A target object risk behavior early warning method and related device

Abstract

The application discloses a target object risk behavior early warning method and related equipment. In the specific method, emotion analysis is first performed based on a preset emotion analysis model and a plurality of historical information published by a target object, and then it is determined whether the target object is emotionally abnormal. Then, a preset risk behavior analysis model is used to determine the risk behavior type that the target object is most likely to have. Finally, information mining is performed on the historical information corresponding to the risk behavior type, so that various types of key information related to the risk behavior type are clearly presented. Finally, risk behavior early warning of the target object is realized based on the key information related to the risk behavior, which can perform deep risk behavior information mining on the emotionally abnormal object and perform early warning based on the mined risk behavior information, thereby improving the accuracy of risk behavior early warning of the target object.

Description

Technical Field

[0001] This application relates to the field of computer technology, and in particular to a method and related equipment for early warning of risky behavior of a target object. Background Technology

[0002] With the rapid development of social media and online social networks, the public's emotions and psychological state are more easily influenced by the diverse opinions expressed on social media, leading to behaviors that pose certain risks to society. Therefore, monitoring the negative emotions that internet users may experience and the risky behaviors resulting from these negative emotions can effectively address the problems of social media and society, making related technologies increasingly important.

[0003] Traditional solutions typically employ manual methods such as offline surveys, online reporting, and screening of specific key elements to monitor user risk behavior. However, these methods often consume excessive social and human resources, making it difficult to meet large-scale monitoring demands. Current technologies, often based on natural language processing, can only determine a user's emotional inclination based on their posts and comments, but cannot accurately capture the potential risky behaviors and related details based on these emotional inclinations, resulting in low accuracy in predicting online user risk behavior.

[0004] Therefore, how to solve the problem of low accuracy in monitoring the risky behavior of network users in existing technologies has become a technical problem that urgently needs to be solved by those skilled in the art. Summary of the Invention

[0005] In view of the above problems, in order to solve the problem of low accuracy in monitoring network user risk behavior in the prior art, this application provides a method and related equipment for early warning of risk behavior of target objects.

[0006] The embodiments of this application disclose the following technical solutions:

[0007] Firstly, this application discloses a method for early warning of risky behavior of a target object, including:

[0008] The system retrieves multiple historical messages published by a target object within a preset time period; these historical messages are those published by the target object on a network platform.

[0009] Based on a preset emotion analysis model and the multiple historical information entries, it is determined whether the target object has an abnormal emotional state; among the multiple historical information entries published by the object with an abnormal emotional state, there are multiple historical information entries indicating negative emotions.

[0010] If it is determined that the target object's emotional state is abnormal, then risk behavior analysis is performed on the multiple historical information entries representing negative emotions according to a preset risk behavior analysis model to obtain multiple risk behavior types corresponding to each of the multiple historical information entries representing negative emotions, as well as a first risk behavior type; the first risk behavior type includes: among the multiple risk behavior types corresponding to each of the multiple historical information entries representing negative emotions, the risk behavior type whose frequency of occurrence ranks higher than a preset ranking; the preset risk behavior analysis model is trained based on a deep learning model;

[0011] Based on a preset information mining algorithm, risk behavior information mining is performed on historical information corresponding to the first risk behavior type to obtain risk behavior information corresponding to the first risk behavior type.

[0012] Based on the risk behavior information corresponding to the first risk behavior type, a risk behavior warning is issued.

[0013] Optionally, determining whether the target object has an abnormal emotional state based on a preset emotion analysis model and the multiple pieces of historical information specifically includes:

[0014] Based on the preset sentiment analysis model, sentiment analysis is performed on the multiple historical information entries to obtain the sentiment category represented by each of the multiple historical information entries; the sentiment category includes at least: positive sentiment and negative sentiment;

[0015] Determine the proportion of the historical information representing negative emotions among the multiple pieces of historical information;

[0016] If the proportion of historical information representing negative emotions in the multiple pieces of historical information exceeds a preset threshold, then the target object is determined to have an abnormal emotional state.

[0017] Optionally, before determining whether the target object has an abnormal emotional state based on a preset emotion analysis model and the multiple pieces of historical information, the method further includes:

[0018] The multiple historical information entries are preprocessed to obtain multiple processed historical information entries; all of the processed historical information entries are stored in a preset format.

[0019] Vector transformation is performed on the multiple processed historical information entries to obtain the word vector spaces corresponding to each of the multiple processed historical information entries.

[0020] Optionally, the step of performing risk behavior analysis on the multiple historical information entries representing negative emotions according to a preset risk behavior analysis model to obtain multiple risk behavior types corresponding to each of the multiple historical information entries representing negative emotions, as well as a first risk behavior type, specifically includes:

[0021] Based on the preset risk behavior analysis model, event analysis is performed on the multiple historical information entries representing negative emotions to determine the event type corresponding to each of the multiple historical information entries representing negative emotions; the event type represents the type of event that leads the target object to exhibit risky behavior;

[0022] By using the preset risk behavior analysis model, the multiple historical information entries representing negative emotions, and the corresponding event types, the risk behavior types corresponding to each of the multiple historical information entries representing negative emotions, as well as the first risk behavior type, are determined.

[0023] Optionally, the step of issuing a risk behavior warning based on the risk behavior information corresponding to the first risk behavior type specifically includes:

[0024] Based on the risk behavior information corresponding to the first risk behavior type, determine the warning level corresponding to the risk behavior information;

[0025] Based on the warning level and the risk behavior information, a risk behavior warning signal is generated for the target object.

[0026] Optionally, the risk behavior information includes: information about the people, time, and location involved when the target object performs the first risk behavior type.

