Social network false user detection method based on neural network stacking integration
Through the neural network stack integration method, combining the prediction results of LSTM and multiple base learners for integration, the problem of poor detection accuracy and generalization performance of fake users in social networks is solved, and higher detection accuracy and adaptability are achieved .
Patent Information
- Application Number
- CN202411136100.7
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-08-19
- Publication Date
- 2024-12-13
AI Technical Summary
The existing social network fake user detection methods have low accuracy and poor generalization performance, making it difficult to effectively detect new fake accounts.
Using a method based on neural network stack integration, user data is preprocessed through the LSTM neural network model, and multiple basic learners (multi-layer perceptron model, convolutional neural network model and recurrent neural network model) are used to predict it. The stacking module integrates the prediction results into input features, and finally makes the final prediction by the meta-learner.
It improves the accuracy and generalization ability of fake user detection, can more comprehensively evaluate user reputation, reduce the risk of overfitting, and adapt to changes in different user groups and social network platforms.
Smart Images
Figure CN119128840A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of Internet account detection, and in particular to a social network fake user detection method based on neural network stacking integration. Background Art
[0002] There are many works related to detecting fake users in social networks. It is difficult for graph-based algorithms to detect new fake accounts because they have a small number of connections with users and may not reflect significant structural differences from new real accounts.
[0003] P. Kondeti et al. used machine learning algorithms to detect fake accounts on social media platforms. They used four machine learning algorithms: support vector machine (SVM), logistic regression (LR), random forest (RF) and K-nearest neighbor (KNN). In addition, two data normalization techniques: Z-Score and Min-Max were used to improve the accuracy of the algorithm. S. Sahoo et al. proposed a machine learning-based framework for real-time detection of fake accounts on Twitter. They identified malicious content and accounts by analyzing different features of accounts, such as the number of tweets, the number of fans, the number of followings, the number of likes, the number of lists, the number of URLs, and the user's location, thereby protecting Twitter user accounts from attacks. However, the accuracy of the solution they proposed was not high, and the generalization performance of the model was not good. Summary of the invention
[0004] The technical problem to be solved by the present invention is to provide a social network fake user detection method based on neural network stacking integration, which can improve the accuracy and generalization ability of overall prediction.
[0005] The technical solution adopted by the present invention to solve the technical problem is: to provide a social network fake user detection method based on neural network stacking integration, comprising the following steps:
[0006] Extract the data set of users to be tested from social networks;
[0007] The LSTM neural network model is used to preprocess the data set of the user to be tested to obtain input data;
[0008] Inputting the input data into a false user detection model to obtain a detection result of the user to be tested;
[0009] Wherein, the fake user detection model includes:
[0010] A plurality of base learners, each base learning period predicts whether the user to be tested is a false user according to the input data to obtain a plurality of prediction results;
[0011] The stacking module is used to stack and integrate multiple prediction results to form input features;
[0012] A meta-learner is used to predict whether the user to be tested is a fake user according to the input features.
[0013] The data set of the user to be tested is extracted from the social network, specifically:
[0014] Construct a GET request to access the personal page of the user to be tested in the social network;
[0015] Parse the returned HTML data, extract the information of the user to be tested, and obtain a data set based on the user information to be tested.
[0016] The data set includes: account ID, gender, nickname, personal introduction and published content.
[0017] The fake user detection model is trained in the following way:
[0018] Extracting datasets of multiple users from social networks;
[0019] Analyze each user's data set and create a label for each user based on the analysis results;
[0020] Using an LSTM neural network model to preprocess the data sets of the multiple users to obtain training data;
[0021] Divide the training data into training set and test set;
[0022] Use the training set to train each base learner to obtain multiple groups of prediction results;
[0023] Each set of prediction results is stacked and integrated to obtain multiple input features;
[0024] Train the meta-learner using multiple input features;
[0025] After the training is completed, the trained base learner and meta learner are tested using the test set. When the error of the test result is less than the threshold, the trained base learner and meta learner are used as false user detection models.
