Telecom fraud identification and analysis system based on intelligent analysis

Abstract

The present invention relates to the field of telecommunications fraud monitoring and identification technology, and specifically to a telecommunications fraud identification and analysis system based on intelligent analysis. The system of the present invention includes a data acquisition module, a fraud type classification module, a telecommunications fraud judgment module, and a reminder module. The fraud type classification module obtains the fraud type of historical fraud text data, and then the telecommunications fraud judgment module accurately matches and analyzes the current call text information with each fraud type to determine whether the current call is telecommunications fraud. Through the system of the present invention, the retrieval process is simplified and the retrieval speed is improved, while the ability to identify and judge fraud content is improved, which helps to better protect the rights and interests of users.

Description

Technical Field

[0001] The present invention relates to the technical field of telecommunication fraud monitoring and identification, and in particular to a telecommunication fraud identification and analysis system based on intelligent analysis. Background Art

[0002] The development of telecommunications fraud identification technology stems from the rapid development of big data, artificial intelligence and network security technology. With the explosive growth of data volume and the improvement of computing power, traditional rules and expert systems have been unable to cope with complex and changeable fraud methods. Machine learning and deep learning algorithms can automatically extract and learn features from massive data to improve recognition efficiency and accuracy. Advances in natural language processing technology have made the analysis of text and voice data more in-depth. At the same time, cloud computing and distributed processing technologies such as Hadoop and Spark provide powerful data processing and storage capabilities, laying the foundation for real-time monitoring and rapid response.

[0003] The conventional method of identifying telecommunications fraud is to have the user mark the mobile phone number, analyze the content of the related calls, and then compare it with the content stored in the fraud database to determine the degree of match. However, with the continuous evolution and upgrading of fraud methods, it is no longer enough to simply use the type or method of fraud as a retrieval standard for classification and storage. Since each fraud case has differences in inducement techniques, context descriptions, and keyword usage, the size of the database continues to expand, resulting in a significant increase in the time required to retrieve historical cases, which in turn affects the retrieval efficiency and even causes problems of misidentification and missed identification. Summary of the invention

[0004] In order to solve the above technical problems, the purpose of the present invention is to provide a telecommunications fraud identification and analysis system based on intelligent analysis.

[0005] According to the telecommunications fraud identification and analysis system based on intelligent analysis provided in the embodiment of the present application, the technical solutions adopted specifically include: A data acquisition module, used to obtain historical fraud text data and current call text information; A fraud type classification module, used to analyze the historical fraud text data, obtain the sentiment fluctuation value of the historical fraud text data, and obtain the fraud type according to the sentiment fluctuation value; A telecommunications fraud determination module, configured to analyze the degree of guidance of the user party based on the current call text information, and analyze the degree of matching between the current call text information and the fraud type, to determine whether the current call is telecommunications fraud; The reminder module is used to remind the user of the risk of the call when the telecommunications fraud judgment module determines that the current call is telecommunications fraud.

[0006] In some embodiments of the present application, the fraud type classification module includes a historical data preprocessing unit, a sentiment fluctuation value unit, and a clustering unit; The historical data preprocessing unit is used to preprocess the historical fraud text data; The emotion fluctuation value unit is used to analyze the historical fraud text data to obtain the emotion fluctuation value of the historical fraud text data; The clustering unit is used to perform cluster analysis on the historical fraud text data according to the sentiment fluctuation value and in combination with the high-frequency adjective vectors and high-frequency transaction noun vectors in the historical fraud text data to obtain the fraud type.

[0007] In some embodiments of the present application, the sentiment fluctuation value unit is configured to: analyze the number of transaction nouns and adjectives in the historical fraud text data and the positional relationship between the two to obtain the sentiment fluctuation value of the historical fraud text data.

[0008] In some embodiments of the present application, the clustering unit is configured to: use the sentiment fluctuation value and the high-frequency adjective vectors and high-frequency transaction noun vectors in the historical fraud text data as three-dimensional indicators, use Euclidean distance to calculate the difference distance of each historical fraud text data, and use the K-means clustering method to perform cluster analysis on the historical fraud text data to obtain the fraud type.

