A risk management evaluation method and system for the digital economy based on cloud computing

Abstract

An embodiment of the present invention provides a digital economy risk management assessment method and system based on cloud computing, belonging to the technical field of risk assessment. The method includes: collecting initial behavior data corresponding to a target object and initial news data related to the behavior of the target object according to a cloud computing platform corresponding to the target object; performing association analysis on the initial behavior data to obtain a target association relationship, and filling the initial behavior data according to the target association relationship to obtain target behavior data; performing anomaly identification on the target behavior data to obtain target anomaly data corresponding to the target object, and determining target behavior characteristics corresponding to the target object according to the target anomaly data; performing news view identification on the initial news data to obtain a target behavior view related to the behavior of the target object; calculating a distance according to the target behavior characteristics and the target behavior view to obtain gap information, and performing risk assessment on the target object according to the gap information to obtain a risk assessment result.

Description

Technical Field

[0001] The present invention relates to the technical field of risk assessment, and particularly to a risk management and assessment method and system for the digital economy based on cloud computing. Background Art

[0002] The current risk assessment system mainly relies on transaction data for analysis and judgment. Although this method can reflect market fluctuations and trading behaviors to a certain extent, its limitations are becoming increasingly prominent. Merely relying on transaction data itself often makes it difficult to capture deeper risk factors, such as the macroeconomic environment and industry dynamics. These factors are equally important for risk, but existing risk assessment technologies cannot fully consider this multi-dimensional information. Therefore, there is a problem of low accuracy in existing technologies when conducting risk assessment, making it difficult to comprehensively and timely identify potential risks, thus failing to effectively serve as a warning and affecting the scientificity and accuracy of decision-making. Summary of the Invention

[0003] The main purpose of the embodiments of the present invention is to provide a risk management and assessment method and system for the digital economy based on cloud computing, aiming to solve the problem of low accuracy in risk assessment in related technologies.

[0004] In the first aspect, the embodiments of the present invention provide a risk management and assessment method for the digital economy based on cloud computing, including:

[0005] Collect initial behavior data corresponding to the target object and initial news data related to the behavior of the target object according to the cloud computing platform corresponding to the target object;

[0006] Perform correlation analysis on the initial behavior data to obtain a target correlation relationship, and perform data filling on the initial behavior data according to the target correlation relationship to obtain target behavior data;

[0007] Perform anomaly identification on the target behavior data to obtain target anomaly data corresponding to the target object, and determine target behavior characteristics corresponding to the target object according to the target anomaly data;

[0008] Perform news view identification on the initial news data to obtain target behavior views related to the behavior of the target object;

[0009] Calculate the distance between the target behavior characteristics and the target behavior views according to the target behavior characteristics and the target behavior views to obtain gap information between the target behavior characteristics and the target behavior views, and perform risk assessment on the target object according to the gap information to obtain a risk assessment result corresponding to the target object.

[0010] In the second aspect, the embodiments of the present invention provide a risk management and assessment system for the digital economy based on cloud computing, including:

[0011] A data collection module, configured to collect initial behavior data corresponding to the target object and initial news data related to the behavior of the target object according to the cloud computing platform corresponding to the target object;

[0012] A data processing module, configured to perform correlation analysis on the initial behavior data to obtain a target correlation relationship, and perform data filling on the initial behavior data according to the target correlation relationship to obtain target behavior data;

[0013] An anomaly recognition module, configured to perform anomaly recognition on the target behavior data to obtain target anomaly data corresponding to the target object, and determine target behavior characteristics corresponding to the target object according to the target anomaly data;

[0014] An opinion recognition module, configured to perform news opinion recognition on the initial news data to obtain target behavior opinions related to the behavior of the target object;

[0015] A risk assessment module, configured to calculate a distance between the target behavior characteristics and the target behavior opinions to obtain gap information therebetween, and perform risk assessment on the target object according to the gap information to obtain a risk assessment result corresponding to the target object.

[0016] An embodiment of the present invention provides a method and system for risk management assessment of the digital economy based on cloud computing. The method includes: collecting initial behavior data of a target object and initial news data related to the initial behavior data through a cloud computing platform, so as to comprehensively capture the behavior characteristics of the target object and whether the behavior characteristics are consistent with external information. Then, by performing correlation analysis on the initial behavior data, the correlation relationship between different behaviors is identified, and the target behavior data is further improved through data filling to enhance the integrity and accuracy of the data. Furthermore, through anomaly recognition technology, target anomaly data in the behavior of the target object can be discovered, and based on the identified target anomaly data, the target behavior characteristics of the target object can be determined, which can help to more accurately understand the behavior pattern of the target object. By performing news opinion recognition on the initial news data, target behavior opinions related to the behavior of the target object can be obtained. Then, based on the target behavior characteristics and the target behavior opinions, distance calculation is performed to obtain the gap information between the target behavior characteristics and the target behavior opinions, and based on the gap information, risk assessment of the target object is carried out to obtain the risk assessment result corresponding to the target object. Furthermore, by combining the target behavior characteristics and the target behavior opinions for risk assessment, the risk level of the target object can be judged more comprehensively. The behavior characteristics provide an analysis of the internal behavior pattern, while the target behavior opinions reflect the reference information provided by the outside world for the behavior involved by the target object. Through comprehensive analysis, the generated risk assessment result can provide a strong reference basis for decision-makers to help them better identify and manage potential risks. It also solves the problem of low accuracy in risk assessment in related technologies. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings required for the description of the embodiments will be briefly introduced below. Obviously, the drawings in the following description are some embodiments of the present application. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0018] Figure 1 It is a schematic flowchart of a method for risk management assessment of the digital economy based on cloud computing provided by an embodiment of the present invention;

[0019] Figure 2 It is a schematic diagram of the module structure of a system for risk management assessment of the digital economy based on cloud computing provided by an embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0020] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of them. All other embodiments obtained by ordinary technicians in this field based on the embodiments of the present invention without making any creative efforts shall fall within the scope of protection of the present invention.

[0021] The flowcharts shown in the accompanying drawings are only examples and do not necessarily include all the contents and operations / steps, nor must they be executed in the order described. For example, some operations / steps may also be decomposed, combined or partially merged, so the actual execution order may change according to actual conditions.

[0022] It should be understood that the terms used in this specification are only for the purpose of describing particular embodiments and are not intended to limit the present invention. As used in the specification and appended claims, the singular forms "a," "an," and "the" are intended to include the plural forms unless the context clearly indicates otherwise.

[0023] Embodiments of the present invention provide a cloud computing-based digital economy risk management and assessment method and system. The cloud computing-based digital economy risk management and assessment method can be applied to terminal devices, such as tablet computers, laptop computers, desktop computers, personal digital assistants, and wearable devices. The terminal device can also be a server or a server cluster.

