An interactive risk prediction method and system based on artificial intelligence
By obtaining browsing content and parameters in real time in the interactive port, calculating the selection probability and keyword groups, and determining the risk value, the problem that account-based supervision cannot identify the account usage in real time is solved, and security is improved, real-time account usage is achieved. Supervision.
Patent Information
- Application Number
- CN202411131614.3
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2024-08-19
- Publication Date
- 2024-11-05
- Estimated Expiration
- 2044-08-19
AI Technical Summary
The existing account-based supervision method is tested when logging in, and the account usage cannot be identified in real time, resulting in a high risk of data leakage and the inability to provide real-time risk identification and supervision.
Insert the interactive recording port in the interactive port, obtain browsing content and browsing parameters in real time, calculate exposure and selection probability, select browsing content in descending order of selection probability, perform text recognition and keyword extraction, determine risk values based on keyword groups, and use AI model Carry out real-time supervision.
Real-time supervision of account usage is realized, security is improved, changes in account usage can be identified in a timely manner, and the risk of data leakage is reduced.
Smart Images

Figure CN118657382B_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of risk prediction, and in particular to an interactive risk prediction method and system based on artificial intelligence. Background Art
[0002] Many existing companies or fields will build databases dedicated to members. The contents of these databases are corporate secrets. Once leaked, it will cause certain losses. Therefore, it is necessary to supervise the use process of members. The conventional supervision method is based on account-based supervision and issuing internal accounts. This method is feasible, but once the account is misused, it will still cause data leakage. The reason for this problem is that the account-based supervision method is instantaneous and is only detected at login. If a more real-time risk identification solution can be provided, it can greatly improve security. Summary of the invention
[0003] The purpose of the present invention is to provide an interactive risk prediction method and system based on artificial intelligence to solve the problems raised in the above background technology.
[0004] To achieve the above object, the present invention provides the following technical solutions:
[0005] An interactive risk prediction method based on artificial intelligence, the method comprising:
[0006] Inserting an interactive recording port into the interactive port, and acquiring browsing content and browsing parameters in a preset period in real time based on the interactive recording port; the browsing parameters include the number of browsing times and single browsing duration of each browsing content;
[0007] Calculating the exposure of each browsing content based on the browsing parameters, and determining the selection probability of each browsing content based on the exposure;
[0008] Select browsing content in descending order of selection probability, create a text library, perform text recognition on the text library, and extract keywords;
[0009] The keywords of a certain interaction record port at different times are arranged in chronological order to obtain keyword groups at different times, and the risk value is determined according to the keyword groups.
[0010] As a further solution of the present invention: the step of inserting an interaction recording port into the interaction port and acquiring browsing content and browsing parameters within a preset time period in real time based on the interaction recording port includes:
[0011] Inserting an interaction record port into an interaction port, and simultaneously creating and publishing a port insertion table; the port insertion table is used to indicate which interaction ports have the interaction record port inserted into them;
[0012] Receive confirmation information fed back by the user and activate the corresponding interaction recording port; wherein the confirmation process is performed at least once each time the interaction port is run;
[0013] Based on the interactive recording port, the browsing instructions are monitored in real time. When a browsing instruction is detected, the browsing content pointed to by the browsing instruction is obtained, and the effective time of the browsing content is recorded in real time;
[0014] The browsing period is determined according to the generation time and effective duration of the browsing instruction as a browsing parameter; wherein the number of browsing periods corresponds to the number of browsing times, and the length of each browsing period corresponds to the single browsing duration.
[0015] As a further solution of the present invention: the step of calculating the exposure of each browsing content based on the browsing parameters and determining the selection probability of each browsing content based on the exposure includes:
[0016] Based on the browsing content, statistics are collected on the browsing period of the browsing content in all interactive ports within a preset time length;
[0017] Create a time point according to a preset time step, and for each time point, obtain the total number of browsing time periods including the time point;
[0018] Count the total number of each time point in chronological order as a feature array;
[0019] Determining a selection probability based on the feature array;
[0020] The relationship between the selection probability and the feature array is:
[0021] ;
[0022] In the formula, is the selection probability, is the preset normalization function used to convert Convert to a value between 0 and 1. and is the preset weight; is the first numerical values, is the total number of values in the feature array.