[0009] In some embodiments of the present application, the telecommunications fraud determination module includes a current data preprocessing unit, a guided degree unit, a matching degree unit, and a fraud risk unit; The current data preprocessing unit is used to preprocess the current call text information; The guidance degree unit is used to analyze the interaction relationship between the user and the other party in the current call text information to obtain the guidance degree of the user; The matching degree unit is used to analyze the similarity between the current call text information and the transaction nouns and adjectives in the fraud types, and obtain the matching degree between the current call text information and all the fraud types; The fraud risk unit is used to obtain the fraud risk of the current call according to the guidance degree and the matching degree.

[0010] In some embodiments of the present application, the telecommunications fraud determination module further includes a determination unit, in which a risk threshold is preset and configured as follows: Comparing the fraud risk with the preset risk threshold; If the fraud risk is greater than or equal to the preset risk threshold, the current call is determined to be telecommunications fraud; If the fraud risk is less than the preset risk threshold, the current call is determined to be a normal call.

[0011] In some embodiments of the present application, the guidance degree unit is configured to: analyze the proportion of the number of conversation words of the other party in the current call text information, and analyze the correlation between the conversation content of the other party and the user in the current call text information to obtain the guidance degree of the user.

[0012] In some embodiments of the present application, the matching degree unit is configured to: analyze the number of transaction nouns with greater similarity between the current call text information and the transaction nouns in the fraud type, analyze the maximum similarity between a transaction noun in the current call text information and all transaction nouns in the fraud type, and analyze the similarity between the adjectives in the current call text information and the adjectives that appear most frequently in the fraud type, to obtain the matching degree between the current call text information and all the fraud types.

[0013] In some embodiments of the present application, the system further comprises: The data collection module is used to collect the text information of the current call into the fraud database when the telecommunications fraud judgment module determines that the current call is telecommunications fraud.

[0014] In some embodiments of the present application, the data acquisition module is also used to obtain current call-related information, which includes device IP address, device information, and caller number; and to obtain blacklist data and whitelist, which include device IP address blacklist, device information blacklist, and caller number blacklist, and the whitelist includes official numbers; the telecommunications fraud judgment module is also configured to compare the current call-related information with the blacklist and the whitelist to perform a preliminary screening of the current call.

[0015] Compared with the prior art, the telecommunications fraud identification and analysis system based on intelligent analysis provided by the present invention has the following beneficial effects: The present invention obtains historical fraud text data and current call text information through a data acquisition module to provide a data basis for subsequent analysis. The fraud type classification module analyzes the historical fraud text data, obtains the emotional fluctuation value of the historical fraud text data, and obtains the fraud type according to the emotional fluctuation value. The system of the present invention classifies the historical fraud text data by the emotional fluctuation value of the historical fraud text data, avoiding the problem of inaccurate classification caused by the complexity of the fraud methods, and providing a basic guarantee for subsequent classification matching. The telecommunications fraud judgment module analyzes the degree of guidance of the user party based on the current call text information, and analyzes the degree of matching between the current call text information and the fraud type, and determines whether the current call is telecommunications fraud; the system of the present invention no longer compares and analyzes the current call with all the data in the fraud database, but only needs to compare all types of fraud classifications, which greatly improves the retrieval speed, and thus improves the response speed; at the same time, it improves the ability to identify and judge fraud content, which helps to better protect the rights and interests of users. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] In order to more clearly illustrate the technical solutions and advantages in the embodiments of the present invention or the prior art, the drawings required for use in the embodiments or the prior art descriptions are briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0017] Figure 1 A schematic diagram of the basic composition of a telecommunications fraud identification and analysis system based on intelligent analysis provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0018] In order to further explain the technical means and effects adopted by the present invention to achieve the predetermined invention purpose, the specific implementation method, structure, features and effects of the telecommunications fraud identification and analysis system based on intelligent analysis proposed by the present invention are described in detail below in combination with the accompanying drawings and preferred embodiments. In the following description, different "one embodiment" or "another embodiment" does not necessarily refer to the same embodiment. In addition, specific features, structures or characteristics in one or more embodiments may be combined in any suitable form.