[0027] Secondly, this application discloses a risk behavior early warning system for a target object, comprising:

[0028] The acquisition module is used to acquire multiple pieces of historical information published by the target object within a preset time period; the historical information refers to the historical information published by the target object on the network platform.

[0029] The emotion analysis module is used to determine whether the target object has an abnormal emotional state based on a preset emotion analysis model and the multiple historical information; among the multiple historical information released by the object with an abnormal emotional state, there are multiple historical information indicating negative emotions;

[0030] The risk behavior analysis module is used to analyze the multiple historical information entries representing negative emotions according to a preset risk behavior analysis model when it is determined that the target object's emotional state is abnormal. This analysis yields multiple risk behavior types corresponding to each of the multiple historical information entries representing negative emotions, as well as a first risk behavior type. The first risk behavior type includes risk behavior types whose frequency of occurrence ranks higher than a preset ranking among the multiple risk behavior types corresponding to each of the multiple historical information entries representing negative emotions. The preset risk behavior analysis model is trained based on a deep learning model.

[0031] The information mining module is used to mine risk behavior information from historical information corresponding to the first risk behavior type based on a preset information mining algorithm, so as to obtain risk behavior information corresponding to the first risk behavior type; the risk behavior information includes: information on the person, location, and time involved in performing the first risk behavior type.

[0032] The early warning module is used to issue early warnings for risk behaviors based on the risk behavior information corresponding to the first risk behavior type.

[0033] Optionally, the sentiment analysis module is specifically used for:

[0034] Based on the preset sentiment analysis model, sentiment analysis is performed on the multiple historical information entries to obtain the sentiment category represented by each of the multiple historical information entries; the sentiment category includes at least: positive sentiment and negative sentiment;

[0035] Determine the proportion of the historical information representing negative emotions among the multiple pieces of historical information;

[0036] If the proportion of historical information representing negative emotions in the multiple pieces of historical information exceeds a preset threshold, then the target object is determined to have an abnormal emotional state.

[0037] Optionally, the system further includes: a preprocessing module; the preprocessing module is specifically used for:

[0038] The multiple historical information entries are preprocessed to obtain multiple processed historical information entries; all of the processed historical information entries are stored in a preset format.

[0039] Vector transformation is performed on the multiple processed historical information entries to obtain the word vector spaces corresponding to each of the multiple processed historical information entries.

[0040] Optionally, the risk behavior analysis module is specifically used for:

[0041] Based on the preset risk behavior analysis model, event analysis is performed on the multiple historical information entries representing negative emotions to determine the event type corresponding to each of the multiple historical information entries representing negative emotions; the event type represents the type of event that leads the target object to exhibit risky behavior;

[0042] By using the preset risk behavior analysis model, the multiple historical information entries representing negative emotions, and the corresponding event types, the risk behavior types corresponding to each of the multiple historical information entries representing negative emotions, as well as the first risk behavior type, are determined.

[0043] Optionally, the early warning module is specifically used for:

[0044] Based on the risk behavior information corresponding to the first risk behavior type, determine the warning level corresponding to the risk behavior information;

[0045] Based on the warning level and the risk behavior information, a risk behavior warning signal is generated for the target object.

[0046] Optionally, the risk behavior information includes: information about the people, time, and location involved when the target object performs the first risk behavior type.

[0047] Thirdly, this application discloses an electronic device, which includes: a processor, a memory, and a system bus;

[0048] The processor and the memory are connected via the system bus;

[0049] The memory is used to store one or more programs, the one or more programs including instructions, which, when executed by the processor, cause the processor to perform the target object risk behavior early warning method.

[0050] Fourthly, this application discloses a computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements the aforementioned method for early warning of risky behavior of the target object.

[0051] Compared with existing technologies, this application has the following beneficial effects: This application discloses a method and related equipment for early warning of risky behavior of a target object. Specifically, the method first acquires multiple historical messages published by the target object on a network platform within a preset time period. Based on a preset sentiment analysis model and the multiple historical messages, it determines whether the target object's emotional state is abnormal. Among the multiple historical messages published by users with abnormal emotional states, there are multiple messages indicating negative emotions. If it is determined that the target object's emotional state is abnormal, risk behavior analysis is performed on the multiple historical messages indicating negative emotions according to the preset risk behavior analysis model, obtaining multiple risk behavior types corresponding to each of the multiple historical messages indicating negative emotions, and a first risk behavior type. The first risk behavior type includes: risk behavior types whose frequency of occurrence is higher than a preset ranking among the multiple risk behavior types corresponding to each of the multiple historical messages indicating negative emotions. Risk behavior information mining is performed on the historical information corresponding to the first risk behavior type to obtain behavioral information corresponding to the first risk behavior type. The behavioral information includes: information about the person, location, and time involved in performing the first risk behavior type. Finally, based on the risk behavior information corresponding to the first risk behavior type, a risk behavior warning for the target object can be completed. In the above method, firstly, the target object with abnormal emotions is identified based on sentiment analysis. Then, the most likely type of risk behavior of the target object is determined through a preset risk behavior analysis model. Finally, information mining is performed on the historical information corresponding to this risk behavior type to clearly present various key information related to this risk behavior type. Finally, a risk behavior warning for the target object is realized based on the key information related to the risk behavior. This method can perform in-depth risk behavior information mining on objects with abnormal emotions and use the mined risk behavior information to issue warnings, thus improving the accuracy of risk behavior warnings for the target object. Attached Figure Description

[0052] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0053] Figure 1 A flowchart illustrating a method for early warning of risky behavior of a target object provided in an embodiment of this application;

[0054] Figure 2 This application provides a schematic diagram of the structure of a target object risk behavior early warning system according to an embodiment of the present application;

[0055] Figure 3 This is a schematic diagram of the structure of an electronic device for early warning of risky behavior of a target object, provided in an embodiment of this application. Detailed Implementation

[0056] As described earlier, with the rapid development of social media and online social networks, the public's emotions and psychological state are more easily influenced by the diverse opinions on social media, leading to behaviors that pose certain risks to society. Therefore, providing early warnings about the negative emotions that internet users may experience and the risky behaviors caused by these negative emotions can effectively address the problems of social media and society, making related technologies increasingly important.