[0026] The data set of each user is analyzed, and a label is created for each user according to the analysis result, specifically:
[0027] Determine whether the user has been identified as a fake user or robot by the social network platform based on the accounts in the dataset;
[0028] If the user has been determined to be a fake user or robot by the social networking platform, the user will be labeled as a fake user;
[0029] Determine whether the user is a certified account of the social networking platform based on the accounts in the data set;
[0030] If the user is a verified account of the social networking platform, the user is labeled as a real user;
[0031] If the user is not determined to be a fake user or robot by the social networking platform and is not a certified account of the social networking platform, the user is evaluated using preset rules and the user is labeled accordingly based on the evaluation results.
[0032] The base learner adopts K-fold cross validation method during training.
[0033] The number of base learners is three, namely: a multi-layer perceptron model, a convolutional neural network model and a recurrent neural network model.
[0034] The technical solution adopted by the present invention to solve the technical problem is: to provide a social network fake user detection device based on neural network stacking integration, comprising:
[0035] The extraction module is used to extract the data set of the user to be tested from the social network;
[0036] A preprocessing module, used to preprocess the data set of the user to be tested using an LSTM neural network model to obtain input data;
[0037] A detection module, used to input the input data into a false user detection model to obtain a detection result of the user to be tested;
[0038] Wherein, the fake user detection model includes:
[0039] A plurality of base learners, each base learning period predicts whether the user to be tested is a false user according to the input data to obtain a plurality of prediction results;
[0040] The stacking module is used to stack and integrate multiple prediction results to form input features;
[0041] A meta-learner is used to predict whether the user to be tested is a fake user according to the input features.
[0042] The technical solution adopted by the present invention to solve its technical problem is: to provide an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the steps of the above-mentioned social network fake user detection method based on neural network stacking integration when executing the computer program.
[0043] The technical solution adopted by the present invention to solve its technical problem is: providing a computer-readable storage medium on which a computer program is stored, and when the computer program is executed by a processor, the steps of the above-mentioned social network fake user detection method based on neural network stacking integration are implemented.
[0044] Beneficial Effects
[0045] Due to the adoption of the above technical solution, the present invention has the following advantages and positive effects compared with the prior art: the present invention stacks and integrates multiple base learners, analyzes user behavior patterns and interaction methods through different levels and perspectives, thereby more comprehensively evaluating the user's reputation and improving the accuracy of user reputation evaluation. The method can process complex social network data and establish deep associations between different data features and patterns. Through stacking and integration, the risk of overfitting can be effectively reduced, and the generalization ability of the model can be improved, thereby better adapting to changes in different user groups and social network platforms. BRIEF DESCRIPTION OF THE DRAWINGS
[0046] Figure 1 is a flow chart of a method for detecting fake users on a social network according to a first embodiment of the present invention;
[0047] Figure 2 is a technical flow chart of a method for detecting fake users on a social network according to a first embodiment of the present invention;
[0048] Figure 3 This is a schematic diagram of data set division in the first embodiment of the present invention;
[0049] Figure 4 4 is a comparison chart of the accuracy performance of the first embodiment of the present invention and other benchmark models. DETAILED DESCRIPTION
[0050] The present invention will be further described below in conjunction with specific embodiments. It should be understood that these embodiments are only used to illustrate the present invention and are not intended to limit the scope of the present invention. In addition, it should be understood that after reading the content taught by the present invention, those skilled in the art can make various changes or modifications to the present invention, and these equivalent forms fall within the scope limited by the appended claims of the application equally.
[0051] The first embodiment of the present invention relates to a method for detecting fake users in a social network based on a neural network stacking integration, such as Figure 1 As shown, the following steps are included:
[0052] Step 1: Extract the data set of the user to be tested from the social network. This step is as follows: first, use the saved cookies to construct a GET request to access the personal page of the user to be tested in the social network (such as Weibo); then, parse the returned HTML data, extract the information of the user to be tested, and obtain a data set based on the user's profile. The data set includes: account ID, gender, nickname, personal introduction, and published content.