[0024] The following embodiments of the present invention are described in detail with reference to the accompanying drawings. In the absence of conflict, the following embodiments and features therein may be combined with each other.

[0025] Please refer to Figure 1 , Figure 1 A flowchart of a cloud computing-based digital economy risk management assessment method provided in an embodiment of the present invention.

[0026] like Figure 1 As shown, the digital economy risk management assessment method based on cloud computing includes steps S101 to S105.

[0027] Step S101: collecting initial behavior data corresponding to the target object and initial news data related to the target object's behavior based on the cloud computing platform corresponding to the target object.

[0028] Exemplarily, the cloud computing platform corresponding to the target object is determined, and initial behavior data corresponding to the target object is collected based on the cloud computing platform. The initial behavior data includes but is not limited to log data, transaction data, user behavior data, etc.

[0029] For example, based on the target's initial behavioral data, key themes or specific keywords are identified when collecting news information. These themes or keywords will be closely related to the initial behavioral data. The cloud computing platform's efficient processing capabilities will then filter and collect initial news data matching these keywords or themes from a wide range of sources (such as news websites, social media, and forums). This initial news data is relevant news information collected specifically based on the target's initial behavioral data. For example, if the target's initial behavioral data indicates a purchase of Company A's stock, the cloud computing platform will automatically collect news reports, market trends, industry analysis, and other information related to Company A. This collected initial news data will provide important evidence for subsequent risk assessments, ensuring that risk assessments based on the initial behavioral data are more accurate and comprehensive.

[0030] Step S102: performing association analysis on the initial behavior data to obtain a target association relationship, and performing data filling on the initial behavior data according to the target association relationship to obtain target behavior data.

[0031] Exemplarily, association rule algorithms such as Apriori and FP-Growth are used to perform association analysis on the initial behavior data to obtain related behaviors with association relationships in the initial behavior data.

[0032] For example, missing or incomplete information is identified in the initial behavioral data, and based on the existing target association relationships, the possible values of the missing data are inferred. The inferred values are then added to the initial behavioral data to generate complete target behavioral data. This allows for effective association analysis of the initial behavioral data, and the resulting target association relationships are used to fill in the missing data, resulting in more complete and accurate target behavioral data, providing a solid data foundation for subsequent applications such as risk assessment.

[0033] In some embodiments, obtaining the target behavior data by filling the initial behavior data according to the target association relationship includes: performing a missing detection on the initial behavior data to obtain target missing data corresponding to a first parameter in the initial behavior data; obtaining target adjacent data corresponding to the target missing data under the first parameter from the initial behavior data; obtaining a second parameter corresponding to the first parameter according to the target association relationship, and obtaining initial association data corresponding to the target adjacent data under the second parameter from the initial behavior data; establishing a target conversion relationship corresponding to the first parameter and the second parameter according to the target adjacent data and the initial association data; obtaining a target missing position corresponding to the target missing data, and obtaining target association data corresponding to the second parameter at the target missing position from the initial behavior data; predicting the target missing data according to the target conversion relationship in combination with the target association data to obtain target filling data; and filling the target missing data with the target filling data to obtain the target behavior data.

[0034] Exemplarily, a comprehensive missing detection is performed on the initial behavior data to find the first parameter with missing information in the data and identify the missing data under the first parameter, and these data are the target missing data to be filled. Under the first parameter, the context data or adjacent data corresponding to the target missing data is found from the initial behavior data to obtain the target adjacent data corresponding to the target missing data under the first parameter.

[0035] Exemplarily, according to the target association relationship, the second parameter associated with the first parameter is identified. From the initial behavior data, the corresponding data of the target adjacent data under the second parameter is extracted, and these data are called initial association data.

[0036] Exemplarily, statistical analysis or machine learning methods are used to analyze the conversion relationship between the target adjacent data and the initial association data, and then this conversion relationship is determined as the target conversion relationship between the first parameter and the second parameter, which will become the basis for predicting the target missing data.

[0037] Exemplarily, in the initial behavior data, the specific position of the target missing data is determined, so that at the target missing position, the target association data corresponding to the second parameter is extracted, and then the previously established target conversion relationship and the target association data are used to predict the possible value of the target missing data to obtain the target filling data.

[0038] Exemplarily, the predicted target filling data is inserted into the corresponding missing position of the initial behavior data to complete the data filling process, and then after the data filling, the complete target behavior data is finally obtained.

[0039] Exemplarily, detect the missing parts in the initial behavior data, predict and fill in the missing data using the target correlation relationship, and finally obtain a complete and accurate target behavior data, thereby providing a solid foundation for subsequent data analysis and risk assessment support.

[0040] Step S103: Perform anomaly identification on the target behavior data to obtain the target anomaly data corresponding to the target object, and determine the target behavior characteristics corresponding to the target object according to the target anomaly data.

[0041] Exemplarily, according to anomaly detection methods, such as statistical methods (e.g., Z-score, Grubbs' test), machine learning methods (e.g., Isolation Forest, One-Class SVM), or deep learning methods (e.g., Autoencoder), perform anomaly identification on the target behavior data to identify the anomaly points that are significantly different from most of the data, and these anomaly points are the target anomaly data.

[0042] Exemplarily, analyze the distribution characteristics of the target anomaly data, and then extract the key features that can represent the behavior characteristics of the target object from the target anomaly data. The key features include but are not limited to frequency, duration, intensity, change trend, etc., so as to perform feature fusion on the key features to obtain the target behavior characteristics corresponding to the target object.

[0043] In some embodiments, the performing anomaly identification on the target behavior data to obtain the target anomaly data corresponding to the target object includes: obtaining any one sub-behavior data from the target behavior data, and splitting the target behavior data into first behavior data and second behavior data according to the sub-behavior data; calculating the data difference between the first behavior data and the second behavior data to obtain the corresponding behavior difference value; obtaining the maximum difference value from all the behavior difference values, and calculating the target probability corresponding to each behavior difference value less than or equal to the maximum difference value using statistical significance; determining a preset threshold, and splitting the target behavior data according to the preset threshold and the target probability to obtain target segmentation data; performing anomaly identification on each sub-segmentation data in the target segmentation data to obtain the target anomaly data corresponding to the target object.

[0044] Exemplarily, arbitrarily select one sub-behavior data from the target behavior data as the segmentation point. This sub-behavior data can be a specific time point, event, or behavior pattern, and then according to the selected sub-behavior data, split the target behavior data into two parts, namely the first behavior data and the second behavior data.

[0045] Exemplarily, methods such as mean comparison, standard deviation analysis, and feature distance calculation are used to compare the first row of data and the second row of data, calculate the differences in their numerical values, features, or behavior patterns, generate behavior difference values, and then select the largest difference value from all the calculated behavior difference values as the maximum difference value, which represents the maximum inconsistency between the two parts after data segmentation.

