicator name GDP growth rate, year 2023, value 0.29 percent, and unit as percentage.

[0029] For acquiring unstructured social network information text data, a cluster based on a distributed web crawler framework can be used. This cluster consists of multiple physical servers, each running multiple crawler processes to collect different types of media source data. These sources are clearly divided into at least two categories, such as official or mainstream media as the first type of media source and social media or online forums as the second type of media source.

[0030] Specifically, the first type of media sources targets websites including national-level media and highly authoritative financial media. The crawler is configured to access specific sections of these websites (such as "International" and "Finance") daily at set times to capture news headlines, text, publication time, sources, and other information. The second type of media sources targets platforms including social media and online forums. The system simulates logins and calls to their open application programming interfaces (within permitted permissions) to capture post content, author information, reposts, and comment counts under specific topics. For Q&A communities, the crawler focuses on capturing questions and answers related to keywords such as investment and debt. To circumvent anti-crawler mechanisms, the crawler cluster is configured with a dynamic Internet Protocol (IP) proxy pool and randomized user agent headers to simulate normal user access behavior.

[0031] All the raw text data captured, along with its metadata such as Uniform Resource Locators, capture time, and source type identifiers, is stored in a document-oriented, non-relational database within a collection of web information data for flexible text processing.

[0032] S02: Quantify the text data of social network information to generate at least one social network information risk indicator.

[0033] Specifically, this step includes: classifying media sources into official or mainstream media as the first type of media source and social media or online forums as the second type of media source; calculating a first sentiment index for the online information text data of the first type of media source; calculating a second sentiment index for the online information text data of the second type of media source; and calculating an information differentiation risk index based on the difference between the first and second sentiment indices, as one of the social network information risk indicators.

[0034] Understandably, before calculating the first and second sentiment indices, the following steps are also included: collecting online information text data from the Internet using web crawler technology; constructing a keyword library containing national identifiers and bilateral financial cooperation business terms, as well as an exclusion keyword library; and using the keyword library to filter the collected online information text data to extract text related to the financial cooperation theme.

[0035] Understandably, after data collection, the system needs to clean and filter the massive amount of raw data to ensure the accuracy of subsequent analysis. This data is then used to construct and store in a keyword database, a country identifier database, a bilateral financial cooperation business database, and an exclusion keyword database. The keyword database is predefined by domain experts and can be dynamically updated. The country identifier database includes the full name, abbreviation, and major cities of the target country; for example, for a specific country, the database includes relevant identifiers. The bilateral financial cooperation business database contains various terms related to financial cooperation, such as "bilateral investment," "currency swap," "debt restructuring," "preferential loans," and "infrastructure financing." The exclusion keyword database contains terms that may cause confusion but are unrelated to financial cooperation, such as "cultural exchange," "sports events," "tourist visas," and "general diplomatic visits."

[0036] Specifically, the system iterates through each document in the collection used to store network information data and applies a Boolean logic rule to determine whether a document is relevant. A document is considered relevant only if its text content simultaneously meets the following three conditions: 'contains at least one country identifier', 'contains at least one bilateral financial cooperation business term', and 'does not contain any exclusion keywords'. Only documents deemed relevant are retained and marked as relevant corpus for further processing. In this way, a large amount of irrelevant information can be effectively filtered out, greatly improving the signal-to-noise ratio.

[0037] The filtered relevant data is automatically split into two processing pipelines based on the source type identifier at the time of collection: the mainstream media pipeline and the non-mainstream media pipeline; the system will calculate the sentiment index for the data in these two pipelines respectively.

[0038] To calculate the sentiment index, the calculation steps for the first and second sentiment indices include: establishing a media authority weighting system, classifying media sources into different authority levels and assigning corresponding weight coefficients; scoring the sentiment of the online information text data of each media source to obtain a basic sentiment score; and weighting and aggregating the basic sentiment score with the corresponding weight coefficients to calculate the first and second sentiment indices respectively.

[0039] The media authority weighting system is a major innovation of this application. The system internally maintains a media authority weighting configuration table, which maps specific media sources or user identities to an authority level and weight coefficient. For example, this embodiment adopts the following six-level weighting system.