[0023] As a further solution of the present invention: the steps of selecting browsing content in descending order of selection probability, creating a text library, performing text recognition on the text library, and extracting keywords include:
[0024] receiving a text library size, selecting browsing content in sequence based on a selection probability, and inputting the text library until the text library reaches the text library size; the text library size at least includes the data volume;
[0025] Perform word segmentation on each browsing content in the text library, remove the separator words based on a preset separator word list, and obtain word clusters; the word cluster is a collection of word groups, and each word group corresponds to each browsing content one by one;
[0026] Calculating the TF-IDF value of each word in the word group, and determining the key degree of each word according to the TF-IDF value;
[0027] A key degree range is received, and words are selected as keywords according to the key degree range.
[0028] As a further solution of the present invention, the steps of selecting browsing content in descending order of selection probability, creating a text library, performing text recognition on the text library, and extracting keywords also include:
[0029] Regularly traverse and identify the text library based on the AI recognition model to extract keywords;
[0030] Randomly generate a preset number of key degree ranges, and obtain keywords corresponding to each key degree range;
[0031] Compare the keywords extracted by the AI recognition model with the keywords extracted based on the criticality and calculate the similarity;
[0032] The key degree range corresponding to the maximum similarity is selected as the new key degree range.
[0033] As a further solution of the present invention: the step of arranging the keywords of a certain interactive record port at different times in chronological order to obtain keyword groups at different times, and determining the risk value according to the keyword groups comprises:
[0034] Arrange keywords of a certain interactive record port at different times in chronological order to obtain keyword groups at different times;
[0035] Input the keyword group into a preset word vector conversion model to obtain a vector corresponding to each keyword to obtain a vector group;
[0036] Compare the vector groups at adjacent moments, calculate the difference rate, and determine the risk value based on the difference rate; the greater the difference rate, the greater the risk value.
[0037] The technical solution of the present invention also provides an interactive risk prediction system based on artificial intelligence, the system comprising:
[0038] A browsing information acquisition module, which is used to insert an interactive recording port into the interactive port, and acquire the browsing content and browsing parameters within a preset period in real time based on the interactive recording port; the browsing parameters include the number of views of each browsing content and the duration of a single view;
[0039] A selection probability calculation module, used to calculate the exposure of each browsing content based on the browsing parameters, and determine the selection probability of each browsing content based on the exposure;
[0040] The keyword extraction module is used to select browsing content in descending order of selection probability, create a text library, perform text recognition on the text library, and extract keywords;
[0041] The risk value calculation module is used to arrange the keywords of a certain interaction record port at different times according to the time sequence, obtain the keyword groups at different times, and determine the risk value according to the keyword groups.
[0042] As a further solution of the present invention: the browsing information acquisition module includes:
[0043] A record port insertion unit, used to insert an interactive record port into an interactive port, and to simultaneously create and publish a port insertion table; the port insertion table is used to indicate which interactive ports have the interactive record port inserted into them;
[0044] A recording port activation unit, used to receive confirmation information fed back by the user and activate the corresponding interactive recording port; wherein the confirmation process is performed at least once each time the interactive port is run;
[0045] A recording execution unit, used to monitor browsing instructions in real time based on the interactive recording port, and when a browsing instruction is detected, obtain the browsing content pointed to by the browsing instruction, and record the effective time of the browsing content in real time;
[0046] The parameter generation unit is used to determine the browsing period according to the generation time and effective duration of the browsing instruction as the browsing parameter; wherein the number of browsing periods corresponds to the number of browsing times, and the length of each browsing period corresponds to the single browsing duration.
[0047] As a further solution of the present invention: the selection probability calculation module includes:
[0048] A time period statistics unit, used to count the browsing time periods of the browsing content in all interactive ports within a preset time length based on the browsing content;
[0049] A total number acquisition unit, used to create a time point according to a preset time step, and for each time point, acquire the total number of browsing periods including the time point;
[0050] The total number counting unit is used to count the total number of each time point in chronological order as a feature array;
[0051] An array recognition unit, used to determine a selection probability based on the feature array;
[0052] The relationship between the selection probability and the feature array is:
[0053] ;
[0054] In the formula, is the selection probability, is the preset normalization function used to convert Convert to a value between 0 and 1. and is the preset weight; is the first numerical values, is the total number of values in the feature array.