[0053] Step 2, preprocessing the data set of the user to be tested using the LSTM neural network model to obtain input data. The LSTM neural network model in this step is a special type of recurrent neural network (RNN), which can effectively solve the long-term dependency problem in ordinary RNN, and control the flow and forgetting of information through its internal gating mechanism. The main purpose of the LSTM neural network model is to process sequence data, and it can effectively capture long-term dependencies. Using the LSTM neural network model for preprocessing in this embodiment can help improve the accuracy of detecting fake users in social platforms, because both user attention time and content release time are important factors to be considered in a social platform.
[0054] The LSTM neural network model consists of one or more LSTM units. Each LSTM unit contains a memory unit (cellstate) and three gates (inputgate, forgetgate, outputgate) to control the flow and forgetting of information. The typical LSTM structure is as follows:
[0055] i t =σ(W i ·[h t-1 ,x t ]+b i )
[0056] f t =σ(W f ·[h t-1 ,x t ]+b f )
[0057] o t =σ(W o ·[h t-1 ,x t ]+b o )
[0058]
[0059]
[0060] ht =o t tanh(C t )
[0061] Among them, x t is the input of the current time step, h t-1 is the hidden state of the previous time step, i t 、f t , o t is the output of the input gate, forget gate, and output gate, C t is the memory cell state at the current time step, h t is the hidden state of the current time step. For each time step, the LSTM receives an input x t and the hidden state h at the previous time step t-1 The input can be a single value, a vector, or a sequence, depending on the specific task and data structure. The output of the LSTM is usually the hidden state h at the current time step. t , or it can be a predicted value or the output of a specific task, such as the category label of text classification or the predicted value of time series.
[0062] Step 3: input the input data into a false user detection model to obtain a detection result of the user to be tested.
[0063] like Figure 2 As shown, the fake user detection model in this step includes:
[0064] Three base learners, each base learning period predicts whether the user to be tested is a false user according to the input data to obtain multiple prediction results;
[0065] The stacking module is used to stack and integrate multiple prediction results to form input features;
[0066] A meta-learner is used to predict whether the user to be tested is a fake user according to the input features.
[0067] Among them, the three base learners are: multi-layer perceptron model (MLP), convolutional neural network model (CNN) and recurrent neural network model (RNN). The reasons for selecting MLP, CNN and RNN as base learners in this implementation are as follows: First, MLP is suitable for processing tasks with structured data and a small number of features. The model is simple, easy to implement and train, but it has certain limitations on the expression of nonlinear relationships in data; CNN focuses on processing images and spatial data, and can effectively capture local features and hierarchical structures; RNN is good at processing sequence data and can consider time relationships. By stacking neural networks with different specialties together, the generalization performance of the model is enhanced and the performance will be improved.
[0068] In this implementation, the fake user detection model can be trained in the following way:
[0069] The same method as step 1 is used to obtain the data sets of multiple users. After obtaining the data sets of these users, it is necessary to analyze the data sets of each user and label each user ID. The label "0" indicates that the user is a fake user, and the label "1" indicates that the user is a real user. In this embodiment, the user is labeled in the following manner: first, it is determined whether the user has been determined as a fake user or robot by the social network platform according to the account in the data set. If the user has been determined as a fake user or robot by the social network platform, the user is labeled as a fake user; then, it is determined whether the user is a certified account of the social network platform according to the account in the data set. If the user is a certified account of the social network platform, the user is labeled as a real user; if the user is not determined as a fake user or robot by the social network platform and is not a certified account of the social network platform, the user is evaluated using the preset rules (see Table 1), and the user is labeled with a corresponding label according to the evaluation result, that is, if the evaluation score exceeds the threshold, the user is labeled as a real user, and if the evaluation score is lower than the threshold, the user is labeled as a fake user.
[0070] Table 1 Fake user judgment rules
[0071]
[0072]
[0073] After the data is labeled, the data set is input into the LSTM neural network model for preprocessing to obtain training data. It is divided into a training set and a test set. The training set is used to train the base learners, while the test set is used to finally evaluate the performance of the stacked set. The training set is passed to all base learners, and then the prediction results of the base learners are used as the input of the meta learner. This division process is shown in Figure 3 middle.
[0074] In this embodiment, the original data is divided into multiple non-overlapping subsets. For example, a K-fold cross validation method can be used. Then, for each subset, one is selected as a validation set and the rest are used as training sets. Then, for each base learner, the complete training set is used for training, and the trained base learner is used to predict the test set.