[0040] Central media (weight coefficient 1.5): given the highest weight because it directly reflects the national will and the policy orientation of the highest level.

[0041] Financial media (weighting coefficient 1.3): This category has a high weighting because it possesses professionalism and depth in the financial field and has a strong guiding influence on professionals.

[0042] High-influence personal media (weight coefficient 1.2): The weight is higher than that of ordinary media because its views spread quickly and widely, and can quickly ignite online information.

[0043] Local mainstream media (weight coefficient 1.0): set as the baseline weight, representing the official or semi-official voice in the region.

[0044] Portal websites (weight coefficient 0.8): They have a low weight because their content is mostly reprinted and lacks originality. Their influence is mainly reflected in traffic rather than in leading viewpoints.

[0045] Ordinary social media users or forum users (weight coefficient 0.3): Represent grassroots public opinion, they are numerous but their opinions are scattered and their individual influence is weak, so they are given the lowest weight.

[0046] When processing each document, the system will query this table based on the source address or author's identity to obtain its weight coefficient.

[0047] Understandably, the steps for sentiment scoring of online information text data from various media sources include: processing sentences in the online information text data based on a pre-set sentiment dictionary, degree adverb dictionary, and negation word dictionary; identifying positive sentiment words, negative sentiment words, degree modifiers, and negation words in the sentences; and calculating the basic sentiment score of the sentences according to pre-set scoring rules and the identified words.

[0048] Specifically, the system first segments the document into sentences, and for each sentence, it performs dictionary loading, word recognition, and scoring operations.

[0049] The dictionary loading process specifically includes: loading a positive sentiment dictionary, a negative sentiment dictionary, a degree adverb dictionary, and a negation word dictionary. The positive sentiment dictionary merges with a general sentiment dictionary and is supplemented by experts with positive words related to financial cooperation, such as "win-win", "opportunity", "milestone", and "inject vitality". The negative sentiment dictionary is supplemented with negative words such as "debt trap", "default risk", "predatory", and "opaque". The degree adverb dictionary is divided into different levels, such as "very", "extremely", and "quite" (corresponding to a multiplier of 1.5), "relatively", "slightly", and "somewhat" (corresponding to a multiplier of 0.8). The negation word dictionary includes words such as "not", "no", "not", and "absolutely not" (corresponding to a multiplier of -1).

[0050] The word recognition and scoring process specifically includes: the system segments the sentence into words and matches them with the dictionary. The final score of the sentence is calculated as follows: first, the base scores (valued as positive 1) of all identified positive sentiment words and the base scores (valued as negative 1) of all identified negative sentiment words are summed. Then, this sum is multiplied by the multipliers corresponding to all identified negative words and all degree adverbs in the sentence. Finally, the product is divided by the sentence length plus one and the logarithm is taken. Dividing by the logarithm of the sentence length is to adjust the sentiment density of long sentences and avoid long sentences from scoring too high due to the accumulation of sentiment words.

[0051] After sentence scoring is completed, the total sentiment score of a document is a weighted average of the scores of all its sentences, with the title sentence having a higher weight.

[0052] Finally, a weighted aggregation is performed to calculate the first sentiment index and the second sentiment index.

[0053] The first sentiment index is calculated by multiplying the sentiment score of each document in the mainstream media channel by its corresponding media weight, summing the results, and then dividing by the sum of the media weights of these documents. The second sentiment index is calculated in the same way, but the summation range includes all documents in non-mainstream media channels.

[0054] Ultimately, the information divide risk index is calculated by taking the absolute value of the difference between the first sentiment index and the second sentiment index. This value is output as one of the social network information risk indicators. A higher value, such as 0.8, means that mainstream media are generally optimistic (index of 0.6), while social media is generally pessimistic (index of -0.2). This reveals a huge information gap and is a high-risk signal that urgently needs attention.

[0055] Understandably, the steps for quantifying social network information text data also include: obtaining policy implementation indicator data reflecting the strength of policy implementation in target cooperative countries; standardizing the policy implementation indicator data and the sentiment index of social network information from non-mainstream media; calculating the absolute difference between the standardized policy implementation indicators and the sentiment index of social network information to generate a policy expectation deviation index, which serves as one of the social network information risk indicators.