[0046] Exemplarily, using statistical methods, calculate the probability when each behavior difference value is less than or equal to the maximum difference value, that is, the target probability. This reflects the significance level of each behavior difference value in the overall data. Then, according to business requirements or statistical experience, determine a preset threshold for judging whether the behavior difference value reaches the significance level, and then compare the target probability with the preset threshold: if the target probability is less than or equal to the preset threshold, it is considered that the behavior difference value is significant and the segmentation point is retained. If the target probability is greater than the preset threshold, the segmentation point is excluded. After screening, a series of significant segmentation points are obtained, and these segmentation points divide the target behavior data into several sub-segmented data, which are then determined as the target segmented data.

[0047] Exemplarily, apply anomaly recognition methods (such as statistical methods, machine learning methods) to each sub-segmented data in the target segmented data to detect anomaly points in each sub-segmented data, and finally obtain the target anomaly data corresponding to the target object.

[0048] In some embodiments, the obtaining the target anomaly data corresponding to the target object by performing anomaly recognition on each sub-segmented data in the target segmented data includes: obtaining an anomaly score for each sub-data in the sub-segmented data to obtain an anomaly score value corresponding to the sub-data; performing data clustering on all the anomaly score values according to a preset quantity to obtain first clustering data and second clustering data corresponding to the anomaly score values; obtaining a first quantity corresponding to the first clustering data and a second quantity corresponding to the second clustering data; calculating the difference between the first quantity and the second quantity to obtain a target difference, and obtaining sub-anomaly data corresponding to the sub-segmented data from the first clustering data and the second clustering data according to the target difference; and fusing all the sub-anomaly data to obtain the target anomaly data corresponding to the target object.

[0049] Exemplarily, use anomaly detection algorithms such as Isolation Forest, Local Outlier Factor (LOF), or statistical-based methods (such as Z-Score or IQR) to obtain an anomaly score value corresponding to each sub-data in the sub-segmented data.

[0050] Exemplarily, according to a preset quantity, data clustering is performed on all abnormal scores using K-Means or DBSCAN to obtain first clustering data and second clustering data corresponding to the abnormal scores. For example, if the preset quantity is 2, then all abnormal scores are clustered using K-Means to obtain two clustering results, which are respectively denoted as the first clustering data and the second clustering data.

[0051] Exemplarily, calculate the number of data in the first clustering data, denoted as the first quantity, calculate the number of data in the second clustering data, denoted as the second quantity, and then calculate the difference between the first quantity and the second quantity to obtain the target difference. Then, when the target difference is less than or equal to the preset quantity, it indicates that the data gap between the first clustering data and the second clustering data is small, and there is no abnormal data in the sub-divided data. When the target difference is greater than the preset quantity, it indicates that the clustering with the smaller quantity in the first clustering data and the second clustering data is used as the abnormal data clustering, and then data is extracted from the clustering with the smaller quantity. These data are the sub-abnormal data corresponding to the sub-divided data. Thus, all sub-abnormal data are merged to obtain the target abnormal data of the target object.

[0052] In some embodiments, the obtaining of the abnormal score corresponding to each sub-data in the sub-divided data includes: analyzing the sub-divided data based on the isolation forest algorithm to obtain the target isolation tree corresponding to the sub-divided data; obtaining the current path length corresponding to the sub-data according to the target isolation tree, and calculating the average path length corresponding to the sub-data according to the current path length; determining a reference factor, and performing abnormal scoring according to the average path length and the reference factor to determine the abnormal score corresponding to each sub-data in the sub-divided data.

[0053] Exemplarily, an abnormal detection model is constructed using the isolation forest algorithm, and the corresponding isolation tree structure is obtained. Thus, the isolation forest model is trained using the sub-divided data to generate multiple target isolation trees.

[0054] Exemplarily, calculate the current path length of each sub-data in the target isolation tree, and take the average of all current path lengths to characterize its abnormal degree. That is, for each sub-data, traverse each target isolation tree and record the path length from the root node to the leaf node. For each sub-data, calculate the average of its path lengths in all isolation trees, that is, the average path length.

[0055] Exemplarily, the reference factor is used to standardize the path length corresponding to the sub-data. Different reference factors are set for different average path lengths. For example, if the average path length is less than 2, the reference factor is set to 0; if the average path length is equal to 2, the reference factor is set to 1; if the average path length is greater than 2, the reference factor is obtained according to the following formula:

[0056]

[0057] Where w represents the average path length When the reference factor is greater than 2, represents the average path length, c represents a fixed constant, and ln represents the natural logarithm.

[0058] Exemplarily, after determining the reference factor based on the average path length, the ratio of the average path length and the reference factor is calculated and the negative number is taken to obtain the corresponding numerical value, and the data corresponding to the exponential of the numerical value with base 2 is taken to determine the abnormal score corresponding to the sub-data.

[0059] In some embodiments, determining the target behavior characteristics corresponding to the target object based on the target abnormal data includes: obtaining target neighboring data corresponding to the target abnormal data from the target behavior data; determining the target change trend corresponding to the object under the target abnormal data based on the target neighboring data and the target abnormal data; extracting time domain features and frequency domain features of the target abnormal data to obtain target-related features corresponding to the target abnormal data; and performing feature fusion on the target-related features and the target change trend to obtain the target behavior characteristics corresponding to the target object.

[0060] For example, for each identified target abnormal data, its neighboring data points are obtained. These neighboring data points can be the previous and next data points in time, or the adjacent data points in space, so as to obtain the target neighboring data.

[0061] For example, the target neighboring data and the target abnormal data are compared by calculating the rate of change, trend direction, etc., and the behavior change trend of the target object under the abnormal data is analyzed. For example, the overall change pattern of the target object behavior before and after the target abnormal data is determined, such as increase, decrease or fluctuation, so as to obtain the target change trend.

[0062] For example, time-domain and frequency-domain features are extracted from the target anomaly data. Time-domain features can include mean, variance, and peak values, while frequency-domain features can be extracted using methods such as Fourier transforms to extract frequency components and periodicity information. These features are then acquired to provide a more comprehensive understanding of the characteristics of the target anomaly data. The extracted time-domain and frequency-domain features are then fused with the previously determined target change trends. This can be achieved through feature weighting and combination to form a comprehensive behavioral feature description, ensuring that the fused features accurately reflect the target behavioral characteristics of the target object under the target anomaly data.

[0063] Exemplarily, based on the fused features and the change trend, determine the target behavior characteristics of the target object under the target abnormal data. The target behavior characteristics should be able to describe the behavior pattern and potential change law of the target object under abnormal conditions.

[0064] Step S104: Perform news opinion recognition on the initial news data to obtain the target behavior opinions related to the behavior of the target object.

[0065] Exemplarily, the initial news data is news information closely related to the initial behavior data. Then, identify the opinions in the initial news data that involve the initial behavior data, such as the opinions related to transactions with the target object or the investment association with the target object, to obtain the target behavior opinions related to the behavior of the target object. That is, the target behavior opinions are the information about the associated objects related to the relevant behavior of the target object.