[0075] For each base learner i, use the training set to train and get the prediction result h i (x). These prediction results serve as the input features of the second-layer stacked ensemble model, which can be expressed as:
[0076] h i (x) = Train(Xtrain ,y train )
[0077] Among them, X train is the feature of the training set, y train is the target value of the training set.
[0078] Finally, the output of each base learner is used as the input feature of the meta-learner, and the true target value is used as its output to train the meta-learner. i (x) are stacked to form the input feature H(x) of the meta-learner, which is used to train the meta-learner g(x). The meta-learner is usually a linear combination or other complex model, which is used to integrate the prediction results of each base learner and finally make an overall prediction. It can be expressed as:
[0079] g(x)=Train(H(x),y train )
[0080] Where H(x)=[h1(x),h2(x),...,h m (x)] is the prediction result vector of all base learners, y train is the target value of the training set.
[0081] In order to evaluate the performance of the fake user detection model in this embodiment in detecting fake users in social networks, its accuracy is tested using a data set, and the results are compared with some commonly used machine learning and neural network models. Bayesian Networks, Decision Trees, Neural Networks, and Random Forests are selected as the benchmark models to be compared.
[0082] In order to reflect the good classification effect, the Multi-Trust feature set proposed by X. Chen et al. is used in the performance evaluation process, which has been proved by the authors to achieve good and stable prediction results on 8 machine learning methods. First, the user features are divided into four groups based on empirical analysis, including profile-based features, behavior-based features, feedback-based features, and link-based features. Among them, profile-based features are extracted from user profile files and contain necessary descriptive information about users and their accounts. These profile files are created when users start to join OSN. Then, behavior-based features are hidden in the actions and operations that users perform in OSN every day, such as posting or forwarding. They can be extracted by analyzing user-generated content and can reveal users' behavioral preferences. In addition, feedback-based features come from the information that users receive from others, such as likes, shares, or comments, which may mean other people's attitudes towards the online activities of the user. Finally, link-based features are extracted by analyzing the relationships between users in OSN, which are hidden in the link structure of the network and are usually not easy to change.
[0083] In order to measure the accuracy performance of these methods, four evaluation indicators were considered, including accuracy, precision, recall and Matthews correlation coefficient (MCC).
[0084] They are calculated as follows:
[0085]
[0086]
[0087]
[0088]
[0089] Among them, TP (True Positive) is a true positive example, which refers to the number of samples predicted by the model as positive and actually positive. FP (False Positive) is a false positive example, which refers to the number of samples predicted by the model as positive but actually negative. TN (True Negative) is a true negative example, which refers to the number of samples predicted by the model as negative and actually negative. FN (False Negative) is a false negative example, which refers to the number of samples predicted by the model as negative but actually positive.
[0090] The learning rate is set to 0.001 and the batch size is set to 32, which are common hyperparameters for evaluating machine learning models. The fake user ratio is set between 5% and 30%. The low fake user ratio is set to simulate the ratio in real situations. Through the 5-fold validation method, each fold is trained for 10 epochs. The comparison of the Accuracy, Precision, Recall and MCC results obtained by the fake user detection model in this embodiment and other four traditional models at different fake user ratios is shown in Figure 4 middle.
[0091] like Figure 4 As shown, the false user detection model in this embodiment has higher accuracy performance compared with the other four traditional models. It is worth emphasizing that the false user detection model in this embodiment obviously has a higher precision rate compared with the other four models. The precision rate measures the proportion of samples that are truly positive among the samples predicted by the classifier as positive. Therefore, when detecting false users on social platforms, more attention is paid to the precision rate to reduce the risk of misjudgment. Although the precision rate of the false user detection model in this embodiment often shows a downward trend as the proportion of false users increases, it is still higher than the other four models.
[0092] It is not difficult to find that the present invention stacks and integrates multiple base learners to analyze user behavior patterns and interaction methods through different levels and perspectives, so as to more comprehensively evaluate the user's reputation and improve the accuracy of user reputation evaluation. In addition, this method can process complex social network data and establish deep associations between different data features and patterns. Through stacking and integration, the risk of overfitting can be effectively reduced and the generalization ability of the model can be improved, so as to better adapt to changes in different user groups and social network platforms.