[0057] Traditional solutions typically employ manual methods such as offline surveys, online reporting, and screening of specific key metrics to warn of user risk behavior. However, these methods often consume excessive social and human resources, making it difficult to meet the large-scale demand for risk behavior warnings. Current technologies often rely on natural language processing to categorize online users into positive and negative sentiments. However, this only determines a user's emotional inclination based on their posts and comments, failing to accurately capture the potential risk behaviors and related details inherent in these sentiments, resulting in low accuracy in warning of online user risk behavior.

[0058] Therefore, how to solve the problem of low accuracy in early warning of network user risk behavior in existing technologies has become a technical problem that urgently needs to be solved by those skilled in the art.

[0059] To address the aforementioned issues, this application discloses a method and related equipment for early warning of risky behavior of a target object. Specifically, the method first acquires multiple historical messages posted by the target object on a network platform within a preset time period. Based on a preset sentiment analysis model and the multiple historical messages, it determines whether the target object's emotional state is abnormal. Among the multiple historical messages posted by users with abnormal emotional states, there are several messages indicating negative emotions. If the target object's emotional state is determined to be abnormal, risk behavior analysis is performed on the multiple historical messages indicating negative emotions according to the preset risk behavior analysis model, resulting in multiple risk behavior types corresponding to each of the multiple historical messages indicating negative emotions, and a first risk behavior type. The first risk behavior type includes risk behavior types whose frequency of occurrence is higher than a preset ranking among the multiple risk behavior types corresponding to each of the multiple historical messages indicating negative emotions. Risk behavior information mining is performed on the historical messages corresponding to the first risk behavior type to obtain behavioral information corresponding to the first risk behavior type. The behavioral information includes information about the person, location, and time involved in performing the first risk behavior type. Finally, based on the risk behavior information corresponding to the first risk behavior type, an early warning of risky behavior for the target object can be completed. In the above method, the target objects with abnormal emotions are first identified based on the sentiment analysis of the target objects. Then, the most likely types of risky behaviors of the target objects are determined by a preset risk behavior analysis model. Finally, information mining is performed on the historical information corresponding to this type of risky behavior to clearly present various key information related to this type of risky behavior. Finally, risk behavior warnings for the target objects are realized based on the key information related to the risky behavior. This method can perform in-depth risk behavior information mining on objects with abnormal emotions and use the mined risk behavior information to issue warnings, thereby improving the accuracy of risk behavior warnings for the target objects.

[0060] To enable those skilled in the art to better understand the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present application, and not all embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present application.

[0061] Method Implementation Examples

[0062] See Figure 1 The figure is a flowchart illustrating a method for early warning of risky behavior of a target object provided in an embodiment of this application, specifically including the following steps:

[0063] S101: Obtain multiple historical information items published by the target object within a preset time period; the historical information refers to the historical information published by the target object on the network platform.

[0064] First, a web crawler algorithm is used to obtain all historical information posted by the target user on online platforms within a preset time period. This historical information can include text messages such as posts and comments on social media, as well as interactive behaviors on social media. The target user can be multiple social media users within a specific region, or social media users who have commented and posted about a specific trending event. The target user is not limited to a single social media user or online user; it can be a specific object within a specific event or region. The basic information of the target user can also be determined based on a specific ID on social media. Then, the target user's posting history on the online platform is obtained. The selection criteria for the target user and the preset time period can be set by the user.

[0065] S102: Based on the preset emotion analysis model and the multiple historical information, determine whether the target object has an abnormal emotional state; among the multiple historical information published by the object with an abnormal emotional state, there are multiple historical information indicating negative emotions.

[0066] After obtaining multiple historical messages posted by the target subject within a preset time period, a pre-defined sentiment analysis model is used to analyze these messages and determine if the target subject is experiencing an abnormal emotional state. Subjects with abnormal emotional states typically post multiple messages expressing negative emotions. By analyzing each message using the pre-defined sentiment analysis model, the specific emotional category expressed can be determined. These categories can include happiness, fear, anger, anxiety, and numbness, among others. These diverse emotions can be broadly categorized into negative and positive emotions. To ensure efficiency and accuracy, the sentiment analysis process is further refined by defining the message's emotional category within a broader sense, classifying it as either negative or positive. Each message corresponds to a specific emotional expression, which can be either negative or positive. The process of determining whether a target subject is experiencing an abnormal emotional state based on multiple historical messages involves the following three steps:

[0067] Step 1: Based on the preset sentiment analysis model, perform sentiment analysis on the multiple historical information entries to obtain the sentiment category represented by each of the multiple historical information entries; the sentiment category includes at least: positive sentiment and negative sentiment;

[0068] Step 2: Determine the proportion of the historical information representing negative emotions among the multiple pieces of historical information;

[0069] Step 3: If the proportion of historical information representing negative emotions in the multiple pieces of historical information exceeds a preset threshold, then the target object is determined to be in an abnormal emotional state.