[0066] For example, use algorithms such as topic modeling, sentiment analysis, and opinion mining to extract the overall trend and mainstream views of the associated objects related to the relevant behavior of the target object from the initial news data, so as to obtain the target behavior opinions related to the behavior of the target object.

[0067] For example, if there is a large - amount transaction of the target object A with the associated object B, then the initial news data will perform news opinion recognition on the associated object B to evaluate whether there is a risk in the large - amount transaction of the target object A with the associated object B through the target behavior opinions in the follow - up.

[0068] In some embodiments, obtaining the target behavior view related to the target object's behavior by performing news view recognition on the initial news data includes: performing event element recognition on the initial news data to obtain corresponding target event information, and determining the first association information between the initial news data according to the element co-occurrence information between the target event information; screening out relevant news data associated with the initial news data according to the first association information, and obtaining the second association information between the initial news data and the relevant news data from the first association information; performing sentiment recognition on the initial news data to obtain corresponding first sentiment information and performing sentiment recognition on the relevant news data to obtain corresponding second sentiment information; and determining the third association information between the initial news data and the relevant news data according to the distance between the first sentiment information and the second sentiment information; fusing the second association information and the third association information to determine the target representation information corresponding to the initial news data under the relevant news data; performing key information recognition on the initial news data according to the target representation information to obtain the target key sentences corresponding to the initial news data; performing behavior classification according to the target key sentences to obtain the target behavior view related to the target object's behavior; wherein, the target representation information is obtained according to the following formula:

[0069]

[0070] Wherein, represents the target representation information corresponding to the i-th initial news data, represents the damping coefficient, num represents the number of news corresponding to the relevant news data, represents the weight information corresponding to the second association information, represents the weight information corresponding to the third association information, represents the second association information between the i-th initial news data and the k-th relevant news data; represents the third association information between the i-th initial news data and the k-th relevant news data, represents the second association information between the k-th relevant news data and the j-th relevant news data; represents the third association information between the k-th relevant news data and the j-th relevant news data; represents the target representation information corresponding to the k-th relevant news data under the news data associated with the k-th relevant news data.

[0071] Exemplarily, event element recognition is performed on the initial news data to extract key information describing the event. These elements may include time, location, person, event type, event result, etc., and then the extracted event elements are organized into structured target event information.

[0072] Exemplarily, according to the co-occurrence relationship of elements in the target event information (i.e., elements that co-occur in multiple event information), the first association information between the initial news data is determined, and the first association information can reflect the similarity or relevance of events in different news reports.

[0073] Exemplarily, according to the first association information, relevant news data associated with the initial news data is screened out, and then the second association information between the initial news data and the relevant news data is further extracted from the first association information.

[0074] Exemplarily, sentiment recognition is performed on the initial news data to extract the corresponding first sentiment information (such as positive sentiment, negative sentiment, neutral sentiment, etc.), and sentiment recognition is performed on the relevant news data to extract the corresponding second sentiment information.

[0075] Exemplarily, according to the difference between the first sentiment information and the second sentiment information (such as the similarity of sentiment polarity, the change of sentiment intensity, etc.), the sentiment distance is calculated, and the third association information between the initial news data and the relevant news data is determined. The third association information can reflect the similarity or difference in sentiment expression in different news reports.

[0076] Exemplarily, the second association information (based on event element association) and the third association information (based on sentiment association) are fused according to the following formula to determine the target representation information of the initial news data under the relevant news data. The target representation information may include the core description of the event, the integration of sentiment tendency, etc.:

[0077]

[0078] where represents the target representation information corresponding to the i-th initial news data, represents the damping coefficient, num represents the number of news corresponding to the relevant news data, represents the weight information corresponding to the second association information, represents the weight information corresponding to the third association information, represents the second association information between the i-th initial news data and the k-th relevant news data; represents the third association information between the i-th initial news data and the k-th relevant news data, represents the second association information between the k-th relevant news data and the j-th relevant news data; Represents the third association information between the k-th relevant news data and the j-th relevant news data; Represents the target characterization information corresponding to the k-th relevant news data under the news data associated with the k-th relevant news data.

[0079] Exemplarily, by fusing the second association information based on event elements and the third association information based on sentiment, the characteristics of news data can be comprehensively described from multiple dimensions. This not only considers the structural information of the event but also the expression of sentiment, which helps to more comprehensively understand the news content. The target characterization information can more accurately reflect the core content and sentiment tendency of news data by fusing multiple association information. By fusing the second association information and the third association information, richer data support can be provided to help decision-makers more comprehensively understand the background, impact, and sentiment tendency of news events.

[0080] Exemplarily, use the event elements and sentiment tendency extracted from the target characterization information as screening criteria. If the target characterization information shows that the core of an event in a news report is "meeting decision" and the sentiment tendency is "negative", then filter out the statements related to "meeting decision" and with "negative" sentiment. Thus, each sentence in the initial news data is scored, and the scoring criteria include: whether the sentence contains the key elements in the target characterization information. Whether the sentence contains the sentiment tendency in the target characterization information. Thus, each sentence in the initial news data is analyzed to determine whether it conforms to the characteristics of the key sentence, and then according to the screening strategy, the sentences that meet the criteria are filtered out, and then the filtered key sentences are sorted, and the sentences with a high degree of association and high information density with the target characterization information are preferentially selected, and then the target key sentences are obtained.

[0081] Exemplarily, for the extracted target key sentences, a special classification model is used to conduct in-depth behavior classification to accurately identify the target behavior views related to the behavior of the target object. The target behavior views are that the associated object involved by the target object develops positively, such as achieving achievements, obtaining support, or realizing growth; or the associated object involved by the target object develops negatively, such as encountering setbacks, being hindered, or facing decline; or the associated object involved by the target object develops in a stable state, that is, the behavior and situation of the target object maintain the status quo without significant changes, any one of them.

[0082] Step S105, calculate the distance between the target behavior feature and the target behavior view according to the target behavior feature and the target behavior view to obtain the gap information between the target behavior feature and the target behavior view, and perform a risk assessment on the target object according to the gap information to obtain the risk assessment result corresponding to the target object.

[0083] Exemplarily, the target behavior type corresponding to the target object is obtained by classifying the types of the target behavior characteristics. For example, the target behavior type is increasing investment, decreasing investment, or stabilizing investment. Then, the target behavior type is compared with the target behavior view, such as positive development (such as support, growth), negative development (such as encountering obstacles, decline), or stable development (such as maintaining the status quo). When the target behavior type is increasing investment and the target behavior view is positive development (such as support, growth), the distance between the target behavior type and the target behavior view is calculated using the text similarity calculation formula to obtain the gap information between the target behavior characteristics and the target behavior view. When the gap information is small, it indicates that the risk assessment result corresponding to the target object is low risk; when the target behavior type is increasing investment and the target behavior view is negative development (such as encountering obstacles, decline), the distance between the target behavior type and the target behavior view is calculated using the text similarity calculation formula to obtain the gap information between the target behavior characteristics and the target behavior view. When the gap information is large, it indicates that the risk assessment result corresponding to the target object is high risk.