[0019] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those commonly understood by those skilled in the art to which the present invention belongs. Terms such as "comprises", "comprising" or any other variants thereof are intended to cover non-exclusive inclusion, so that a circuit structure, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such article or device. In the absence of further restrictions, an element defined by the phrase "comprising a ..." does not exclude the presence of other identical elements in the article or device including the element.

[0020] The specific scheme of a telecommunications fraud identification and analysis system based on intelligent analysis provided by the present invention is described in detail below in conjunction with the accompanying drawings.

[0021] See also Figure 1 , which shows the basic composition of a telecommunications fraud identification and analysis system based on intelligent analysis provided by an embodiment of the present invention.

[0022] like Figure 1 As shown, a telecommunications fraud identification and analysis system based on intelligent analysis provided by an embodiment of the present invention mainly includes a data acquisition module 1, a fraud type classification module 2, a telecommunications fraud judgment module 3, a reminder module 4 and a data collection module 5.

[0023] Specifically, the data acquisition module 1 is used to obtain historical fraud text data and current call text information. The fraud database collects a large amount of fraud voice data and SMS data, and converts the voice data into corresponding text content according to language recognition technology, and stores all processed fraud information in the database. Based on this, the data acquisition module 1 is configured to obtain historical fraud text data from the fraud database. In addition, when the user is making a current call or receiving a current text message, the data acquisition module 1 is configured to obtain the voice data of the current call content during the current call, and convert the voice data into corresponding text content according to language recognition technology, that is, to obtain the current call text information. Further, the data acquisition module 1 is also configured to obtain the device IP address, device information, and caller number of the caller or the source of the text message of the current call, as well as the number of historical calls and text messages of both numbers, the duration of a single call, the call type and other information. In addition, the data acquisition module 1 is also configured to obtain its blacklist and whitelist data from relevant government websites, banks and other financial institutions, wherein the blacklist data includes a device IP address blacklist, a device information blacklist, and a caller number blacklist, and the whitelist includes official numbers.

[0024] The fraud type classification module 2 is used to analyze historical fraud text data, obtain the sentiment fluctuation value of the historical fraud text data, and obtain the fraud type according to the sentiment fluctuation value.

[0025] For telecommunications fraud, the fraudulent means generally guide the victims through different emotions and different expressions. By guiding the victims, they are asked for information to carry out fraudulent behavior. In this regard, we can analyze the fraud information obtained from the fraud database, obtain the thinking guidance methods in different fraud cases by semantic splitting in the historical fraud text data, and classify a large amount of historical fraud text data according to the thinking guidance methods, so as to obtain the commonality of the guidance logic under different fraud modes.

[0026] First, the fraud type classification module 2 includes a historical data preprocessing unit 21. The historical data preprocessing unit 21 is used to preprocess the historical fraud text data to obtain word vectors. Specifically, the historical data preprocessing unit 21 is configured to remove non-text symbols in the historical fraud text data based on the historical fraud text data corresponding to a large number of fraud cases obtained from the fraud database, and use word segmentation tools such as THULAC to segment the historical fraud text data into multiple words or vocabulary units. Furthermore, the Word2vec model is used to process the segmented historical fraud text data to generate a vocabulary, and a word vector is assigned to each word in the vocabulary, which is a multidimensional vector; and then the multidimensional data dimensionality reduction technology is used to compress the word vector into a two-dimensional space.