[0056] The policy expectation deviation index compares "public sentiment" with "official actions." "Public sentiment" is represented by the second sentiment index (i.e., the non-mainstream media sentiment index), while "official actions" are quantified by a series of specific policy implementation indicators.

[0057] To this end, the system first needs to obtain these policy implementation indicator data, which include at least one of the following: the number of bilateral investment projects with the target partner country; the turnover of contracted projects completed in the target partner country; and the proportion of bilateral aid to the recipient country's GDP.

[0058] Specifically, this data can be collected from specific data sources. For example, the number of bilateral investment projects comes from the annual foreign direct investment bulletins of relevant departments and the global development finance database, and the data type is integer; the completed turnover of contracted projects also comes from the statistical data of relevant departments, reflecting the actual progress of infrastructure construction and other projects, and the data type is floating point, with the unit being US$100 million; the proportion of bilateral aid to the GDP of the recipient country is calculated by combining the publicly available data of relevant departments and the fiscal data of the recipient country, reflecting the actual strength of the aid and its relative importance to the economy of the recipient country.

[0059] After obtaining the raw values ​​of these policy implementation indicators and the non-mainstream media sentiment index, since the two have completely different scales, they must be standardized. The range standardization method can be used to map the raw values ​​of the policy implementation indicators and the non-mainstream media sentiment index to the range of 0 to 1, so as to obtain the standardized policy implementation indicator values ​​and the standardized non-mainstream media sentiment index values.

[0060] Finally, the policy expectation deviation index is calculated by taking the absolute value of the difference between the standardized policy implementation indicator value and the standardized non-mainstream media sentiment index value. The interpretation of this index is extremely valuable. For example, if the standardized index representing public sentiment is 0.9 (extremely optimistic), while the standardized index representing actual investment is only 0.2 (far lower than other countries), the policy expectation deviation index is as high as 0.7. This indicates that the public has excessively high and unrealistic expectations for cooperation. Once these expectations are not met, they may quickly turn into disappointment and negative emotions, posing a potential risk. Conversely, if the standardized index of public sentiment is 0.1 (extremely pessimistic), while the standardized index of actual investment is 0.8 (huge actual investment), the policy expectation deviation index is also very high. This indicates that the official efforts have not been recognized or acknowledged by the public, and cooperative projects may lack public support, facing a "thankless" predicament. Therefore, the policy expectation deviation index becomes a key quantitative tool for measuring the effectiveness of policy communication and managing social expectations.

[0061] S03: Standardize macroeconomic indicator data.

[0062] Specifically, since macroeconomic indicator data have different dimensions and units, and different indicators contribute to risk in different directions, standardization is necessary to eliminate these differences. In this embodiment, the range standardization method can be used.

[0063] For positive indicators, the larger the value, the higher the risk of financial cooperation. The standardized value is calculated by subtracting the minimum value of the indicator from the original value of the indicator among all the assessed objects, and then dividing the difference by the difference between the maximum and minimum values ​​of the indicator among all the assessed objects.

[0064] For negative indicators, the larger the value, the lower the risk of financial cooperation. For example, the standardized value of GDP per capita is calculated by subtracting the original value of the indicator from the maximum value of the indicator among all the assessed objects, and then dividing the difference by the difference between the maximum and minimum values ​​of the indicator among all the assessed objects.

[0065] After processing, all macroeconomic indicators were converted into dimensionless values ​​between 0 and 1, with larger values ​​representing higher risks. These standardized data constituted part of the assessment matrix.

[0066] S04: Integrate social network information risk indicators with standardized macroeconomic indicator data to construct a weighted decision matrix.

[0067] S05: Based on the weighted decision matrix, a comprehensive evaluation model is used to calculate and output the financial cooperation risk index for the target partner country.

[0068] Specifically, the social network information risk indicators (such as the information differentiation risk index and the policy expectation deviation index) are first combined with standardized macroeconomic indicator data to form an initial decision matrix.