[0084] In some embodiments, calculating the distance between the target behavior characteristics and the target behavior view to obtain the gap information between the target behavior characteristics and the target behavior view, and performing a risk assessment on the target object according to the gap information to obtain the risk assessment result corresponding to the target object includes: representing the target behavior characteristics using the first feature representation layer of the risk assessment model to obtain first feature information; representing the target behavior view using the second feature representation layer of the risk assessment model to obtain second feature information; calculating the distance between the target behavior characteristics and the target behavior view using the distance calculation layer of the risk assessment model according to the first feature information and the second feature information to obtain the gap information between the target behavior characteristics and the target behavior view; classifying the data according to the gap information using the risk classification layer of the risk assessment model to obtain the target risk type between the target behavior characteristics and the target behavior view; and determining the risk assessment result corresponding to the target object according to the target risk type.

[0085] Exemplarily, the first feature representation layer of the risk assessment model is used to represent the target behavior characteristics. For example, key numerical features are extracted from the target behavior characteristics, such as the frequency, intensity, influence, etc. of the behavior. Then, the extracted features are normalized to ensure that the numerical ranges of different features are the same, which is convenient for subsequent calculations. Thus, the normalized features are represented in the form of vectors or matrices as the first feature information.

[0086] For example, the second feature representation layer is used to characterize the target behavior opinion. For example, specific features such as positive or negative tendency, sentiment intensity, and opinion proportion are extracted from the target behavior opinion. The extracted features are normalized to ensure that they have the same numerical range as the first feature information. The normalized opinion features are represented as vectors or matrices as the second feature information.

[0087] Exemplarily, the distance calculation layer is used to calculate the gap between the first feature information and the second feature information. For example, the distance between the first feature information and the second feature information is calculated according to a distance measurement method (such as Euclidean distance, cosine similarity, etc.), and the calculation result is converted into a specific gap value to represent the difference between the target behavior characteristics and the target behavior viewpoints to obtain gap information.

[0088] For example, the risk classification layer can be used to classify data based on gap information. For example, different risk types can be defined based on the range or threshold of the gap information. For example, low risk: the gap is small, and the behavioral characteristics are consistent with the behavioral viewpoint. Medium risk: the gap is moderate, and there is a certain deviation between the behavioral characteristics and the behavioral viewpoint. High risk: the gap is large, and the behavioral characteristics are significantly inconsistent with the behavioral viewpoint. Based on the value of the gap information, the data can be classified into the target risk type (e.g., low risk, medium risk, high risk).

[0089] Exemplarily, a risk assessment result of a target object is determined according to a target risk type, the target risk type is associated with the target object, a risk assessment result of the target object is generated, and then the target risk type is determined as the risk assessment result of the target object.

[0090] In some embodiments, after obtaining the risk assessment result, the method further includes: determining preset information, comparing the preset information with the risk assessment result, and when the risk assessment result meets the preset information, generating a risk description statement corresponding to the risk assessment result based on the target behavior characteristics and the target behavior viewpoint; determining a risk reminder result based on the risk description statement and the risk assessment result, and sending the risk reminder result to the target communication device.

[0091] For example, if the preset information indicates a medium or high risk level, then when the risk assessment result is any of the preset information, it is determined that the risk assessment result meets the preset conditions, thereby extracting key information related to the risk assessment result from the target behavior characteristics and target behavior perspectives, such as the specific manifestations of the target behavior characteristics (such as frequency and intensity) and the specific content of the target behavior perspectives (such as positive and negative tendencies and emotional intensity). Based on the extracted key information, a concise risk description statement is generated, such as: "The target object's behavior frequency has increased significantly, but the proportion of negative evaluations of related objects involved in the target object is too high, posing a high risk."

[0092] For example, a risk alert method is determined based on the risk assessment results, and a risk alert result corresponding to the target object is determined based on the risk alert method and risk description. The risk description is then sent to the target communication device corresponding to the target object based on the risk alert method in the risk alert result. This allows for timely identification of potential risks and the transmission of risk information to relevant personnel through effective communication methods, providing support for risk management and decision-making.

[0093] See also Figure 2 , Figure 2 A cloud computing-based digital economic risk management and assessment system 200 is provided in an embodiment of the present application. The cloud computing-based digital economic risk management and assessment system 200 includes a data acquisition module 201, a data processing module 202, an anomaly identification module 203, a viewpoint identification module 204, and a risk assessment module 205. The data acquisition module 201 is used to collect the initial behavior data corresponding to the target object and the initial news data related to the target object's behavior according to the cloud computing platform corresponding to the target object; the data processing module 202 is used to perform correlation analysis on the initial behavior data to obtain a target correlation relationship, and fill in the data of the initial behavior data according to the target correlation relationship. Obtain target behavior data; an anomaly identification module 203, used to perform anomaly identification on the target behavior data to obtain target anomaly data corresponding to the target object, and determine the target behavior characteristics corresponding to the target object based on the target anomaly data; an opinion identification module 204, used to perform news opinion identification on the initial news data to obtain a target behavior opinion related to the target object behavior; a risk assessment module 205, used to perform distance calculation based on the target behavior characteristics and the target behavior opinion to obtain gap information between the target behavior characteristics and the target behavior opinion, and perform risk assessment on the target object based on the gap information to obtain a risk assessment result corresponding to the target object.

[0094] In some embodiments, during the process of filling the initial behavior data according to the target association relationship to obtain the target behavior data, the data processing module 202 performs the following:

[0095] Perform missing detection on the initial behavior data to obtain target missing data corresponding to a first parameter in the initial behavior data;

[0096] Obtain target adjacent data corresponding to the target missing data under the first parameter from the initial behavior data;

[0097] Obtain a second parameter corresponding to the first parameter according to the target association relationship, and obtain initial associated data corresponding to the target adjacent data under the second parameter from the initial behavior data;

[0098] Establish a target conversion relationship corresponding to the first parameter and the second parameter according to the target adjacent data and the initial associated data;

[0099] Obtain a target missing position corresponding to the target missing data, and obtain target associated data corresponding to the second parameter at the target missing position from the initial behavior data;

[0100] Perform data prediction on the target missing data according to the target conversion relationship and the target associated data to obtain target filling data;

[0101] Fill the target missing data with the target filling data to obtain the target behavior data.