[0070] In determining whether a target is experiencing emotional abnormalities, the proportion of negative emotional messages among multiple historical posts published by the target is used for assessment. If the proportion of negative emotional messages exceeds a preset threshold, the target is considered to be in an abnormal emotional state. This method can quickly and accurately determine whether a target is experiencing an abnormal emotional state based on a preset emotion analysis model and a preset threshold. Specifically, the preset emotion analysis model can be trained as follows:

[0071] First, a portion of historical information from multiple objects obtained from historical time periods is randomly selected as training and testing data. Based on the emotion categories described above, such as happiness, anger, anxiety, etc., these historical information are manually labeled one by one. After the labeled information is converted into data that meets the requirements of the preset emotion analysis model format, it is input into the emotion analysis model for training. At the same time, a portion of historical information is used to test and adjust the output results to further improve the model's generalization ability.

[0072] In this embodiment, training and testing can be performed using k-fold cross-validation. All input data is uniformly divided into k parts, and then training and validation are performed k times. In each iteration, one part is selected as the validation set, and the rest are used as the training set. Specifically, the overall data is first split, and then in each round, different parts are selected as the test set, with the rest used as the training set. This process of training and testing is repeated, and finally, the average result of the k tests is obtained, which serves as the final performance metric of the model.

[0073] Furthermore, to ensure the model adapts to various new data, it can be updated and adjusted regularly. This includes collecting new data, manually labeling it, and adding these new training samples to the existing training data to supplement the model's training. Simultaneously, the model's performance can be evaluated and corrected based on newly collected test data to ensure that its sentiment analysis capabilities remain at their optimal level.

[0074] In this embodiment, the preset sentiment analysis model adopts the BERT model or the LSTM model. This embodiment does not specifically limit the type of model used for training.

[0075] S103: If it is determined that the target object's emotional state is abnormal, then risk behavior analysis is performed on the multiple historical information representing negative emotions according to the preset risk behavior analysis model to obtain multiple risk behavior types corresponding to each of the multiple historical information representing negative emotions and a first risk behavior type; the first risk behavior type includes: among the multiple risk behavior types corresponding to each of the multiple historical information representing negative emotions, the risk behavior type whose frequency of occurrence is higher than the preset ranking; the preset risk behavior analysis model is trained based on a deep learning model.

[0076] When an abnormal emotion is identified in a target individual, a risk behavior analysis model is used to analyze their historical information expressing negative emotions. This analysis yields a unique risk behavior type and a primary risk behavior type for each instance of negative emotion. The primary risk behavior type is the risk behavior model whose frequency of occurrence among multiple negative emotion-expressing risk behavior types ranks higher than a preset ranking. This preset ranking can be set manually. For example, when the preset ranking is 3, the top three risk behavior types are selected as the primary risk behavior type based on their frequency of occurrence.

[0077] The purpose of obtaining the first risk behavior type is to determine the most likely risk behavior type of the target object within a preset time period. By pre-determining the most likely risk behavior type of the target object within a certain period, the prediction range of the target object's risk behavior can be effectively narrowed, limiting the most likely risk behavior type of the target object, thereby providing accurate data for subsequent steps of risk behavior information mining. The risk behavior type can be pre-defined. Specifically, by using a preset risk behavior analysis model and multiple historical pieces of information representing negative emotions, the process of obtaining the first risk behavior type can be achieved through the following two steps:

[0078] Step 1: Based on the preset risk behavior analysis model, perform event analysis on the multiple historical information entries representing negative emotions to determine the event type corresponding to each of the multiple historical information entries representing negative emotions; the event type represents the type of event that leads the target object to exhibit risky behavior;

[0079] Step 2: Using the preset risk behavior analysis model, the multiple historical information entries representing negative emotions, and the corresponding event types, determine the risk behavior types corresponding to each of the multiple historical information entries representing negative emotions, as well as the first risk behavior type.

[0080] Each piece of historical information expressing negative emotion corresponds to a specific risk behavior type, which in turn includes a specific time frame. Therefore, before determining the risk behavior type corresponding to historical information, we first determine the event type based on a pre-defined risk behavior model, such as social events, family events, personal events, etc. After determining the event type, we then determine the corresponding risk behavior type based on the event type.

[0081] By identifying the event types corresponding to historical information expressing negative emotions, and then determining specific risk behavior types based on those event types, the efficiency of acquiring risk behavior types can be effectively improved. After obtaining the risk behavior type corresponding to each historical information expressing negative emotions, the frequency of occurrence of each risk behavior type is obtained and ranked. The risk behavior type ranked higher than a preset ranking is identified as the first risk behavior type. This yields the most likely risk behavior type for the target object within a preset timeframe, i.e., the first risk behavior type.

[0082] Specifically, the training process for the pre-defined risk behavior analysis model can be obtained through the following methods:

[0083] First, historical information on all expressions of negative emotions over a past period is obtained. A portion of this information is randomly selected as training and testing data. Based on the aforementioned risk behavior categories, this information is manually labeled. The labeled information is then converted into data that meets the format requirements of the risk behavior analysis model and fed into the risk behavior early warning model for training. At the same time, the model is validated and adjusted using test data to further improve its generalization ability.

[0084] In this embodiment, training and testing can be performed using k-fold cross-validation. All data is evenly divided into k parts, and then training and validation are performed k times. In each iteration, one part is selected as the validation set, and the rest are used as the training set. The overall data is split, and in each round, a different part is selected as the test set, and the rest are used as the training set. Training and testing are repeated cyclically, and finally, the average result of the k tests is obtained, which is used as the final performance metric of the model.

[0085] Furthermore, to ensure the model adapts to various new data, it can be updated and adjusted regularly. This includes collecting new data, manually labeling it, and adding these new training samples to the existing training data to supplement the model's training. Simultaneously, the model's performance can be evaluated and corrected based on newly collected test data to ensure that its predictive ability remains at its optimal level.