[0069] Next, the entropy weight method is used to determine the objective weight of each indicator. The core idea of ​​the entropy weight method is that the greater the numerical difference of an indicator among different evaluation objects, the smaller its information entropy, indicating that the indicator provides a greater amount of information and should be given a higher weight. Through this method, the system can automatically calculate the weight of all indicators, including the "information differentiation risk index".

[0070] Then, the calculated weights are used to weight the initial decision matrix to obtain a weighted decision matrix.

[0071] Finally, the optimal solution distance method is applied to calculate the final financial cooperation risk index. This method first determines an "optimal solution" (i.e., an ideal solution where each indicator takes the optimal value) and a "worst solution" (i.e., a virtual solution where each indicator takes the worst value) from the weighted matrix. Then, it calculates the Euclidean distance between the evaluation vector of each real country and the optimal and worst solutions. Finally, it calculates the relative proximity of each country, which is the financial cooperation risk index output by this application. The calculation method is as follows: divide the distance between the country's evaluation vector and the worst solution by the sum of the distances between the country's evaluation vector and the optimal and worst solutions. The value of this relative proximity is... The range is from 0 to 1. In the settings of this application, the closer a country's value is to 1, the closer it is to the optimal solution and the farther it is from the worst solution, thus the lower its financial cooperation risk and the higher its cooperation stability; conversely, the closer the value is to 0, the higher its risk; for example, if the calculated risk index of a certain country in 2023 is 0.35, while that of another country is 0.68, it indicates that the overall risk of financial cooperation with the latter is lower than that with the former; the server stores the calculated index and each sub-indicator in the results database, and can push it to the front-end visualization dashboard for display through the application programming interface.

[0072] Understandably, macroeconomic indicators include at least one of the following categories: governance stability indicators, including the degree of governance stability and the quality of regulation; social stability indicators, including the rule of law, the global peace index, and the unemployment rate; domestic economic stability indicators, including debt repayment capacity, GDP per capita, and GDP growth rate; and trade and investment stability indicators, including net inflows of foreign direct investment and the proportion of bilateral trade to GDP.

[0073] Among them, the primary indicator of governance stability includes the following sub-indicators: the "Governance Stability" indicator, which is derived from the World Bank Global Governance Index and ranges from -2.5 to +2.5. It is a negative indicator, meaning that the higher the value, the lower the risk; and the "Supervisory Quality" indicator, which is also derived from the World Bank Global Governance Index and has the same value range and risk direction as the former.

[0074] Regarding the primary indicator of social stability, its sub-indicators include: the "Rule of Law" indicator from the World Bank's Global Governance Index, which ranges from -2.5 to +2.5 and is a negative indicator; and the "Global Peace Index" published by the Institute for Economics and Peace, where a lower score represents greater peace and is a positive indicator, meaning a higher score indicates higher risk.

[0075] Regarding the primary indicator of economic stability, its sub-indicators include: the "debt service capacity" indicator from the World Bank, which is usually measured as a percentage of total external debt to gross national income and is a positive indicator; the "GDP per capita" indicator from the World Bank, denominated in US dollars and is a negative indicator; and the "GDP growth rate" indicator from the World Bank, expressed as a percentage and is a negative indicator.

[0076] The primary indicator for trade and investment stability includes the following sub-indicators: the "net foreign direct investment inflow as a percentage of GDP" from the World Bank, expressed as a percentage and considered a negative indicator; and the "bilateral trade volume as a percentage of total trade volume" from relevant departments or the International Monetary Fund, expressed as a percentage and considered a negative indicator.

[0077] Understandably, the steps for calculating and outputting the financial cooperation risk index using a comprehensive evaluation model specifically involve employing the entropy weight-optimal solution distance method, including: calculating the weights of social network information risk indicators and macroeconomic indicator data using the entropy weight method, and constructing a weighted decision matrix based on these weights; determining the optimal and worst-case scenarios based on the weighted decision matrix; calculating the Euclidean distance between the evaluation vectors of each target cooperating country and the optimal and worst-case scenarios; calculating the relative proximity based on the Euclidean distance, and using this relative proximity as the final financial cooperation risk index.