[0102] In some embodiments, during the process of performing anomaly recognition on the target behavior data to obtain target anomaly data corresponding to the target object, the anomaly recognition module 203 performs the following:

[0103] Obtain any sub-behavior data from the target behavior data, and divide the target behavior data into first behavior data and second behavior data according to the sub-behavior data;

[0104] Calculate the data difference between the first behavior data and the second behavior data to obtain a corresponding behavior difference value;

[0105] Obtain the maximum difference value from all the behavior difference values, and calculate the target probability corresponding to each behavior difference value being less than or equal to the maximum difference value using statistical significance;

[0106] Determine a preset threshold, and divide the target behavior data according to the preset threshold and the target probability to obtain target segmentation data;

[0107] Perform anomaly recognition on each sub-segmentation data in the target segmentation data to obtain the target anomaly data corresponding to the target object.

[0108] In some embodiments, during the process of performing anomaly recognition on each sub-segmentation data in the target segmentation data to obtain the target anomaly data corresponding to the target object, the anomaly recognition module 203 performs the following:

[0109] Perform anomaly scoring on each sub-data in the sub-segmentation data to obtain the anomaly score corresponding to the sub-data;

[0110] Perform data clustering on all the anomaly scores according to a preset quantity to obtain the first clustering data and the second clustering data corresponding to the anomaly scores;

[0111] Obtain the first quantity corresponding to the first clustering data and the second quantity corresponding to the second clustering data;

[0112] Calculate the difference between the first quantity and the second quantity to obtain the target difference, and obtain the sub-anomaly data corresponding to the sub-segmentation data from the first clustering data and the second clustering data according to the target difference;

[0113] Fuse all the sub-anomaly data to obtain the target anomaly data corresponding to the target object.

[0114] In some embodiments, during the process of performing anomaly scoring on each sub-data in the sub-segmentation data to obtain the anomaly score corresponding to the sub-data, the anomaly recognition module 203 performs the following:

[0115] Analyze the sub-segmentation data based on the isolation forest algorithm to obtain the target isolation tree corresponding to the sub-segmentation data;

[0116] Obtain the current path length corresponding to the sub-data according to the target isolation tree, and perform mean calculation according to the current path length to obtain the average path length corresponding to the sub-data;

[0117] Determine the reference factor, and perform anomaly scoring according to the average path length and the reference factor to determine the anomaly score corresponding to each sub-data in the sub-segmentation data.

[0118] In some embodiments, during the process of determining the target behavior characteristics corresponding to the target object according to the target anomaly data, the anomaly recognition module 203 performs the following:

[0119] Obtain the target adjacent data corresponding to the target anomaly data from the target behavior data;

[0120] Determine the target change trend corresponding to the object under the target abnormal data according to the target proximity data and the target abnormal data;

[0121] Extract time-domain features and frequency-domain features from the target abnormal data to obtain target-related features corresponding to the target abnormal data;

[0122] Perform feature fusion on the target-related features and the target change trend to obtain the target behavior features corresponding to the target object.

[0123] In some embodiments, during the process of the view recognition module 204 performing news view recognition on the initial news data to obtain target behavior views related to the behavior of the target object, it performs:

[0124] Perform event element recognition on the initial news data to obtain corresponding target event information, and determine the first association information between the initial news data according to the element co-occurrence information between the target event information;

[0125] Filter out relevant news data associated with the initial news data according to the first association information, and obtain the second association information between the initial news data and the relevant news data from the first association information;

[0126] Perform sentiment recognition on the initial news data to obtain corresponding first sentiment information and perform sentiment recognition on the relevant news data to obtain corresponding second sentiment information;

[0127] And determine the third association information between the initial news data and the relevant news data according to the distance between the first sentiment information and the second sentiment information;

[0128] Fuse the second association information and the third association information to determine the target representation information corresponding to the initial news data under the relevant news data;

[0129] Perform key information recognition on the initial news data according to the target representation information to obtain the target key sentences corresponding to the initial news data;

[0130] Perform behavior classification according to the target key sentences to obtain the target behavior views related to the behavior of the target object;

[0131] Among them, the target representation information is obtained according to the following formula:

[0132]

[0133] Among them, represents the target representation information corresponding to the i-th initial news data, represents the damping coefficient, and num represents the number of news corresponding to the relevant news data. represents the weight information corresponding to the second associated information. represents the weight information corresponding to the third associated information. represents the second associated information between the i-th initial news data and the k-th relevant news data. represents the third associated information between the i-th initial news data and the k-th relevant news data. represents the second associated information between the k-th relevant news data and the j-th relevant news data. represents the third associated information between the k-th relevant news data and the j-th relevant news data. represents the target characterization information corresponding to the k-th relevant news data among the news data associated with the k-th relevant news data.

[0134] In some embodiments, during the process of the risk assessment module 205 calculating the distance based on the target behavior characteristics and the target behavior view to obtain the gap information between the target behavior characteristics and the target behavior view, and performing a risk assessment on the target object based on the gap information to obtain the risk assessment result corresponding to the target object, the following operations are executed:

[0135] Use the first feature characterization layer of the risk assessment model to perform feature representation on the target behavior characteristics to obtain first feature information.

[0136] Use the second feature characterization layer of the risk assessment model to perform feature representation on the target behavior view to obtain second feature information.

[0137] Use the distance calculation layer of the risk assessment model to calculate the distance based on the first feature information and the second feature information to obtain the gap information between the target behavior characteristics and the target behavior view.

[0138] Use the risk classification layer of the risk assessment model to perform data classification based on the gap information to obtain the target risk type between the target behavior characteristics and the target behavior view.

[0139] Determine the risk assessment result corresponding to the target object based on the target risk type.

[0140] In some embodiments, after the risk assessment module 205 obtains the risk assessment result, the following operations are further executed:

[0141] Determine preset information, compare the preset information with the risk assessment result, and when the risk assessment result meets the preset information, generate a risk description statement corresponding to the risk assessment result according to the target behavior characteristics and the target behavior view;

[0142] Determine a risk reminder result according to the risk description statement and the risk assessment result, and send the risk reminder result to a target communication device.

[0143] In some embodiments, the digital economy risk management and assessment system 200 based on cloud computing can be applied to a terminal device.

[0144] It should be noted that those skilled in the art can clearly understand that for the convenience and brevity of description, the specific working process of the above-described digital economy risk management and assessment system 200 based on cloud computing can refer to the corresponding process in the foregoing embodiments of the digital economy risk management and assessment method based on cloud computing, and will not be described in detail here.

[0145] The embodiment of the present invention further provides a storage medium for computer-readable storage. The storage medium stores one or more programs, and the one or more programs can be executed by one or more processors to implement the steps of any one of the digital economy risk management and assessment methods provided in the specification of the embodiment of the present invention.

[0146] Among them, the storage medium may be an internal storage unit of the terminal device described in the foregoing embodiment, such as the hard disk or memory of the terminal device. The storage medium may also be an external storage device of the terminal device, such as a plug-in hard disk equipped on the terminal device, a smart media card (SMC), a secure digital (SD) card, a flash card, etc.

