User account pushing method and device, electronic equipment and storage medium

Abstract

Embodiments of the present application disclose a user account pushing method and device, electronic equipment and a storage medium; the method comprises: in the case where it is detected that a user account is reported, obtaining operation time sequence data of the user account in a preset time period before the user account is reported; determining a change feature of a specified operation in the preset time period based on the operation time sequence data; calculating a violation probability of the user account based on the change feature; and pushing the user account based on a violation level determined based on the violation probability. According to the change trend of the specified operation of the user account, the probability that the user account is a violation account is determined, and the user account is pushed based on the violation level determined based on the probability that the user account is a violation account. In the case where the artificial auditing capacity is limited, the human resource waste caused by random auditing can be avoided, thereby targeted auditing of the user account can be realized, and the detection rate of violation accounts is improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of computer, in particular to a user account pushing method and device, electronic equipment and storage medium. BACKGROUND

[0002] With the rapid development of computer technology and Internet technology, users can publish and transmit information on the network to achieve fast information exchange. The Internet brings convenience for information exchange, but also may lead to the transmission of bad information, causing bad influence. In order to maintain the safety and stability of the Internet, a corresponding reporting mechanism can be provided to audit the user account publishing information and determine whether it is in violation.

[0003] The user account auditing method is usually manual auditing. With the increasing number of user accounts that need to be audited, due to the limitation of manual auditing capacity, only a part of user accounts can be manually audited, and the extraction process is random and blind, which makes it difficult to achieve targeted auditing, resulting in low detection rate of illegal accounts. SUMMARY

[0004] The embodiments of the present application provide a user account pushing method and device, electronic equipment and storage medium, which can realize targeted auditing of user accounts under the condition of limited manual auditing capacity, and improve the detection rate of illegal accounts.

[0005] The embodiments of the present application provide a user account pushing method, comprising:

[0006] In the case of detecting that a user account is reported, operation time sequence data of the user account in a preset time period before being reported is obtained, the operation time sequence data comprising operation data of the user account performing a specified operation arranged in time sequence;

[0007] Based on the operation time sequence data, a change feature of the specified operation in the preset time period is determined, the change feature representing a change trend of the specified operation in the preset time period;

[0008] Based on the change feature, a violation probability of the user account is calculated;

[0009] Based on the violation probability, the user account is pushed.

[0010] The embodiments of the present application also provide a user account recommendation device, comprising:

[0011] The acquisition module is configured to, in the case of detecting that a user account is reported, acquire operation time sequence data of the user account in a preset time period before being reported, the operation time sequence data comprising operation data of the user account performing a specified operation arranged in time sequence;

[0012] a feature determination module configured to determine a variation feature of the specified operation in the preset time period based on the operation time series data, the variation feature representing a variation trend of the specified operation in the preset time period;

[0013] a probability calculation module configured to calculate a violation probability of the user account based on the variation feature;

[0014] a level determination module configured to determine a violation level of the user account based on the violation probability;

[0015] a pushing module configured to push the user account according to the violation level.

[0016] In some embodiments, the preset time period includes a plurality of sub-time periods, and the feature determination module further includes:

[0017] a feature extraction unit configured to perform feature extraction on the operation time series data to obtain a feature vector sequence, the feature vector sequence including a feature vector corresponding to each sub-time period;

[0018] a fusion unit configured to perform data fusion processing on the feature vectors to obtain the variation feature of the specified operation in the preset time period.

[0019] In some embodiments, the fusion unit is further configured to:

[0020] for each feature vector of a current sub-time period, obtain a forward hidden layer state corresponding to a feature vector of a previous sub-time period, the previous sub-time period being a period before the current sub-time period;

[0021] calculate the variation feature of the specified operation in the preset time period according to the forward hidden layer state and each of the feature vectors.

[0022] In some embodiments, the fusion unit is further configured to:

[0023] perform forward encoding on each feature vector of a current sub-time period to obtain a forward hidden layer state of each sub-time period;

[0024] perform backward encoding on each feature vector of a current sub-time period to obtain a backward hidden layer state of each sub-time period;

[0025] concatenate the forward hidden layer state and the backward hidden layer state of each sub-time period to obtain the variation feature of the specified operation in the preset time period.