[0055] As a further solution of the present invention: the keyword extraction module includes:
[0056] A text library creation unit, configured to receive a text library scale, select browsing contents in sequence based on a selection probability, and input the text library until the text library reaches the text library scale; the text library scale includes at least a data volume;
[0057] A word segmentation unit is used to segment each browsing content in the text library, remove the separator words based on a preset separator word list, and obtain a word cluster; the word cluster is a collection of word groups, and each word group corresponds to each browsing content one by one;
[0058] A key degree calculation unit, used to calculate the TF-IDF value of each word in the word group, and determine the key degree of each word according to the TF-IDF value;
[0059] The keyword selection unit is used to receive a key degree range and select words as keywords according to the key degree range.
[0060] Compared with the prior art, the beneficial effects of the present invention are: the present invention inserts an interactive record port in the interactive port, monitors the usage records of the account, extracts keywords from the usage records, and compares the keywords at different times to determine whether the usage of the account has changed significantly, thereby performing real-time supervision on the account and greatly improving security. BRIEF DESCRIPTION OF THE DRAWINGS
[0061] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings in the following description are only some embodiments of the present invention.
[0062] Figure 1 The flowchart of the interactive risk prediction method based on artificial intelligence is shown in FIG.
[0063] Figure 2 This is a structural diagram of the interactive risk prediction system based on artificial intelligence. DETAILED DESCRIPTION
[0064] In order to make the technical problems, technical solutions and beneficial effects to be solved by the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.
[0065] Figure 1 The flowchart of the interactive risk prediction method based on artificial intelligence is as follows. In an embodiment of the present invention, an interactive risk prediction method based on artificial intelligence includes:
[0066] Step S100: inserting an interaction recording port into the interaction port, and acquiring browsing content and browsing parameters in a preset period in real time based on the interaction recording port; the browsing parameters include the number of browsing times and single browsing duration of each browsing content;
[0067] The technical solution of the present invention is applied to a limited information interaction platform. The information centralized management party obtains information regularly, stores it in the database, and then shares it to some designated ports. These are internal materials and are only available to member users. Member users can access the internal materials through the ports.
[0068] The device used by member users to interact with the platform is called an interactive port. The platform administrator will insert an interactive recording port in the interactive port. The interactive recording port is used to obtain the browsing process of member users in real time, specifically, to obtain the browsing content and browsing parameters within a preset time period in real time. The browsing parameters include the number of views of each browsing content and the duration of a single browsing. In layman's terms, it means what content the user has viewed and how long each content was viewed.
[0069] It should be noted that before obtaining the browsing process of member users, the platform administrator must clearly inform the member users. On the one hand, it is convenient for management, prevents the outflow of information, and ensures that it is only available for member users to view. On the other hand, the member users are informed that their browsing process is under supervision. If the member users still choose to view the information, they are assumed to be granted monitoring permissions.
[0070] Step S200: calculating the exposure of each browsing content based on the browsing parameters, and determining the selection probability of each browsing content based on the exposure;
[0071] The exposure of each browsed content is calculated based on the browsing parameters, and the browsing parameters include the number of browse times and the duration of a single browse. The more the number of browse times, the higher the exposure. The longer the single browse duration, the higher the exposure. The exposure is converted into a selection probability according to preset rules. The higher the exposure, the greater the selection probability.
[0072] Regarding the above-mentioned number of browsing times and single browsing time, since the peripheral device of the interactive port is generally a display screen with a certain size, one browsing process often only browses one document. If it has a split-screen function, two documents can be browsed. This application pins a document (one browsing content) once, that is, one browsing process is regarded as one number of browsing times, and the duration of each browsing is called a single browsing time. It is worth mentioning that in actual use, member users will perform multiple operations when comparing documents. At this time, the number of browsing times of some documents will be very high, and the duration of each browsing will be very short.