[0147] Those of ordinary skill in the art can understand that all or some of the steps in the methods disclosed above, and the functional modules / units in systems and devices, can be implemented as software, firmware, hardware, and their appropriate combinations. In a hardware embodiment, the division between the functional modules / units mentioned above does not necessarily correspond to the division of physical components; for example, one physical component can have multiple functions, or one function or step can be executed by several physical components in cooperation. Some or all physical components can be implemented as software executed by a processor, such as a central processing unit, a digital signal processor, or a microprocessor, or implemented as hardware, or implemented as an integrated circuit, such as an application-specific integrated circuit. Such software can be distributed on a computer-readable medium, which can include a computer storage medium (or non-transitory medium) and a communication medium (or transitory medium). As is well known to those of ordinary skill in the art, the term computer storage medium includes volatile and non-volatile, removable and non-removable media implemented in any method or technology for storing information, such as computer-readable instructions, data structures, program modules, or other data. The computer storage medium includes, but is not limited to, RAM, ROM, EEPROM, flash memory, or other memory technologies, CD-ROM, digital versatile disk (DVD), or other optical disk storage, magnetic cartridges, tapes, magnetic disk storage, or other magnetic storage devices, or any other medium that can be used to store the desired information and can be accessed by a computer. In addition, it is well known to those of ordinary skill in the art that a communication medium typically contains computer-readable instructions, data structures, program modules, or other data in a modulated data signal such as a carrier wave or other transmission mechanism, and can include any information delivery medium.

[0148] It should be understood that the term "and / or" used in the specification and appended claims of the present invention refers to any combination and all possible combinations of one or more of the associated listed items, and includes these combinations. It should be noted that in this article, the term "comprises", "comprising", or any other variation thereof is intended to cover a non-exclusive inclusion, such that a process, method, article, or system that comprises a series of elements includes not only those elements but also other elements not expressly listed, or elements that are inherent to such process, method, article, or system. Without further limitation, an element defined by the statement "comprising an..." does not exclude the presence of additional identical elements in the process, method, article, or system that comprises the element.

[0149] The serial numbers of the embodiments of the present invention above are only for description and do not represent the advantages or disadvantages of the embodiments. As mentioned above, the above are only specific embodiments of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art within the technical scope disclosed by the present invention can easily think of various equivalent modifications or substitutions, and these modifications or substitutions should all be covered within the protection scope of the present invention. Therefore, the protection scope of the present invention shall be subject to the protection scope of the claims.

Claims

1. A risk management assessment method for the digital economy based on cloud computing, characterized in that, The method includes: Collecting the initial behavior data corresponding to the target object and the initial news data related to the behavior of the target object according to the cloud computing platform corresponding to the target object; Performing association analysis on the initial behavior data to obtain a target association relationship, and filling the initial behavior data according to the target association relationship to obtain target behavior data; Performing anomaly identification on the target behavior data to obtain target anomaly data corresponding to the target object, and determining target behavior characteristics corresponding to the target object according to the target anomaly data; Performing news view identification on the initial news data to obtain target behavior views related to the behavior of the target object; Calculating the distance according to the target behavior characteristics and the target behavior views to obtain the gap information between the target behavior characteristics and the target behavior views, and performing risk assessment on the target object according to the gap information to obtain a risk assessment result corresponding to the target object; Wherein, filling the initial behavior data according to the target association relationship to obtain target behavior data includes: Performing missing detection on the initial behavior data to obtain target missing data corresponding to a first parameter in the initial behavior data; Obtaining target adjacent data corresponding to the target missing data under the first parameter from the initial behavior data; Obtaining a second parameter corresponding to the first parameter according to the target association relationship, and obtaining initial associated data corresponding to the target adjacent data under the second parameter from the initial behavior data; Establishing a target conversion relationship corresponding to the first parameter and the second parameter according to the target adjacent data and the initial associated data; Obtaining a target missing position corresponding to the target missing data, and obtaining target associated data corresponding to the second parameter at the target missing position from the initial behavior data; Performing data prediction on the target missing data according to the target conversion relationship in combination with the target associated data to obtain target filling data; Filling the target missing data according to the target filling data to obtain the target behavior data; Wherein, performing news view identification on the initial news data to obtain target behavior views related to the behavior of the target object includes: Performing event element identification on the initial news data to obtain corresponding target event information, and determining first association information between the initial news data according to the element co-occurrence information between the target event information; Screening out relevant news data associated with the initial news data according to the first association information, and obtaining second association information between the initial news data and the relevant news data from the first association information; Performing sentiment identification on the initial news data to obtain corresponding first sentiment information and performing sentiment identification on the relevant news data to obtain corresponding second sentiment information; And determining third association information between the initial news data and the relevant news data according to the distance between the first sentiment information and the second sentiment information; Fuse the second associated information and the third associated information to determine the target representation information corresponding to the initial news data under the relevant news data; Identify key information from the initial news data according to the target representation information to obtain the target key sentence corresponding to the initial news data; Perform behavior classification according to the target key sentence to obtain the target behavior view related to the behavior of the target object; Among them, the target representation information is obtained according to the following formula: ； Among them, represents the target representation information corresponding to the i-th initial news data, represents the damping coefficient, and num represents the number of news corresponding to the relevant news data, represents the weight information corresponding to the second association information, represents the weight information corresponding to the third association information, represents the second association information between the i-th initial news data and the k-th relevant news data; represents the third association information between the i-th initial news data and the k-th relevant news data, represents the second association information between the k-th relevant news data and the j-th relevant news data; represents the third association information between the k-th relevant news data and the j-th relevant news data; represents the target representation information corresponding to the k-th relevant news data among the news data associated with the k-th relevant news data; Among them, the distance is calculated according to the target behavior feature and the target behavior view to obtain the gap information between the target behavior feature and the target behavior view, and the target object is risk-assessed according to the gap information to obtain the risk assessment result corresponding to the target object, including: Use the first feature representation layer of the risk assessment model to perform feature representation on the target behavior feature to obtain the first feature information; Use the second feature representation layer of the risk assessment model to perform feature representation on the target behavior view to obtain the second feature information; Use the distance calculation layer of the risk assessment model to calculate the distance according to the first feature information and the second feature information to obtain the gap information between the target behavior feature and the target behavior view; Use the risk classification layer of the risk assessment model to perform data classification according to the gap information to obtain the target risk type between the target behavior feature and the target behavior view; Determine the risk assessment result corresponding to the target object according to the target risk type.