[0027] In order to classify a large amount of historical fraud text data, it is necessary to judge the similarity between the corresponding related fraud texts. For conventional historical fraud text data matching judgment, it is usually obtained by analyzing the similarity of the word vectors of related nouns of the same type. However, there are many types of current fraud methods, and the corresponding related nouns of the same type are weakened in the fraud information, making the final matching effect poor. However, the changes in the specific emotional guidance of related telecommunications fraud are difficult to weaken. For this reason, we need to analyze and judge the emotional guidance before and after the transaction words in the historical fraud text data to obtain the corresponding changes in emotional fluctuations. In this way, classification is carried out according to the corresponding emotional guidance method. When the user's call content is subsequently detected in real time, the other party's emotional guidance method before and after the transaction words is judged to match and judge with various types of fraud.

[0028] Based on the above analysis, in this embodiment, the fraud type classification module 2 is configured to analyze historical fraud text data, obtain the sentiment fluctuation value of the historical fraud text data, and obtain the fraud type according to the sentiment fluctuation value.

[0029] Specifically, the fraud type classification module 2 also includes an emotion fluctuation value unit 22. The emotion fluctuation value unit 22 is used to analyze the historical fraud text data and obtain the emotion fluctuation value of the historical fraud text data. Further, the emotion fluctuation value unit 22 is configured to: analyze the number of transaction nouns and adjectives in the historical fraud text data and the positional relationship between the two, and obtain the emotion fluctuation value of the historical fraud text data.

[0030] First, the segmented historical fraud text data is divided into nouns, adjectives, verbs, etc. according to semantics. The main purpose of the fraud text corresponding to the historical fraud text data is to guide the victim to conduct relevant fund transactions with it by adopting various methods, so as to achieve the purpose of fraud. In this regard, the fraud type classification module 2 is configured to obtain the relevant transaction nouns appearing in the historical fraud text data, and obtain the number and position of the corresponding transaction nouns by matching the commonly used transaction nouns with their corresponding transaction noun vectors. At the same time, the fraud type classification module 2 is configured to obtain the corresponding adjectives in the historical fraud text data, obtain the number of word vectors corresponding to the adjectives and the positional relationship between the adjectives and the transaction nouns. Then, the emotional guidance trend of the historical fraud text data is obtained by analyzing the change in the number of adjectives in the historical fraud text data. The emotional fluctuation value of a single historical fraud text data is obtained by analyzing the emotional guidance trend of the historical fraud text data and the emotional difference before and after the number of times the transaction nouns appear and the corresponding position.

[0031] Furthermore, the calculation formula for the sentiment fluctuation value of historical fraud text data is: In the formula, Indicates that in the fraud database The sentiment fluctuation value of historical fraud text data; Indicates The number of times transaction terms appear in historical fraud text data; Indicates In the historical fraud text data The word vector of the adjective before the transaction noun; Indicates In the historical fraud text data The word vector of the adjective after the transaction noun; Indicates In the historical fraud text data The word interval between a transaction noun and the adjective vector closest to it, where the word interval represents the number of other word vectors that exist between two word vectors; Indicates In the historical fraud text data The word interval between a transaction noun and the nearest adjective vector behind it, where the word interval represents the number of other word vectors between the two word vectors; Indicates The adjective vectors that appear most frequently in the historical fraud text data; It means to find the covariance of two sets of vectors.

[0032] express and The covariance of The difference between the adjective word vector and the word distance before and after the transaction noun. When the value is smaller, it means that the sentiment change of the corresponding historical fraud text data is more uniform. When the value is larger, it means that the sentiment change of the corresponding historical fraud text data is more intense. Indicates The adjective vector that appears most frequently in the historical fraud text data represents the The dominant emotional mode in historical fraud text data; the emotional fluctuation value of the historical fraud text data is obtained through the emotional fluctuation degree in the historical fraud text data and its dominant emotional mode.

[0033] The fraud type classification module 2 also includes a clustering unit 23. The clustering unit 23 is used to perform cluster analysis on the historical fraud text data according to the sentiment fluctuation value and in combination with the high-frequency adjective vectors and high-frequency transaction noun vectors in the historical fraud text data to obtain the fraud type.