[0086] Taking the TextCNN (Text Convolutional Neural Network) model as an example, the steps to train it to obtain the preset risk behavior analysis model are as follows:

[0087] Step 1: Map each word in the historical information to a low-dimensional space using word embedding methods to obtain the word vector of each word, thereby converting the text in each piece of historical information into a two-dimensional matrix.

[0088] Step 2: Perform convolution operations on the above two-dimensional matrix using multiple convolution kernels of different sizes to obtain the feature map corresponding to each two-dimensional matrix. Each feature map represents a local feature in each historical information text.

[0089] Step 3: Perform max pooling on each feature map to obtain its corresponding scalar. The scalar corresponding to each feature map is used to represent the most important feature value in the feature map, thereby reducing the feature dimension and computational complexity.

[0090] Step 4: Concatenate all pooled feature values ​​into a long vector, which serves as the global feature representation for each piece of historical information.

[0091] Step 5: To prevent overfitting, some feature values ​​are randomly discarded between the splicing layer and the fully connected layer to increase the model's generalization ability.

[0092] Step 6: Input the concatenated feature vector into a fully connected layer, and let the SoftMax function output the probability of each class to perform a multi-class classification task.

[0093] S104: Based on a preset information mining algorithm, perform risk behavior information mining on the historical information corresponding to the first risk behavior type to obtain risk behavior information corresponding to the first risk behavior type.

[0094] After obtaining the most likely type of risky behavior of the target object within a preset time period, i.e., the first type of risky behavior, the historical information published by the target object corresponding to the first type of risky behavior is reconfirmed, and risky behavior information is mined based on a preset information mining algorithm. This process mines important information such as people, time, location, and keywords that may be involved when the first type of risky behavior occurs, thus clearly presenting various key information related to the risky behavior of the target object. In the subsequent risky behavior warning process, this behavioral information related to the trend behavior can be used to issue corresponding risky behavior warnings.

[0095] In this embodiment, the preset information mining algorithm can employ named entity recognition, word frequency statistics, and inverse document keyword extraction algorithms to mine behavioral information that may be involved in the first risk behavior type.

[0096] Named entity recognition (NER) is a natural language processing technique used to identify and classify named entities with specific meanings from text, such as names of people, places, and organizations. NER typically trains a model to identify these entities and classifies them into predefined categories. Word frequency statistics refers to the frequency with which a word appears in a document or corpus; a higher word frequency indicates that the word appears more frequently in the text.

[0097] Inverse document frequency (IVF) keyword extraction is an algorithm used to extract keywords from a collection of documents. This algorithm determines the importance of keywords by calculating the inverse document frequency (IVF) of a word within the entire collection. A higher IVF value indicates that the word appears less frequently in the collection and has higher distinctiveness.

[0098] By combining named entity recognition, word frequency statistics, and inverse document frequency (IVF) keyword extraction algorithms, we can extract and rank keywords. By calculating the word frequency (Word Frequency) of each word in the text and its IVF within the entire document set, and then multiplying the Word Frequency by the IVF to obtain a weight value, we can rank the keywords based on these weight values. This yields keywords closely related to risky behavior types, which can be used to describe information such as people, locations, and times associated with the primary risky behavior type.

[0099] S105: Based on the risk behavior information corresponding to the first risk behavior type, issue a risk behavior warning.

[0100] Finally, based on the risk behavior information corresponding to the first type of risk behavior, a risk behavior warning for the target can be completed. This risk behavior warning process can be accomplished through the following two steps:

[0101] Step 1: Determine the warning level corresponding to the risk behavior information based on the risk behavior information corresponding to the first risk behavior type;

[0102] Step 2: Generate a risk behavior warning signal for the target object based on the warning level and the risk behavior information.

[0103] After obtaining the risk behavior information of the first risk behavior type, the corresponding warning level can be determined based on the specific risk behavior information. The determination of the warning level can be decided based on the clarity of the risk behavior information. The clearer and more specific the risk behavior information related to the first risk behavior type is, the higher the corresponding warning level will be. It can also be determined based on the pre-set correspondence between various risk behavior information and warning levels. For example, when the time difference between the occurrence time of the risk behavior indicated in the risk behavior information and the current time is less than the pre-set time interval, the warning level is confirmed as the dangerous level at this time. The specific warning level can be a specific value, or can be described by adjectives such as dangerous, urgent, weak, etc. This embodiment does not specifically limit the manifestation form of the specific warning level.

[0104] Finally, according to the warning level and the relevant risk behavior information, a risk behavior warning signal for the target object is generated, and the risk behavior warning for the target object can be completed. In the risk behavior warning signal, it includes the risk behavior information related to the target object and the warning level. Through the risk behavior warning signal of the target object, the risk behavior warning for the target object can be accurately achieved.

[0105] As an optional implementation manner, before step S102, the following two steps are further included:

[0106] Step 1: Perform data preprocessing on the multiple historical information to obtain multiple processed historical information; the multiple processed historical information are all stored in a pre-set format;

[0107] Step 2: Perform vector conversion on the multiple processed historical information to obtain a word vector space corresponding to each of the multiple processed historical information.

[0108] Before judging whether the emotional state of the target object is abnormal based on the pre-set emotional analysis model and historical information, the historical information needs to be converted into a format that meets the requirements of the pre-set emotional analysis model. First, it is necessary to perform data preprocessing on the multiple historical information. The process of data preprocessing includes cleaning, word segmentation, removing stop words, etc. By performing data preprocessing on the multiple historical information, the data accuracy and applicability of the historical information can be improved. Among them, data cleaning is mainly to delete duplicate data, process missing values, and unify the data format, etc. Word segmentation is mainly to identify and cut continuous text into independent words and tags. Removing stop words is mainly to remove some common words in the text that do not contribute much to the theme or emotion of the text, such as "de", "he", "shi", etc.