[0073] Step S300: selecting browsing contents in descending order of selection probability, creating a text library, performing text recognition on the text library, and extracting keywords;
[0074] The purpose of constructing a text library in this application is to extract appropriate keywords based on the text library. The text library needs to be of a certain size, which may be only a few thousand articles, while the reading volume of member users may be very large. Therefore, it is necessary to select some documents from the documents read by member users, count the selected documents, and obtain the text library; the documents are the browsing content in this application; after the text library is created, the text library is identified to extract keywords.
[0075] Among them, the process of selecting some documents from the documents read by member users is determined by the selection probability in the above content. The greater the exposure, the higher the selection probability, and the corresponding document has a greater possibility of being a part of the text library.
[0076] From the above content, we can also know that the more times a member user watches and the longer the single viewing time, the higher the exposure.
[0077] Step S400: Arrange keywords of a certain interaction record port at different times in chronological order to obtain keyword groups at different times, and determine risk values according to the keyword groups;
[0078] After the above operation, each interaction record port can obtain some keywords, that is, a keyword group. The keyword groups at different times are counted, and the risk value can be determined according to the difference between the keyword groups at different times. The greater the difference between the keyword groups at different times, the higher the risk value.
[0079] For step S100 in the above content, a specific embodiment is provided as follows:
[0080] The step of inserting an interactive recording port into the interactive port and acquiring browsing content and browsing parameters within a preset period in real time based on the interactive recording port includes:
[0081] Inserting an interaction record port into an interaction port, and simultaneously creating and publishing a port insertion table; the port insertion table is used to indicate which interaction ports have the interaction record port inserted into them;
[0082] Receive confirmation information fed back by the user and activate the corresponding interaction recording port; wherein the confirmation process is performed at least once each time the interaction port is run;
[0083] Based on the interactive recording port, the browsing instructions are monitored in real time. When a browsing instruction is detected, the browsing content pointed to by the browsing instruction is obtained, and the effective time of the browsing content is recorded in real time;
[0084] The browsing period is determined according to the generation time and effective duration of the browsing instruction as a browsing parameter; wherein the number of browsing periods corresponds to the number of browsing times, and the length of each browsing period corresponds to the single browsing duration.
[0085] Based on the description of the process of step S100, the above content provides a more detailed solution, which inserts an interaction record port into an interaction port, and simultaneously creates and publishes a port insertion table. The meaning of publishing is to inform all member users which interaction ports have interaction record ports.
[0086] Each time a member user uses an interactive port, the member user uploads confirmation information. After receiving the confirmation information from the user, the corresponding interactive recording port is activated to monitor the interactive process. The activation process occurs before each use, which means that when using the interactive recording port, the user clearly informs the platform to allow recording and monitoring.
[0087] The browsing instructions are monitored based on the interactive recording port. When the browsing instructions are monitored, the browsing content pointed to by the browsing instructions is obtained. The effective time of the browsing content is recorded in real time based on the browsing recording port. The effective time is the time when it is pinned on the display screen.
[0088] The browsing period is determined according to the generation time and effective duration of the browsing instruction as a browsing parameter; wherein the number of browsing periods corresponds to the number of displays, and the length of each browsing period corresponds to the single display duration.
[0089] Regarding step S200, the step of calculating the exposure of each browsing content based on the browsing parameters and determining the selection probability of each browsing content based on the exposure includes:
[0090] Based on the browsing content, statistics are collected on the browsing period of the browsing content in all interactive ports within a preset time length;
[0091] Create a time point according to a preset time step, and for each time point, obtain the total number of browsing time periods including the time point;
[0092] Count the total number of each time point in chronological order as a feature array;
[0093] A selection probability is determined based on the feature array.
[0094] In actual application, all interactive ports may be browsing the browsing content. The statistical scheme provided by the present application is to count the browsing time periods of the browsing content in all interactive ports within a preset time length (one day or one working period) for any browsing content, based on the browsing content. Then, create time points according to the preset time steps. For each time point, obtain which browsing time periods intersect with the time point, and record the number of intersecting browsing time periods as the feature of each time point; count the total number of each time point in chronological order as the exposure, count the exposure of each time point as an exposure array, and analyze the feature array to determine the selection probability.