2. The method according to claim 1, wherein The abnormal identification of the target behavior data to obtain the target abnormal data corresponding to the target object includes: Obtain any sub-behavior data from the target behavior data, and divide the target behavior data according to the sub-behavior data to obtain the first behavior data and the second behavior data; Calculate the data difference between the first behavior data and the second behavior data to obtain the corresponding behavior difference value; Obtain the maximum difference value from all the behavior difference values, and use statistical significance to calculate the target probability corresponding to each behavior difference value less than or equal to the maximum difference value; Determine a preset threshold, and divide the target behavior data according to the preset threshold and the target probability to obtain the target segmentation data; Perform abnormal identification on each sub-segmentation data in the target segmentation data to obtain the target abnormal data corresponding to the target object.

3. The method according to claim 2, wherein The abnormal identification of each sub-segmentation data in the target segmentation data to obtain the target abnormal data corresponding to the target object includes: Perform abnormal scoring on each sub-data in the sub-segmentation data to obtain the abnormal score corresponding to the sub-data; Perform data clustering on all the abnormal scores according to a preset quantity to obtain the first clustering data and the second clustering data corresponding to the abnormal scores; Obtain the first quantity corresponding to the first clustering data and the second quantity corresponding to the second clustering data; Calculate the difference between the first quantity and the second quantity to obtain a target difference, and obtain the sub-abnormal data corresponding to the sub-segmentation data from the first clustering data and the second clustering data according to the target difference; Fuse all the sub-abnormal data to obtain the target abnormal data corresponding to the target object.

4. The method according to claim 3, wherein The abnormal scoring of each sub-data in the sub-segmentation data to obtain the abnormal score corresponding to the sub-data includes: Analyze the sub-segmentation data based on the isolation forest algorithm to obtain the target isolation tree corresponding to the sub-segmentation data; Obtain the current path length corresponding to the sub-data according to the target isolation tree, and perform mean calculation according to the current path length to obtain the average path length corresponding to the sub-data; Determine a reference factor, and perform abnormal scoring according to the average path length and the reference factor to determine the abnormal score corresponding to each sub-data in the sub-segmentation data.

5. The method according to claim 1, characterized in that The determining the target behavior characteristics corresponding to the target object according to the target abnormal data includes: Obtain the target adjacent data corresponding to the target abnormal data from the target behavior data; Determine the target change trend corresponding to the object under the target abnormal data according to the target adjacent data and the target abnormal data; Extract time-domain features and frequency-domain features from the target abnormal data to obtain the target related features corresponding to the target abnormal data; Perform feature fusion on the target related features and the target change trend to obtain the target behavior characteristics corresponding to the target object.

6. The method according to claim 1, wherein After obtaining the risk assessment result, the method further includes: Determine preset information, compare the preset information with the risk assessment result, and when the risk assessment result meets the preset information, generate a risk description statement corresponding to the risk assessment result according to the target behavior characteristics and the target behavior view; Determine a risk reminder result according to the risk description statement and the risk assessment result, and send the risk reminder result to the target communication device.

7. A digital economy risk management assessment system based on cloud computing, characterized in that, Including: A data collection module, configured to collect the initial behavior data corresponding to the target object and the initial news data related to the behavior of the target object according to the cloud computing platform corresponding to the target object; A data processing module for performing association analysis on the initial behavior data to obtain a target association relationship, and filling the initial behavior data according to the target association relationship to obtain target behavior data; wherein, the filling the initial behavior data according to the target association relationship to obtain target behavior data includes: performing missing detection on the initial behavior data to obtain target missing data corresponding to a first parameter in the initial behavior data; obtaining target adjacent data corresponding to the target missing data under the first parameter from the initial behavior data; obtaining a second parameter corresponding to the first parameter according to the target association relationship, and obtaining initial associated data corresponding to the target adjacent data under the second parameter from the initial behavior data; establishing a target conversion relationship corresponding to the first parameter and the second parameter according to the target adjacent data and the initial associated data; obtaining a target missing position corresponding to the target missing data, and obtaining target associated data corresponding to the second parameter at the target missing position from the initial behavior data; predicting the target missing data according to the target conversion relationship and the target associated data to obtain target filling data; filling the target missing data with the target filling data to obtain the target behavior data; An anomaly recognition module for performing anomaly recognition on the target behavior data to obtain target anomaly data corresponding to the target object, and determining target behavior characteristics corresponding to the target object according to the target anomaly data; An opinion recognition module, which is used to perform news opinion recognition on the initial news data to obtain target behavior opinions related to the behavior of the target object. Among them, the process of performing news opinion recognition on the initial news data to obtain target behavior opinions related to the behavior of the target object includes: performing event element recognition on the initial news data to obtain corresponding target event information, and determining the first association information between the initial news data according to the co-occurrence information of the elements between the target event information; screening out relevant news data associated with the initial news data according to the first association information, and obtaining the second association information between the initial news data and the relevant news data from the first association information; performing sentiment recognition on the initial news data to obtain corresponding first sentiment information and performing sentiment recognition on the relevant news data to obtain corresponding second sentiment information; and determining the third association information between the initial news data and the relevant news data according to the distance between the first sentiment information and the second sentiment information; fusing the second association information and the third association information to determine the target representation information corresponding to the initial news data under the relevant news data; performing key information recognition on the initial news data according to the target representation information to obtain the target key sentences corresponding to the initial news data; performing behavior classification according to the target key sentences to obtain the target behavior opinions related to the behavior of the target object; among them, the target representation information is obtained according to the following formula: ； Among them, represents the target characterization information corresponding to the i-th initial news data, represents the damping coefficient, num represents the number of news corresponding to the relevant news data, represents the weight information corresponding to the second association information, represents the weight information corresponding to the third association information, represents the second association information between the i-th initial news data and the k-th relevant news data; represents the third association information between the i-th initial news data and the k-th relevant news data, represents the second association information between the k-th relevant news data and the j-th relevant news data; represents the third association information between the k-th relevant news data and the j-th relevant news data; represents the target characterization information corresponding to the k-th relevant news data among the news data associated with the k-th relevant news data; A risk assessment module, which is used to calculate the distance between the target behavior characteristics and the target behavior opinions to obtain the gap information between the target behavior characteristics and the target behavior opinions, and perform risk assessment on the target object according to the gap information to obtain the risk assessment result corresponding to the target object; among them, the process of calculating the distance between the target behavior characteristics and the target behavior opinions to obtain the gap information between the target behavior characteristics and the target behavior opinions, and performing risk assessment on the target object according to the gap information to obtain the risk assessment result corresponding to the target object includes: using the first feature representation layer of the risk assessment model to perform feature representation on the target behavior characteristics to obtain first feature information; using the second feature representation layer of the risk assessment model to perform feature representation on the target behavior opinions to obtain second feature information; using the distance calculation layer of the risk assessment model to calculate the distance according to the first feature information and the second feature information to obtain the gap information between the target behavior characteristics and the target behavior opinions; using the risk classification layer of the risk assessment model to perform data classification according to the gap information to obtain the target risk type between the target behavior characteristics and the target behavior opinions; determining the risk assessment result corresponding to the target object according to the target risk type.

Images