[0109] After preprocessing the historical information, each preprocessed historical information is transformed into a corresponding word vector space. In this embodiment, the Word2vec model is used to convert each preprocessed historical information into a word vector space and save it for subsequent processing. Specifically, the Continuous Bag of Words (CBOW) or Skip-gram model of Word2vec can be selected.

[0110] The CBOW model predicts the target word from context words. Given a target word and its context, the goal of the CBOW model is to maximize the following log-likelihood function:

[0111]

[0112] Among them, w t It is the target word, w t-n ,…w t-1 ,w t+1 ,…,w t+n It is its context word, n is the size of the context window, and T is the total number of words in the training set.

[0113] Skip-gram models predict context words from a target word. Given a target word and its context, the goal of a skip-gram model is to maximize the following log-likelihood function:

[0114]

[0115] Among them, w t It is the target word, w t+j It is its context word, n is the size of the context window, and T is the total number of words in the training set.

[0116] By performing vector transformation on multiple historical data entries after data preprocessing, we obtain their respective word vector spaces, enabling the position of a word in the vector space to reflect its semantic meaning. Through word vector representation, we can convert words into a form that machine hardware can understand and capture the semantic similarities and relationships between words.

[0117] This embodiment discloses a method for early warning of risky behavior of a target object. Specifically, the method first acquires multiple historical messages posted by the target object on a network platform within a preset time period. Based on a preset sentiment analysis model and the multiple historical messages, it determines whether the target object's emotional state is abnormal. Among the multiple historical messages posted by users with abnormal emotional states, there are several messages indicating negative emotions. If the target object's emotional state is determined to be abnormal, risk behavior analysis is performed on the multiple historical messages indicating negative emotions according to the preset risk behavior analysis model, resulting in multiple risk behavior types corresponding to each of the multiple historical messages indicating negative emotions, and a first risk behavior type. The first risk behavior type includes risk behavior types whose frequency of occurrence is higher than a preset ranking among the multiple risk behavior types corresponding to each of the multiple historical messages indicating negative emotions. Risk behavior information mining is performed on the historical messages corresponding to the first risk behavior type to obtain behavioral information corresponding to the first risk behavior type. The behavioral information includes information about the person, location, and time involved in performing the first risk behavior type. Finally, based on the risk behavior information corresponding to the first risk behavior type, a risk behavior warning for the target object can be completed. In the above method, the target objects with abnormal emotions are first identified based on the sentiment analysis of the target objects. Then, the most likely types of risky behaviors of the target objects are determined by a preset risk behavior analysis model. Finally, information mining is performed on the historical information corresponding to this type of risky behavior to clearly present various key information related to this type of risky behavior. Finally, risk behavior warnings for the target objects are realized based on the key information related to the risky behavior. This method can perform in-depth risk behavior information mining on objects with abnormal emotions and use the mined risk behavior information to issue warnings, thereby improving the accuracy of risk behavior warnings for the target objects.

[0118] The following describes a risk behavior early warning system for a target object provided by an embodiment of this application. The risk behavior early warning system for a target object described below can be referred to in correspondence with the risk behavior early warning method for a target object described above.

[0119] System Implementation Examples

[0120] Reference Figure 2 The figure is a schematic diagram of the structure of a target object risk behavior early warning system provided in an embodiment of this application, which specifically includes the following modules:

[0121] The acquisition module 100 is used to acquire multiple pieces of historical information published by the target object within a preset time period; the historical information is the historical information published by the target object on the network platform;

[0122] The emotion analysis module 200 is used to determine whether the target object has an abnormal emotional state based on a preset emotion analysis model and the multiple pieces of historical information; among the multiple pieces of historical information published by the object with an abnormal emotional state, there are multiple pieces of historical information indicating negative emotions;

[0123] The risk behavior analysis module 300 is used to analyze the multiple historical information entries representing negative emotions according to a preset risk behavior analysis model when it is determined that the target object's emotional state is abnormal. This analysis yields multiple risk behavior types corresponding to each of the multiple historical information entries representing negative emotions, as well as a first risk behavior type. The first risk behavior type includes risk behavior types whose frequency of occurrence is higher than a preset ranking among the multiple risk behavior types corresponding to each of the multiple historical information entries representing negative emotions. The preset risk behavior analysis model is trained based on a deep learning model.

[0124] The information mining module 400 is used to mine risk behavior information from historical information corresponding to the first risk behavior type based on a preset information mining algorithm, and obtain risk behavior information corresponding to the first risk behavior type; the risk behavior information includes: information on the person, location, and time involved in performing the first risk behavior type;

[0125] The early warning module 500 is used to issue a risk behavior early warning based on the risk behavior information corresponding to the first risk behavior type.

[0126] Optionally, the sentiment analysis module 200 is specifically used for:

[0127] Based on the preset sentiment analysis model, sentiment analysis is performed on the multiple historical information entries to obtain the sentiment category represented by each of the multiple historical information entries; the sentiment category includes at least: positive sentiment and negative sentiment;

[0128] Determine the proportion of the historical information representing negative emotions among the multiple pieces of historical information;

[0129] If the proportion of historical information representing negative emotions in the multiple pieces of historical information exceeds a preset threshold, then the target object is determined to have an abnormal emotional state.

[0130] Optionally, the system further includes: a preprocessing module; the preprocessing module is specifically used for:

[0131] The multiple historical information entries are preprocessed to obtain multiple processed historical information entries; all of the processed historical information entries are stored in a preset format.