[0078] Suppose we are evaluating three countries, country A, country B, and country C, and we have three indicators, such as GDP per capita, information differentiation risk, and policy expectation deviation risk.

[0079] After standardization, all indicators are processed to obtain a decision matrix in which all values ​​are in the range of 0 to 1, and the larger the value, the higher the risk.

[0080] First, calculate the proportion of the value of the i-th country in the total value of all countries for the j-th indicator. Then, use this proportion to calculate the information entropy of the j-th indicator by multiplying the proportion of each country by its natural logarithm, summing the results for all countries, and then multiplying by a negative value equal to the reciprocal of the natural logarithm of the total number of countries. Finally, calculate the weight of the j-th indicator by subtracting its information entropy from 1 and then dividing by the sum of the subtractions of 1 and their corresponding information entropies for all indicators. Through this calculation, the system can determine the weights of each indicator, indicating that some indicators contain more information and therefore have higher weights in this evaluation.

[0081] Multiply each column of the decision matrix by its corresponding weight to construct a weighted matrix.

[0082] Each element of the optimal solution is the minimum value of the corresponding column in the weighted matrix, because smaller values ​​indicate lower risk. Conversely, each element of the worst solution is the maximum value of the corresponding column in the weighted matrix. For each country, calculate the Euclidean distance from its evaluation vector to the optimal and worst solutions.

[0083] The relative proximity, or risk index, is obtained by dividing the distance between each country's evaluation vector and the worst-case scenario by the sum of the distances between that evaluation vector and the best-case scenario. The final output is the risk index for each country, for example: {Country A: 0.35, Country B: 0.68, Country C: 0.21}, which indicates that Country B has the lowest risk and Country C has the highest risk.

[0084] like Figure 2 As shown, this application also discloses a financial risk assessment system that integrates network information and macroeconomic data, comprising: a data acquisition module for acquiring heterogeneous data from the target partner country, the heterogeneous data including structured macroeconomic indicator data and unstructured social network information text data; a network information processing module for quantifying the social network information text data to generate at least one social network information risk indicator, wherein the social network information risk indicator includes an information differentiation risk index obtained by calculating the sentiment differences of different information sources, and a policy expectation deviation index obtained by calculating the expectation difference between public sentiment and policy implementation; a macroeconomic data processing module for standardizing the structured macroeconomic indicator data; and a risk assessment module configured to integrate the social network information risk indicator generated by the social network information processing module and the macroeconomic indicator data processed by the macroeconomic data processing module, and calculate and output a financial cooperation risk index for the target partner country using a comprehensive evaluation model.

[0085] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the above-described division of functional modules is used as an example. In practical applications, the above functions can be assigned to different functional modules as needed, that is, the internal structure of the device can be divided into different functional modules to complete all or part of the functions described above. The specific working process of the system, device, and unit described above can be referred to the corresponding process in the foregoing method embodiments, and will not be repeated here.

[0086] In the several embodiments provided in this application, it should be understood that the disclosed systems, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of modules or units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces, or indirect coupling or communication connection between apparatuses or units, and may be electrical, mechanical, or other forms.

[0087] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0088] Furthermore, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0089] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) or processor to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory, random access memory, magnetic disks, or optical disks.

[0090] The above are all preferred embodiments of this application and are not intended to limit the scope of protection of this application. Any feature disclosed in this specification (including the abstract and drawings) may be replaced by other equivalent or similar features unless specifically stated otherwise. That is, unless specifically stated otherwise, each feature is only one example of a series of equivalent or similar features.

Claims

1. A financial risk assessment method of fusing network information and macro data, characterized in that, include: Obtain heterogeneous data from the target partner countries, including structured macroeconomic indicator data and unstructured social network information text data, wherein the social network information text data comes from at least two different types of media sources; The social network information text data is quantified to generate at least one social network information risk indicator; The macroeconomic indicator data are standardized. A weighted decision matrix is ​​constructed by integrating the social network information risk indicators with the standardized macroeconomic indicator data. Based on the weighted decision matrix, a comprehensive evaluation model is used to calculate and output the financial cooperation risk index for the target partner country.