[0095] The relationship between the selection probability and the feature array is:
[0096] ;
[0097] In the formula, is the selection probability, is the preset normalization function used to convert Convert to a value between 0 and 1. and is the preset weight; is the first numerical values, is the total number of values in the feature array.
[0098] In actual applications, the more times R&D personnel browse, the longer their browsing time. will be bigger, and in addition, Indicates the change of browsing behavior over time. The more times you browse, the larger this value will be, and the greater the final selection probability will be.
[0099] It should be noted that the method for calculating each selection probability in this application is to calculate the sum of the values corresponding to all feature arrays, and then divide each value by the sum to obtain a value from 0 to 1 as the selection probability.
[0100] Regarding step S300, the steps of selecting browsing content in descending order of selection probability, creating a text library, performing text recognition on the text library, and extracting keywords include:
[0101] receiving a text library size, selecting browsing content in sequence based on a selection probability, and inputting the text library until the text library reaches the text library size; the text library size at least includes a data volume;
[0102] Perform word segmentation on each browsing content in the text library, remove the separator words based on a preset separator word list, and obtain word clusters; the word cluster is a collection of word groups, and each word group corresponds to each browsing content one by one;
[0103] Calculating the TF-IDF value of each word in the word group, and determining the key degree of each word according to the TF-IDF value;
[0104] A key degree range is received, and words are selected as keywords according to the key degree range.
[0105] The above content provides a certain keyword extraction scheme. First, the text library size is determined. The text library size indicates how much data there is in the text library at most. Based on the selection probability, browsing content is randomly selected and input into the text library. The content in the text library will increase until the text library size is reached.
[0106] Each browsing content in the text library is segmented into words, and the separator words are removed based on the preset separator word list to obtain word groups. Each browsing content corresponds to a word group, and the text library corresponds to a word group cluster.
[0107] Calculate the TF-IDF value of each word in the word group, and determine the key degree of each word based on the TF-IDF value. The simplest way is to use the TF-IDF value directly as the key degree. At this time, select words whose key degree belongs to the preset key degree range as keywords.
[0108] Among them, regarding the TF-IDF value, TF is the frequency of a word in a document, and IDF is the inverse frequency of the word in all documents. By multiplying the two values, we can get a numerical value that reflects the importance of the word, that is, the TF-IDF value. In fact, a weight can be introduced between the TF-IDF value and the criticality to adjust the value space of the criticality.
[0109] Furthermore, the steps of selecting browsing content in descending order of selection probability, creating a text library, performing text recognition on the text library, and extracting keywords also include:
[0110] Regularly traverse and identify the text library based on the AI recognition model to extract keywords;
[0111] Randomly generate a preset number of key degree ranges, and obtain keywords corresponding to each key degree range;
[0112] Compare the keywords extracted by the AI recognition model with the keywords extracted based on the criticality and calculate the similarity;
[0113] The key degree range corresponding to the maximum similarity is selected as the new key degree range.
[0114] The above content provides an auxiliary adjustment solution. The keyword extraction process based on keywords is actually very dependent on the criticality range. The criticality range is a range, not a threshold. Words greater than or less than the threshold are keywords, but words belonging to the range are keywords. Therefore, the value of the range is very important.
[0115] This application introduces an AI-assisted solution, which uses an AI recognition model to extract keywords, and then randomly generates some key ranges. The AI recognition results evaluate the key ranges. The more consistent the results of using the key ranges are with the AI recognition results, the more accurate the key ranges are. Among them, the consistency is evaluated by similarity, and the greater the similarity, the more consistent it is.
[0116] Regarding step S400, the step of arranging the keywords of a certain interaction record port at different times in chronological order to obtain keyword groups at different times, and determining the risk value according to the keyword groups includes:
[0117] Arrange keywords of a certain interactive record port at different times in chronological order to obtain keyword groups at different times;
[0118] Input the keyword group into a preset word vector conversion model to obtain a vector corresponding to each keyword to obtain a vector group;
[0119] Compare the vector groups at adjacent moments, calculate the difference rate, and determine the risk value based on the difference rate; the greater the difference rate, the greater the risk value.