[0132] Vector transformation is performed on the multiple processed historical information entries to obtain the word vector spaces corresponding to each of the multiple processed historical information entries.

[0133] Optionally, the risk behavior analysis module 300 is specifically used for:

[0134] Based on the preset risk behavior analysis model, event analysis is performed on the multiple historical information entries representing negative emotions to determine the event type corresponding to each of the multiple historical information entries representing negative emotions; the event type represents the type of event that leads the target object to exhibit risky behavior;

[0135] By using the preset risk behavior analysis model, the multiple historical information entries representing negative emotions, and the corresponding event types, the risk behavior types corresponding to each of the multiple historical information entries representing negative emotions, as well as the first risk behavior type, are determined.

[0136] Optionally, the early warning module 500 is specifically used for:

[0137] Based on the risk behavior information corresponding to the first risk behavior type, determine the warning level corresponding to the risk behavior information;

[0138] Based on the warning level and the risk behavior information, a risk behavior warning signal is generated for the target object.

[0139] Optionally, the risk behavior information includes: information about the people, time, and location involved when the target object performs the first risk behavior type.

[0140] Electronic device examples

[0141] See Figure 3 The figure is a schematic diagram of the structure of an electronic device for early warning of risky behavior of a target object provided in an embodiment of this application, including:

[0142] Memory 11 is used to store computer programs;

[0143] The processor 12 is used to implement the steps of the target object risk behavior early warning method described in any of the above method embodiments when executing the computer program.

[0144] In this embodiment, the device can be an in-vehicle computer, a PC (Personal Computer), or a terminal device such as a smartphone, tablet computer, handheld computer, or portable computer.

[0145] The device may include a memory 11, a processor 12, and a bus 13.

[0146] The memory 11 includes at least one type of readable storage medium, such as flash memory, hard disk, multimedia card, card-type memory (e.g., SD or DX memory), magnetic memory, magnetic disk, optical disk, etc. In some embodiments, the memory 11 may be an internal storage unit of the device, such as the hard disk of the device. In other embodiments, the memory 11 may be an external storage device of the device, such as a plug-in hard disk, SmartMedia Card (SMC), Secure Digital (SD) card, Flash Card, etc. Furthermore, the memory 11 may include both internal and external storage units of the device. The memory 11 can be used not only to store application software and various types of data installed on the device, such as program code for executing fault prediction methods, but also to temporarily store data that has been output or will be output. In some embodiments, the processor 12 may be a Central Processing Unit (CPU).

[0147] In some embodiments, processor 12 may be a central processing unit (CPU), controller, microcontroller, microprocessor or other data processing chip, used to run program code stored in memory 11 or process data, such as program code for executing a fault prediction method.

[0148] This bus 13 can be a Peripheral Component Interconnect (PCI) bus or an Extended Industry Standard Architecture (EISA) bus, etc. This bus can be divided into address bus, data bus, control bus, etc. For ease of representation, Figure 3 The bus is represented by a single thick line, but this does not mean that there is only one bus or one type of bus.

[0149] Furthermore, the device may also include a network interface 14, which may optionally include a wired interface and / or a wireless interface (such as a Wi-Fi interface, a Bluetooth interface, etc.), typically used to establish communication connections between the device and other electronic devices.

[0150] Optionally, the device may further include a user interface 15, which may include a display, an input unit such as a keyboard, and optionally, a standard wired interface or a wireless interface. Optionally, in some embodiments, the display may be an LED display, a liquid crystal display, a touch-sensitive liquid crystal display, or an OLED (Organic Light-Emitting Diode) touchscreen, etc. The display may also be appropriately referred to as a screen or display unit, used to display information processed in the device and to display a visual user interface.

[0151] Figure 3 Only devices with components 11-15 are shown; those skilled in the art will understand that... Figure 3 The structure shown does not constitute a limitation on the device and may include fewer or more components than shown, or combine certain components, or have different component arrangements.

[0152] It should be noted that the various embodiments in this specification are described in a progressive manner, and the same or similar parts between the various embodiments can be referred to mutually. Each embodiment focuses on describing the differences from other embodiments. In particular, for the method apparatus, electronic device, and vehicle, since they are basically similar to the method embodiments, the description is relatively simple, and the relevant parts can be referred to the description of the method embodiments. The method apparatus, electronic device, and vehicle described above are merely illustrative. The units described as separate components may or may not be physically separate, and the components indicated as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of the solution in this embodiment according to actual needs. Those skilled in the art can understand and implement this without creative effort.

[0153] The above description is merely one specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A method for early warning of risky behavior of a target object, characterized in that, include: Retrieve multiple historical messages published by the target object within a preset time period; The historical information refers to the historical information published by the target object on the network platform; Based on a preset emotion analysis model and the multiple pieces of historical information, determine whether the target object has an abnormal emotional state. Among the numerous historical messages posted by individuals exhibiting abnormal emotional states, several expressed negative emotions. If it is determined that the target object's emotional state is abnormal, then risk behavior analysis is performed on the multiple historical information indicating negative emotions according to the preset risk behavior analysis model to obtain multiple risk behavior types corresponding to each of the multiple historical information indicating negative emotions and a first risk behavior type; The first risk behavior type includes: among the multiple risk behavior types corresponding to the multiple historical information representing negative emotions, the risk behavior type whose frequency of occurrence ranks higher than a preset ranking; the preset risk behavior analysis model is trained based on a deep learning model; Based on a preset information mining algorithm, risk behavior information mining is performed on historical information corresponding to the first risk behavior type to obtain risk behavior information corresponding to the first risk behavior type. Based on the risk behavior information corresponding to the first risk behavior type, a risk behavior warning is issued; The step of determining whether the target object has an abnormal emotional state based on a preset emotion analysis model and the multiple pieces of historical information specifically includes: Based on the preset sentiment analysis model, sentiment analysis is performed on the multiple historical information entries to obtain the sentiment category represented by each of the multiple historical information entries; the sentiment category includes at least: positive sentiment and negative sentiment; Determine the proportion of the historical information representing negative emotions among the multiple pieces of historical information; If the proportion of the historical information representing negative emotions in the multiple pieces of historical information exceeds a preset threshold, then the target object is determined to have an abnormal emotional state. Before determining whether the target object has an abnormal emotional state based on the preset emotion analysis model and the multiple historical information, the process also includes: The multiple historical information entries are preprocessed to obtain multiple processed historical information entries; all of the processed historical information entries are stored in a preset format. Vector transformation is performed on the multiple processed historical information entries to obtain the word vector spaces corresponding to each of the multiple processed historical information entries.