[0034] In the above, the sentiment fluctuation value unit 22 deeply analyzes the sentiment fluctuation and dominant sentiment mode before and after the transaction nouns of the historical fraud text data corresponding to the fraud case, and obtains the sentiment fluctuation value in the historical fraud text data. In view of the diversity of fraud methods and patterns in fraud cases, it is necessary to further use the comparison of sentiment fluctuation values, high-frequency adjective vectors and high-frequency transaction noun vectors to evaluate the similarity between different historical fraud text data, thereby realizing the classification of historical fraud text data.

[0035] Therefore, further, the clustering unit 23 is configured to: use the sentiment fluctuation value and the high-frequency adjective vector and high-frequency transaction noun vector in the historical fraud text data as three-dimensional indicators, use the Euclidean distance to calculate the difference distance of each historical fraud text data, and use the K-means clustering method to cluster the historical fraud text data to obtain the fraud type. Among them, the high-frequency adjective vector and the high-frequency transaction noun vector represent the adjective with the most occurrences and the transaction noun with the most occurrences in the historical fraud text data, respectively. Further, when determining the optimal k value in the K-means clustering analysis, the elbow method is used to ensure the accuracy and stability of the clustering effect.

[0036] Finally, the different historical fraud text data in the fraud database are divided into multiple clusters through the fraud type classification module 2. The historical fraud text data in each cluster represents a specific fraud type, that is, the fraud type is obtained.

[0037] The telecommunications fraud judgment module 3 is used to analyze the degree of guidance of the user based on the current call text information, and analyze the degree of matching between the current call text information and the fraud type, to determine whether the current call is telecommunications fraud.

[0038] In order to monitor in real time whether the incoming call information received by the user involves telecommunications fraud, the present embodiment takes the following measures: First, the telecommunications fraud judgment module 3 is configured to compare the current call-related information with the blacklist and the whitelist, and to preliminarily screen the current call, that is, the telecommunications fraud judgment module 3 preliminarily screens the incoming call number, the device IP address, the device information, etc., and accurately distinguishes and detects the unfamiliar calls. In this process, the data in the blacklist and the whitelist obtained by the data acquisition module 1 are compared. For the official numbers in the whitelist, normal calls are guaranteed to ensure that normal communication is not disturbed; for the device IP address, device information, and incoming call number in the blacklist, the current call is directly judged to be telecommunications fraud, and the system will immediately issue a warning to the user, thereby improving the judgment efficiency. For other unfamiliar incoming call information, the telecommunications fraud judgment module 3 uniformly analyzes and detects the content of the current call text information. That is, the telecommunications fraud judgment module 3 is configured to analyze the degree of guidance of the user based on the current call text information, and analyze the degree of matching between the current call text information and the fraud type, and judge whether the current call is telecommunications fraud.

[0039] Furthermore, the telecommunications fraud judgment module 3 includes a current data preprocessing unit 31 , a guided degree unit 32 , a matching degree unit 33 and a fraud risk unit 34 .

[0040] Specifically, the current data preprocessing unit 31 is used to preprocess the current call text information.

[0041] When the user is talking to an unfamiliar caller, the guided degree unit 32 is configured to obtain the degree of guidance between the user and the other party through the conversation content between the two. If the degree of guidance of the user during the conversation is relatively high, then the call is a telecommunications fraud and the risk of the user being defrauded is relatively high; if the degree of guidance of the user during the conversation is relatively low, then the call may be a telecommunications fraud but the risk of the user being defrauded is relatively low; therefore, it is necessary to judge the degree of guidance of the user from the call information. In this regard, the guided degree unit 32 is configured to split the current call text information based on the current call text information obtained by the data acquisition module 1 using a word segmentation tool, and after processing the split current call text information, use the word2vec model to process the split current call text information, generate a vocabulary, and assign a word vector to each word in the vocabulary, which is a multi-dimensional vector; and use multi-dimensional data dimensionality reduction technology to compress the word vector into a two-dimensional space.