[0120] The above content provides a specific risk value determination scheme. For a certain interaction record port, its keywords at different times are obtained to obtain keyword groups at different times. Each keyword in the keyword group is input into a preset word vector conversion model to obtain the vector corresponding to each keyword. The keyword group is then converted into a vector group, and the vector groups at adjacent times are compared to calculate the difference rate. The risk value is determined based on the difference rate. The larger the difference rate, the greater the risk value.
[0121] It is worth mentioning that the difference rate is a decreasing function of the similarity. Vector similarity is easy to calculate. Therefore, in actual use, the vector similarity is generally calculated and then the difference rate is calculated.
[0122] Figure 2 : is a structural diagram of an interactive risk prediction system based on artificial intelligence. In an embodiment of the present invention, an interactive risk prediction system based on artificial intelligence, the system 10 includes:
[0123] The browsing information acquisition module 11 is used to insert an interaction record port in the interaction port, and acquire the browsing content and browsing parameters in a preset period in real time based on the interaction record port; the browsing parameters include the number of views of each browsing content and the single browsing duration;
[0124] A selection probability calculation module 12, configured to calculate the exposure of each browsing content based on the browsing parameters, and determine the selection probability of each browsing content based on the exposure;
[0125] The keyword extraction module 13 is used to select browsing content in descending order of selection probability, create a text library, perform text recognition on the text library, and extract keywords;
[0126] The risk value calculation module 14 is used to arrange the keywords of a certain interaction record port at different times according to the time sequence, obtain the keyword groups at different times, and determine the risk value according to the keyword groups.
[0127] Furthermore, the browsing information acquisition module 11 includes:
[0128] A record port insertion unit, used to insert an interactive record port into an interactive port, and to simultaneously create and publish a port insertion table; the port insertion table is used to indicate which interactive ports have the interactive record port inserted into them;
[0129] A recording port activation unit, used to receive confirmation information fed back by the user and activate the corresponding interactive recording port; wherein the confirmation process is performed at least once each time the interactive port is run;
[0130] A recording execution unit, used to monitor browsing instructions in real time based on the interactive recording port, and when a browsing instruction is detected, obtain the browsing content pointed to by the browsing instruction, and record the effective time of the browsing content in real time;
[0131] The parameter generation unit is used to determine the browsing period according to the generation time and effective duration of the browsing instruction as the browsing parameter; wherein the number of browsing periods corresponds to the number of browsing times, and the length of each browsing period corresponds to the single browsing duration.
[0132] Specifically, the selection probability calculation module 12 includes:
[0133] A time period statistics unit, used to count the browsing time periods of the browsing content in all interactive ports within a preset time length based on the browsing content;
[0134] A total number acquisition unit, used to create a time point according to a preset time step, and for each time point, acquire the total number of browsing periods including the time point;
[0135] The total number counting unit is used to count the total number of each time point in chronological order as a feature array;
[0136] An array recognition unit, used to determine a selection probability based on the feature array;
[0137] The relationship between the selection probability and the feature array is:
[0138] ;
[0139] In the formula, is the selection probability, is the preset normalization function used to convert Convert to a value between 0 and 1. and is the preset weight; is the first numerical values, is the total number of values in the feature array.
[0140] Specifically, the keyword extraction module 13 includes:
[0141] A text library creation unit, configured to receive a text library scale, select browsing contents in sequence based on a selection probability, and input the text library until the text library reaches the text library scale; the text library scale includes at least a data volume;
[0142] A word segmentation unit is used to segment each browsing content in the text library, remove the separator words based on a preset separator word list, and obtain a word cluster; the word cluster is a collection of word groups, and each word group corresponds to each browsing content one by one;
[0143] A key degree calculation unit, used to calculate the TF-IDF value of each word in the word group, and determine the key degree of each word according to the TF-IDF value;
[0144] The keyword selection unit is used to receive a key degree range and select words as keywords according to the key degree range.