2. The method according to claim 1, characterized in that, The step involves performing risk behavior analysis on the multiple historical information entries representing negative emotions based on a preset risk behavior analysis model, resulting in multiple risk behavior types corresponding to each of the multiple historical information entries representing negative emotions, as well as a first risk behavior type. Specifically, this includes: Based on the preset risk behavior analysis model, event analysis is performed on the multiple historical information entries representing negative emotions to determine the event type corresponding to each of the multiple historical information entries representing negative emotions; the event type represents the type of event that leads the target object to exhibit risky behavior; By using the preset risk behavior analysis model, the multiple historical information entries representing negative emotions, and the corresponding event types, the risk behavior types corresponding to each of the multiple historical information entries representing negative emotions, as well as the first risk behavior type, are determined.

3. The method according to claim 1, characterized in that, The step of issuing a risk behavior warning based on the risk behavior information corresponding to the first risk behavior type specifically includes: Based on the risk behavior information corresponding to the first risk behavior type, determine the warning level corresponding to the risk behavior information; Based on the warning level and the risk behavior information, a risk behavior warning signal is generated for the target object.

4. The method according to claim 1, characterized in that, The risk behavior information includes: information about the people, time, and location involved when the target object performs the first risk behavior type.

5. A target object risk behavior early warning system, characterized in that, include: The acquisition module is used to acquire multiple historical messages published by the target object within a preset time period; The historical information refers to the historical information published by the target object on the network platform; The emotion analysis module is used to determine whether the target object has an abnormal emotional state based on a preset emotion analysis model and the multiple historical information. Among the numerous historical messages posted by individuals exhibiting abnormal emotional states, several expressed negative emotions. The risk behavior analysis module is used to analyze the multiple historical information indicating negative emotions according to a preset risk behavior analysis model when it is determined that the target object's emotional state is abnormal, thereby obtaining multiple risk behavior types corresponding to each of the multiple historical information indicating negative emotions and a first risk behavior type. The first risk behavior type includes: among the multiple risk behavior types corresponding to the multiple historical information representing negative emotions, the risk behavior type whose frequency of occurrence ranks higher than a preset ranking; the preset risk behavior analysis model is trained based on a deep learning model; The information mining module is used to mine risk behavior information from historical information corresponding to the first risk behavior type based on a preset information mining algorithm, so as to obtain risk behavior information corresponding to the first risk behavior type; the risk behavior information includes: information on the person, location, and time involved in performing the first risk behavior type. The early warning module is used to issue early warnings for risk behaviors based on the risk behavior information corresponding to the first risk behavior type. The sentiment analysis module is specifically used for: Based on the preset sentiment analysis model, sentiment analysis is performed on the multiple historical information entries to obtain the sentiment category represented by each of the multiple historical information entries; the sentiment category includes at least: positive sentiment and negative sentiment; Determine the proportion of the historical information representing negative emotions among the multiple pieces of historical information; If the proportion of the historical information representing negative emotions in the multiple pieces of historical information exceeds a preset threshold, then the target object is determined to have an abnormal emotional state. The system further includes a preprocessing module; the preprocessing module is specifically used for: The multiple historical information entries are preprocessed to obtain multiple processed historical information entries; all of the processed historical information entries are stored in a preset format. Vector transformation is performed on the multiple processed historical information entries to obtain the word vector spaces corresponding to each of the multiple processed historical information entries.

6. The system according to claim 5, characterized in that, The risk behavior analysis module is specifically used for: Based on the preset risk behavior analysis model, event analysis is performed on the multiple historical information entries representing negative emotions to determine the event type corresponding to each of the multiple historical information entries representing negative emotions; the event type represents the type of event that leads the target object to exhibit risky behavior; By using the preset risk behavior analysis model, the multiple historical information entries representing negative emotions, and the corresponding event types, the risk behavior types corresponding to each of the multiple historical information entries representing negative emotions, as well as the first risk behavior type, are determined.

7. The system according to claim 5, characterized in that, The early warning module is specifically used for: Based on the risk behavior information corresponding to the first risk behavior type, determine the warning level corresponding to the risk behavior information; Based on the warning level and the risk behavior information, a risk behavior warning signal is generated for the target object.

8. The system according to claim 5, characterized in that, The risk behavior information includes: information about the people, time, and location involved when the target object performs the first risk behavior type.

9. An electronic device, characterized in that, The device includes: a processor, a memory, and a system bus; The processor and the memory are connected via the system bus; The memory is used to store one or more programs, the one or more programs including instructions, which, when executed by the processor, cause the processor to perform the target object risk behavior early warning method according to any one of claims 1-4.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When executed by the processor, the program implements the target object risk behavior early warning method as described in any one of claims 1-4.

Images