[0042] The guidance degree unit 32 is used to analyze the interaction relationship between the user and the other party in the current call text information to obtain the guidance degree of the user. Further, the guidance degree unit 32 is configured to: analyze the proportion of the number of conversation words of the other party in the current call text information, and analyze the correlation between the conversation content of the other party and the user in the current call text information to obtain the guidance degree of the user.

[0043] In the above, each word vector in the current call text information corresponding to the current call is obtained through the current data preprocessing unit 31. Further, the guidance degree unit 32 is configured to use the LSTM neural network to analyze the vector corresponding to each paragraph through each word vector in each paragraph, and analyze the proportion of the number of words of the other party and the number of words of both parties in each paragraph of the current call text information, and analyze the correlation between the content of the conversation between the other party and the user in each paragraph of the current call text information to analyze the degree of guidance of the user. When the proportion of the content of the conversation between the other party and the user is small, and the correlation between the content of both parties in the current call text information is weak, it is considered that the degree of guidance received by the user is low.

[0044] Furthermore, the calculation formula of the user's degree of guidance is: In the formula, Indicates the current The degree to which users are guided in unfamiliar calls; Indicates the current The number of words spoken by the other party in each unfamiliar call; Indicates the current The number of words spoken by both parties in each unfamiliar call; Indicates the current The number of interactions between the two parties in a stranger call; Indicates the current In the first The content vector of the user at the time of the interaction; Indicates the current In the first The content vector of the other party during the first interaction; Indicates calculating the cosine similarity between the two.

[0045] It indicates the proportion of the other party's words in the conversation between the two parties, that is, the proportion of the other party's conversation content. The larger the proportion, the stronger the other party's dominant position in the conversation, indicating that the user is more guided; It indicates the correlation of the call contents of all interactions between the current caller and the user. The stronger the corresponding overall average correlation, the stronger the degree of guidance of the user.

[0046] The matching degree unit 33 is used to analyze the similarity between the current call text information and the transaction nouns and adjectives in the fraud type, and obtain the matching degree between the current call text information and all fraud types.

[0047] For the current call text information, the corresponding words are divided into nouns, adjectives, verbs, etc. The main purpose of the fraud information is to induce the victim to conduct financial transactions with it. Therefore, the matching degree unit 33 is configured to focus on identifying the transaction nouns in the current call text information. When a corresponding single transaction noun is identified, the current call text information is matched and analyzed with the historical fraud text data corresponding to each fraud type obtained by the fraud type classification module 2 to obtain the matching degree of the current call text information with all fraud types.

[0048] Furthermore, the matching degree unit 33 is configured to: analyze the number of transaction nouns with greater similarity between the current call text information and the transaction nouns in the fraud type, analyze the maximum similarity between a transaction noun in the current call text information and all transaction nouns in the fraud type, and analyze the similarity between the adjective in the current call text information and the adjective that appears most frequently in the fraud type, to obtain the matching degree between the current call text information and all fraud types. The calculation formula for the matching degree between the current call text information and all fraud types is: In the formula, Indicates the current The current call text information corresponding to the first unknown call The matching degree of the fraud type; Indicates the current The nouns that appear in the current call text message corresponding to the unknown caller and the Cosine similarity of noun vectors appearing in the fraud-like type; Indicates getting the current The nouns that appear in the current call text message corresponding to the unknown caller and the The number of nouns with large cosine similarity (greater than 95%) that appear in the fraud-like type; Indicates the current The number of transaction nouns in the current call text message corresponding to the unknown incoming calls; Indicates Any transaction noun vector that appears in the fraud-like type; Indicates the current The current call text message corresponding to the unknown call transaction noun vectors; Indicates the current The current call text message corresponding to the unknown call Transaction terms and The largest covariance of all transaction nouns appearing in the fraud type indicates the similarity between two vectors; Indicates the current The current call text message corresponding to the unknown call adjective vectors; Indicates The most frequent adjective vectors in the fraud-like category; Indicates the current The adjectives that appear in the current call text message corresponding to the first unknown caller are The covariance of the most frequent adjectives in the fraud-like type indicates the similarity between the two vectors. It means to find the covariance of two sets of vectors.