[0145] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present invention should be included in the protection scope of the present invention.
Claims
1. An interactive risk prediction method based on artificial intelligence, characterized in that: The method comprises: Inserting an interactive recording port into the interactive port, and acquiring browsing content and browsing parameters in a preset period in real time based on the interactive recording port; the browsing parameters include the number of browsing times and single browsing duration of each browsing content; Calculating the exposure of each browsing content based on the browsing parameters, and determining the selection probability of each browsing content based on the exposure; Select browsing content in descending order of selection probability, create a text library, perform text recognition on the text library, and extract keywords; Arrange keywords of a certain interactive record port at different times in chronological order to obtain keyword groups at different times, and determine the risk value based on the keyword groups; The step of calculating the exposure of each browsing content based on the browsing parameters and determining the selection probability of each browsing content based on the exposure comprises: Based on the browsing content, statistics are collected on the browsing period of the browsing content in all interactive ports within a preset time length; Create a time point according to a preset time step, and for each time point, obtain the total number of browsing time periods including the time point; Count the total number of each time point in chronological order as a feature array; Determining a selection probability based on the feature array; The relationship between the selection probability and the feature array is: ; In the formula, is the selection probability, is the preset normalization function used to convert Convert to a value between 0 and 1. and is the preset weight; is the first numerical values, is the total number of values in the feature array; The method used in this application to calculate each selection probability is to calculate the sum of the values corresponding to all feature arrays, and then divide each value by the sum to obtain a value between 0 and 1 as the selection probability; The steps of selecting browsing content in descending order of selection probability, creating a text library, performing text recognition on the text library, and extracting keywords also include: Regularly traverse and identify the text library based on the AI recognition model to extract keywords; Randomly generate a preset number of key degree ranges, and obtain keywords corresponding to each key degree range; Compare the keywords extracted by the AI recognition model with the keywords extracted based on the criticality and calculate the similarity; Select the key degree range corresponding to the maximum similarity as the new key degree range; The step of arranging keywords of a certain interactive record port at different times in chronological order to obtain keyword groups at different times, and determining the risk value according to the keyword groups comprises: Arrange keywords of a certain interactive record port at different times in chronological order to obtain keyword groups at different times; Input the keyword group into a preset word vector conversion model to obtain a vector corresponding to each keyword to obtain a vector group; Compare the vector groups at adjacent moments, calculate the difference rate, and determine the risk value based on the difference rate; the greater the difference rate, the greater the risk value.
2. The interactive risk prediction method based on artificial intelligence according to claim 1, characterized in that: The step of inserting an interactive recording port into the interactive port and acquiring browsing content and browsing parameters within a preset period in real time based on the interactive recording port includes: Inserting an interaction record port into an interaction port, and simultaneously creating and publishing a port insertion table; the port insertion table is used to indicate which interaction ports have the interaction record port inserted into them; Receive confirmation information fed back by the user and activate the corresponding interaction recording port; wherein the confirmation process is performed at least once each time the interaction port is run; Based on the interactive recording port, the browsing instructions are monitored in real time. When a browsing instruction is detected, the browsing content pointed to by the browsing instruction is obtained, and the effective time of the browsing content is recorded in real time; The browsing period is determined according to the generation time and effective duration of the browsing instruction as a browsing parameter; wherein the number of browsing periods corresponds to the number of browsing times, and the length of each browsing period corresponds to the single browsing duration.
3. The interactive risk prediction method based on artificial intelligence according to claim 1, characterized in that: The steps of selecting browsing content in descending order of selection probability, creating a text library, performing text recognition on the text library, and extracting keywords include: receiving a text library size, selecting browsing content in sequence based on a selection probability, and inputting the text library until the text library reaches the text library size; the text library size at least includes the data volume; Perform word segmentation on each browsing content in the text library, remove the separator words based on a preset separator word list, and obtain word clusters; the word cluster is a collection of word groups, and each word group corresponds to each browsing content one by one; Calculating the TF-IDF value of each word in the word group, and determining the key degree of each word according to the TF-IDF value; A key degree range is received, and words are selected as keywords according to the key degree range.