[0093] In addition, the present invention can also integrate various types of information, such as the user's historical behavior data, social interaction patterns, content preferences, etc., to comprehensively consider the user's overall reputation in the social network. This comprehensive evaluation can not only more accurately identify potential fraudulent behavior and the spread of false information, but also improve user experience and reduce unnecessary interference and negative effects.
[0094] Finally, the present invention has high flexibility and scalability, and can be customized and optimized according to specific social network platforms and user groups. This makes the method not only applicable to large-scale social network platforms, but also can be effectively applied in small social applications and specific user groups, thereby promoting the security, credibility and user satisfaction of social network platforms.
[0095] The second embodiment of the present invention relates to a social network fake user detection device based on neural network stacking integration, comprising:
[0096] The extraction module is used to extract the data set of the user to be tested from the social network;
[0097] A preprocessing module, used to preprocess the data set of the user to be tested using an LSTM neural network model to obtain input data;
[0098] A detection module, used to input the input data into a false user detection model to obtain a detection result of the user to be tested;
[0099] Wherein, the fake user detection model includes:
[0100] A plurality of base learners, each base learning period predicts whether the user to be tested is a false user according to the input data to obtain a plurality of prediction results;
[0101] The stacking module is used to stack and integrate multiple prediction results to form input features;
[0102] A meta-learner is used to predict whether the user to be tested is a fake user according to the input features.
[0103] The extraction module comprises:
[0104] The access unit is used to construct a GET request to access the personal page of the user to be tested in the social network;
[0105] The parsing and extraction unit is used to parse the returned HTML data, extract the information of the user to be tested, and obtain a data set based on the user information to be tested.
[0106] The data set includes: account ID, gender, nickname, personal introduction and published content.
[0107] The fake user detection model is trained in the following way:
[0108] Extracting datasets of multiple users from social networks;
[0109] Analyze each user's data set and create a label for each user based on the analysis results;
[0110] Using an LSTM neural network model to preprocess the data sets of the multiple users to obtain training data;
[0111] Divide the training data into training set and test set;
[0112] Use the training set to train each base learner to obtain multiple groups of prediction results;
[0113] Each set of prediction results is stacked and integrated to obtain multiple input features;
[0114] Train the meta-learner using multiple input features;
[0115] After the training is completed, the trained base learner and meta learner are tested using the test set. When the error of the test result is less than the threshold, the trained base learner and meta learner are used as false user detection models.
[0116] The data set of each user is analyzed, and a label is created for each user according to the analysis result, specifically:
[0117] Determine whether the user has been identified as a fake user or robot by the social network platform based on the accounts in the dataset;
[0118] If the user has been determined to be a fake user or robot by the social networking platform, the user will be labeled as a fake user;
[0119] Determine whether the user is a certified account of the social networking platform based on the accounts in the data set;
[0120] If the user is a verified account of the social networking platform, the user is labeled as a real user;
[0121] If the user is not determined to be a fake user or robot by the social networking platform and is not a certified account of the social networking platform, the user is evaluated using preset rules and the user is labeled accordingly based on the evaluation results.
[0122] The base learner adopts K-fold cross validation method during training.
[0123] The number of base learners is three, namely: a multi-layer perceptron model, a convolutional neural network model and a recurrent neural network model.
[0124] A third embodiment of the present invention relates to an electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the steps of the social network fake user detection method based on neural network stacking integration of the first embodiment when executing the computer program.
[0125] The fourth embodiment of the present invention relates to a computer-readable storage medium having a computer program stored thereon, characterized in that when the computer program is executed by a processor, the steps of the social network fake user detection method based on neural network stacking integration of the first embodiment are implemented.
[0126] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems, or computer program products. Therefore, the present application may adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment in combination with software and hardware. Moreover, the present application may adopt the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage and optical storage, etc.) that contain computer-usable program code.
[0127] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.