[0049] The above analysis is performed by analyzing the similarity between the transaction nouns in the current call text information and the transaction nouns in the historical fraud text data corresponding to a certain type of fraud, as well as the main sentiment similarity between the two, and combining the number of nouns with high similarity between the two to obtain the similarity of the fraud behavior corresponding to the current call text information. The current call text information is matched with each fraud type in parallel to obtain the corresponding fraud behavior similarity.

[0050] The fraud risk unit 34 is used to obtain the fraud risk of the current call according to the guidance degree and the matching degree.

[0051] The degree of guidance of the user of the current call and the degree of matching between the text information of the current call and all fraud types are obtained respectively through the degree of guidance unit 32 and the degree of matching unit 33. Based on this, the fraud risk unit 34 of this embodiment is configured to obtain the fraud risk of the current call according to the degree of guidance and the degree of matching. The calculation formula of the fraud risk of the current call is: In the formula, Indicates the current The risk of fraud from unfamiliar phone calls; Indicates the current The degree to which users are guided in unfamiliar calls; Indicates the current The maximum matching degree between the current call text information corresponding to each unfamiliar call and various fraud types; represents the linear normalization function.

[0052] The telecommunications fraud judgment module 3 also includes a judgment unit 35, in which a risk threshold is preset, and the preset risk threshold can be 0.85. The judgment unit 35 is configured to: compare the fraud risk with the preset risk threshold; if the fraud risk is greater than or equal to the preset risk threshold, that is , judge the current call as telecommunications fraud; if the fraud risk is less than the preset risk threshold, , determine that the current call is a normal call.

[0053] The reminder module 4 is used to remind the user of the risk of the call when the telecommunication fraud judgment module 3 judges that the current call is telecommunication fraud. The reminder method includes SMS reminder, phone reminder, etc.

[0054] The data collection module 5 is used to collect the text information of the current call into the fraud database when the telecommunications fraud judgment module 3 judges that the current call is telecommunications fraud. The purpose is to expand the data of telecommunications fraud types, thereby improving the judgment accuracy and response speed of telecommunications fraud information.

[0055] In addition, the system of this embodiment monitors telephone communications in real time and preliminarily screens telephone request information based on the fraud labeling list provided by relevant institutions. For non-blacklist telephone information, the call content is further analyzed and risk assessment is performed through vocabulary matching technology. Once it is found that the call content has a large degree of match with the fraud type recorded in the fraud database, the system will immediately issue a warning to the user to remind him / her of the possible risks, thereby helping the user to avoid economic losses in a timely manner.

[0056] It should be noted that the sequence of the above embodiments of the present invention is only for description and does not represent the advantages and disadvantages of the embodiments. The processes depicted in the accompanying drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0057] The various embodiments in this specification are described in a progressive manner, and the same or similar parts between the various embodiments can be referenced to each other, and each embodiment focuses on the differences from other embodiments.

Claims

1. A telecommunications fraud identification and analysis system based on intelligent analysis, It is characterized in that The system comprises: A data acquisition module, used to obtain historical fraud text data and current call text information; A fraud type classification module, used to analyze the historical fraud text data, obtain the sentiment fluctuation value of the historical fraud text data, and obtain the fraud type according to the sentiment fluctuation value; A telecommunications fraud determination module, configured to analyze the degree of guidance of the user party based on the current call text information, and analyze the degree of matching between the current call text information and the fraud type, to determine whether the current call is telecommunications fraud; The reminder module is used to remind the user of the risk of the call when the telecommunications fraud judgment module determines that the current call is telecommunications fraud.