4. An interactive risk prediction system based on artificial intelligence, characterized in that: The system comprises: A browsing information acquisition module, which is used to insert an interactive recording port into the interactive port, and acquire the browsing content and browsing parameters within a preset period in real time based on the interactive recording port; the browsing parameters include the number of views of each browsing content and the duration of a single view; A selection probability calculation module, used to calculate the exposure of each browsing content based on the browsing parameters, and determine the selection probability of each browsing content based on the exposure; The keyword extraction module is used to select browsing content in descending order of selection probability, create a text library, perform text recognition on the text library, and extract keywords; A risk value calculation module is used to arrange the keywords of a certain interaction record port at different times according to the time sequence, obtain the keyword groups at different times, and determine the risk value according to the keyword groups; The selection probability calculation module includes: A time period statistics unit, used to count the browsing time periods of the browsing content in all interactive ports within a preset time length based on the browsing content; A total number acquisition unit, used to create a time point according to a preset time step, and for each time point, acquire the total number of browsing periods including the time point; The total number counting unit is used to count the total number of each time point in chronological order as a feature array; An array recognition unit, used to determine a selection probability based on the feature array; The relationship between the selection probability and the feature array is: ; In the formula, is the selection probability, is the preset normalization function used to convert Convert to a value between 0 and 1. and is the preset weight; is the first numerical values, is the total number of values in the feature array; The method used in this application to calculate each selection probability is to calculate the sum of the values corresponding to all feature arrays, and then divide each value by the sum to obtain a value between 0 and 1 as the selection probability; The step of selecting browsing content in descending order of selection probability, creating a text library, performing text recognition on the text library, and extracting keywords also includes: Regularly traverse and identify the text library based on the AI recognition model to extract keywords; Randomly generate a preset number of key degree ranges, and obtain keywords corresponding to each key degree range; Compare the keywords extracted by the AI recognition model with the keywords extracted based on the criticality and calculate the similarity; Select the key degree range corresponding to the maximum similarity as the new key degree range; The contents of arranging the keywords of a certain interactive record port at different times in chronological order to obtain keyword groups at different times and determining the risk value according to the keyword groups include: Arrange keywords of a certain interactive record port at different times in chronological order to obtain keyword groups at different times; Input the keyword group into a preset word vector conversion model to obtain a vector corresponding to each keyword to obtain a vector group; Compare the vector groups at adjacent moments, calculate the difference rate, and determine the risk value based on the difference rate; the greater the difference rate, the greater the risk value.
5. The interactive risk prediction system based on artificial intelligence according to claim 4 is characterized in that: The browsing information acquisition module includes: A record port insertion unit, used to insert an interactive record port into an interactive port, and to simultaneously create and publish a port insertion table; the port insertion table is used to indicate which interactive ports have the interactive record port inserted into them; A recording port activation unit, used to receive confirmation information fed back by the user and activate the corresponding interactive recording port; wherein the confirmation process is performed at least once each time the interactive port is run; A recording execution unit, used to monitor browsing instructions in real time based on the interactive recording port, and when a browsing instruction is detected, obtain the browsing content pointed to by the browsing instruction, and record the effective time of the browsing content in real time; The parameter generation unit is used to determine the browsing period according to the generation time and effective duration of the browsing instruction as the browsing parameter; wherein the number of browsing periods corresponds to the number of browsing times, and the length of each browsing period corresponds to the single browsing duration.
6. The interactive risk prediction system based on artificial intelligence according to claim 4 is characterized in that: The keyword extraction module includes: A text library creation unit, configured to receive a text library scale, select browsing contents in sequence based on a selection probability, and input the text library until the text library reaches the text library scale; the text library scale includes at least a data volume; A word segmentation unit is used to segment each browsing content in the text library, remove the separator words based on a preset separator word list, and obtain a word cluster; the word cluster is a collection of word groups, and each word group corresponds to each browsing content one by one; A key degree calculation unit, used to calculate the TF-IDF value of each word in the word group, and determine the key degree of each word according to the TF-IDF value; The keyword selection unit is used to receive a key degree range and select words as keywords according to the key degree range.