[0128] These computer program instructions may also be stored in a computer-readable memory that can direct a computer or other programmable data processing device to work in a specific manner, so that the instructions stored in the computer-readable memory produce a manufactured product including an instruction method, which is implemented in the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.
[0129] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.
[0130] The above is only a specific embodiment of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art can easily think of changes or substitutions within the technical scope disclosed by the present invention, which should be included in the protection scope of the present invention. Therefore, the protection scope of the present invention should be based on the protection scope of the claims.
Claims
1. A method for detecting fake users in social networks based on neural network stacking integration, characterized in that: The following steps are involved: Extract the data set of users to be tested from social networks; The LSTM neural network model is used to preprocess the data set of the user to be tested to obtain input data; Inputting the input data into a false user detection model to obtain a detection result of the user to be tested; Wherein, the fake user detection model includes: A plurality of base learners, each base learning period predicts whether the user to be tested is a false user according to the input data to obtain a plurality of prediction results; The stacking module is used to stack and integrate multiple prediction results to form input features; A meta-learner is used to predict whether the user to be tested is a fake user according to the input features.
2. The method for detecting fake users in social networks based on neural network stacking integration according to claim 1, characterized in that: The data set of the user to be tested is extracted from the social network, specifically: Construct a GET request to access the personal page of the user to be tested in the social network; Parse the returned HTML data, extract the information of the user to be tested, and obtain a data set based on the user information to be tested.
3. The method for detecting fake users in social networks based on neural network stacking integration according to claim 1, characterized in that: The data set includes: account ID, gender, nickname, personal introduction and published content.
4. The method for detecting fake users in social networks based on neural network stacking integration according to claim 1, characterized in that: The fake user detection model is trained in the following way: Extracting datasets of multiple users from social networks; Analyze each user's data set and create a label for each user based on the analysis results; Using an LSTM neural network model to preprocess the data sets of the multiple users to obtain training data; Divide the training data into training set and test set; Use the training set to train each base learner to obtain multiple groups of prediction results; Each set of prediction results is stacked and integrated to obtain multiple input features; Train the meta-learner using multiple input features; After the training is completed, the trained base learner and meta learner are tested using the test set. When the error of the test result is less than the threshold, the trained base learner and meta learner are used as false user detection models.
5. The method for detecting fake users in social networks based on neural network stacking integration according to claim 4, characterized in that: The data set of each user is analyzed, and a label is created for each user according to the analysis result, specifically: Determine whether the user has been identified as a fake user or robot by the social network platform based on the accounts in the dataset; If the user has been determined to be a fake user or robot by the social networking platform, the user will be labeled as a fake user; Determine whether the user is a certified account of the social networking platform based on the accounts in the data set; If the user is a verified account of the social networking platform, the user is labeled as a real user; If the user is not determined to be a fake user or robot by the social networking platform and is not a certified account of the social networking platform, the user is evaluated using preset rules and the user is labeled accordingly based on the evaluation results.
6. The method for detecting fake users in social networks based on neural network stacking integration according to claim 4, characterized in that: The base learner adopts K-fold cross validation method during training.
7. The method for detecting fake users in social networks based on neural network stacking integration according to claim 1, characterized in that: The number of base learners is three, namely: a multi-layer perceptron model, a convolutional neural network model and a recurrent neural network model.
8. A social network fake user detection device based on neural network stacking integration, characterized in that: include: The extraction module is used to extract the data set of the user to be tested from the social network; A preprocessing module, used to preprocess the data set of the user to be tested using an LSTM neural network model to obtain input data; A detection module, used to input the input data into a false user detection model to obtain a detection result of the user to be tested; Wherein, the fake user detection model includes: A plurality of base learners, each base learning period predicts whether the user to be tested is a false user according to the input data to obtain a plurality of prediction results; The stacking module is used to stack and integrate multiple prediction results to form input features; A meta-learner is used to predict whether the user to be tested is a fake user according to the input features.
9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the computer program, the steps of the social network fake user detection method based on neural network stacking integration as described in any one of claims 1-7 are implemented.
10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method for detecting fake users in a social network based on neural network stacking integration as described in any one of claims 1 to 7 are implemented.