2. According to the telecommunications fraud identification and analysis system based on intelligent analysis according to claim 1, It is characterized in that The fraud type classification module includes a historical data preprocessing unit, a sentiment fluctuation value unit and a clustering unit; The historical data preprocessing unit is used to preprocess the historical fraud text data; The emotion fluctuation value unit is used to analyze the historical fraud text data to obtain the emotion fluctuation value of the historical fraud text data; The clustering unit is used to perform cluster analysis on the historical fraud text data according to the sentiment fluctuation value and in combination with the high-frequency adjective vectors and high-frequency transaction noun vectors in the historical fraud text data to obtain the fraud type.

3. According to the telecommunications fraud identification and analysis system based on intelligent analysis according to claim 2, It is characterized in that The sentiment fluctuation value unit is configured to: analyze the number of transaction nouns and adjectives in the historical fraud text data and the positional relationship between the two, and obtain the sentiment fluctuation value of the historical fraud text data.

4. According to claim 2, the telecommunications fraud identification and analysis system based on intelligent analysis, It is characterized in that The clustering unit is configured to: use the sentiment fluctuation value and the high-frequency adjective vectors and high-frequency transaction noun vectors in the historical fraud text data as three-dimensional indicators, use Euclidean distance to calculate the difference distance of each historical fraud text data, and use the K-means clustering method to perform cluster analysis on the historical fraud text data to obtain the fraud type.

5. According to claim 1, the telecommunications fraud identification and analysis system based on intelligent analysis, It is characterized in that The telecommunications fraud judgment module includes a current data preprocessing unit, a guided degree unit, a matching degree unit and a fraud risk unit; The current data preprocessing unit is used to preprocess the current call text information; The guidance degree unit is used to analyze the interaction relationship between the user and the other party in the current call text information to obtain the guidance degree of the user; The matching degree unit is used to analyze the similarity between the current call text information and the transaction nouns and adjectives in the fraud types, and obtain the matching degree between the current call text information and all the fraud types; The fraud risk unit is used to obtain the fraud risk of the current call according to the guidance degree and the matching degree.

6. According to claim 5, the telecommunications fraud identification and analysis system based on intelligent analysis, It is characterized in that The telecommunications fraud determination module further includes a determination unit, wherein a risk threshold is preset in the determination unit and is configured as follows: Comparing the fraud risk with the preset risk threshold; If the fraud risk is greater than or equal to the preset risk threshold, the current call is determined to be telecommunications fraud; If the fraud risk is less than the preset risk threshold, the current call is determined to be a normal call.

7. According to claim 5, the telecommunications fraud identification and analysis system based on intelligent analysis, It is characterized in that The guidance degree unit is configured to: analyze the proportion of the other party's conversation words in the current conversation text information, and analyze the correlation between the conversation content between the other party and the user in the current conversation text information to obtain the user's guidance degree.

8. According to claim 5, the telecommunications fraud identification and analysis system based on intelligent analysis, It is characterized in that The matching degree unit is configured to: analyze the number of transaction nouns with greater similarity between the current call text information and the transaction nouns in the fraud type, analyze the maximum similarity between a transaction noun in the current call text information and all transaction nouns in the fraud type, and analyze the similarity between the adjectives in the current call text information and the adjectives that appear most frequently in the fraud type, to obtain the matching degree between the current call text information and all the fraud types.

9. The telecommunications fraud identification and analysis system based on intelligent analysis according to claim 1, It is characterized in that The system further comprises: The data collection module is used to collect the text information of the current call into the fraud database when the telecommunications fraud judgment module determines that the current call is telecommunications fraud.

10. The telecommunications fraud identification and analysis system based on intelligent analysis according to claim 1, It is characterized in that The data acquisition module is also used to obtain current call-related information, which includes device IP address, device information, and incoming call number; and to obtain blacklist data and whitelist, which include device IP address blacklist, device information blacklist, and incoming call number blacklist, and the whitelist includes official numbers; the telecommunications fraud judgment module is also configured to compare the current call-related information with the blacklist and the whitelist to perform a preliminary screening of the current call.

Images