2. The method of claim 1, wherein, The step of quantifying the social network information text data specifically includes: The media sources are divided into official or mainstream media as the first type of media sources, and social media or online forums as the second type of media sources. The first sentiment index is calculated based on the online information text data of the first type of media source; A second sentiment index is calculated based on the online information text data of the second type of media sources; Based on the difference between the first sentiment index and the second sentiment index, an information differentiation risk index is calculated and generated as one of the social network information risk indicators.

3. The method of claim 2, wherein, The calculation steps for the first and second sentiment indices include: Establish a media authority weighting system, classify the media sources into different authority levels and assign corresponding weight coefficients; Sentiment scores were assigned to the online information text data from various media sources to obtain basic sentiment scores. The basic sentiment score and the corresponding weight coefficient are weighted and aggregated to calculate the first sentiment index and the second sentiment index respectively.

4. The method of claim 3, wherein, The steps for performing sentiment scoring on the online information text data from various media sources include: Based on a pre-defined sentiment dictionary, degree adverb dictionary, and negation word dictionary, sentences in online information text data are processed; Identify positive sentiment words, negative sentiment words, degree modifiers, and negation words in sentences; Based on the preset scoring rules and the identified words, the basic sentiment score of the sentence is calculated.

5. The method of claim 2, wherein, Before calculating the first and second sentiment indices, the following steps are also included: The web crawler technology is used to collect the text data of the network information from the Internet; a keyword library containing national identifiers and bilateral financial cooperation business terms, as well as an exclusion keyword library, are constructed; The collected online information text data is filtered using the keyword database to extract text related to the theme of financial cooperation.

6. The method of claim 1, wherein, The macroeconomic indicator data includes at least one of the following categories: Governance stability indicators include the degree of governance stability and the quality of supervision; Social stability indicators include the rule of law and the global peace index; Indicators of its own economic stability, including debt repayment capacity, GDP per capita, and GDP growth rate; Trade and investment stability indicators include net inflows of foreign direct investment and the ratio of bilateral trade to GDP.

7. The method of claim 1, wherein, The step of quantifying the social network information text data further includes: Obtain policy implementation indicator data reflecting the effectiveness of policy implementation in the target cooperative countries; The policy implementation indicator data and the sentiment index of social network information from non-mainstream media are standardized. The absolute difference between the standardized policy implementation indicators and the social network information sentiment index is calculated to generate a policy expectation deviation index, which serves as one of the social network information risk indicators.

8. The method of claim 7, wherein, The policy implementation indicator data includes at least one of the following: The number of bilateral investment projects with the target countries; Turnover on contracted projects in the target cooperative countries; The proportion of bilateral aid to the recipient country's GDP.

9. The method of claim 1, wherein, The step of calculating and outputting the financial cooperation risk index for the target cooperative country using a comprehensive evaluation model specifically involves employing the entropy weight-superiority distance method, including: The weights of the social network information risk index and the macroeconomic index data are calculated using the entropy weight method, and a weighted decision matrix is ​​constructed based on these weights. The optimal and worst solutions are determined based on the weighted decision matrix. Calculate the Euclidean distance between the evaluation vector of each target cooperating country and the optimal and worst solutions; The relative proximity is calculated based on the Euclidean distance, and this relative proximity is used as the final financial cooperation risk index.

10. A financial risk assessment system that fuses network information with macro data, characterized by, include: The data acquisition module is used to acquire heterogeneous data from the target partner countries, including structured macroeconomic indicator data and unstructured social network information text data. The network information processing module is used to quantify the social network information text data and generate at least one social network information risk indicator, wherein the social network information risk indicator includes an information differentiation risk index obtained by calculating the emotional differences of different information sources, and a policy expectation deviation index obtained by calculating the expectation difference between public sentiment and policy implementation. The macro data processing module is used to standardize the macro indicator data; The risk assessment module is configured to integrate social network information risk indicators generated by the social network information processing module and macroeconomic indicator data processed by the macroeconomic data processing module, and use a comprehensive evaluation model to calculate and output a financial cooperation risk index for the target cooperating country.

Images