[0026] In some embodiments, the probability calculation module further includes:

[0027] an attention weight calculation unit, configured to perform linear transformation on the change feature based on a preset weight parameter to obtain an attention weight corresponding to the change feature;

[0028] an attention feature calculation unit, configured to calculate an attention feature corresponding to the change feature based on the attention weight and the preset weight parameter;

[0029] a probability calculation unit, configured to obtain a violation probability of the user account according to the attention feature.

[0030] In some embodiments, the preset weight parameter includes a first weight parameter and a second weight parameter, and the attention weight calculation unit is further configured to:

[0031] calculate a product of the change feature and the first weight parameter to obtain a first attention weight;

[0032] calculate a product of the change feature and the second weight parameter to obtain a second attention weight;

[0033] obtain the attention weight corresponding to the change feature according to the first attention weight and the second attention weight.

[0034] In some embodiments, the preset weight parameter further includes a third weight parameter, and the attention feature calculation unit is further configured to:

[0035] perform linear transformation on the change feature based on the third weight parameter to obtain an adjusted change feature;

[0036] calculate a product of the attention weight and the adjusted change feature to obtain the attention feature corresponding to the change feature.

[0037] In some embodiments, the probability calculation unit is further configured to:

[0038] obtain a target parameter and a target bias;

[0039] perform weighted calculation on the attention feature using the target parameter to obtain a weighted attention feature;

[0040] calculate a sum of the weighted attention feature and the target bias to obtain a target feature vector;

[0041] perform normalization processing on the target feature vector to obtain the violation probability of the user account.

[0042] In some embodiments, the level determination module is further configured to:

[0043] obtain a number of times that the user account is reported in the preset time period;

[0044] multiply the first preset weight and the number of times to obtain a first parameter value;

[0045] multiply the second preset weight and the violation probability to obtain a second parameter value, the first preset weight being less than the second preset weight;

[0046] sum the first parameter value and the second parameter value to obtain a target parameter value;

[0047] determine a violation level of the user account based on the target parameter value.

[0048] In some embodiments, the level determination module is further configured to:

[0049] if the violation probability is greater than or equal to a first probability, determine the violation level of the user account as a first violation level;

[0050] if the violation probability is less than the first probability, determine a size relationship between the violation probability and a second probability, the second probability being lower than the violation probability indicated by the first probability;

[0051] if the violation probability is greater than or equal to the second probability, determine the violation level of the user account as a second violation level;

[0052] if the violation probability is less than the second probability, determine the violation level of the user account as a third violation level.

[0053] In some embodiments, the specified operation includes a publishing operation of specified information through the user account, and the operation data includes a number of times of publishing the specified information.

[0054] Embodiments of the present application also provide an electronic device, which includes a memory storing a plurality of instructions; and a processor loading the instructions from the memory to perform steps in any of the user account pushing methods provided by embodiments of the present application.

[0055] Embodiments of the present application also provide a computer readable storage medium storing a plurality of instructions, the instructions being suitable for being loaded by a processor to perform steps in any of the user account pushing methods provided by embodiments of the present application.

[0056] Embodiments of the present application also provide a computer program product, which includes computer programs / instructions, the computer programs / instructions being executed by a processor to implement steps in any of the user account pushing methods provided by embodiments of the present application.

[0057] The embodiment of the application can acquire operation time sequence data of a user account in a preset time period before the user account is reported in a case where it is detected that the user account is reported; then, according to the operation time sequence data, a change feature of a specified operation in the preset time period is determined, then a violation probability of the user account is calculated by using the change feature, and further, a violation level of the user account is determined based on the violation probability, and finally, the user account is pushed based on the violation level. The change feature of the specified operation is extracted from the operation time sequence data in the preset time period, which can represent the change trend of the user account in the specified operation, then the probability that the user account is a violation account can be determined according to the change trend, the violation level can be determined according to the probability that the user account is a violation account, and the user account is pushed based on the violation level. In a case where the artificial auditing capacity is limited, the human resource waste caused by random auditing can be avoided, so that the targeted auditing of the user account can be realized, and the detection rate of the violation account is improved. BRIEF DESCRIPTION OF DRAWINGS

[0058] In order to more clearly illustrate the technical solutions in the embodiments of the application, the drawings needed to be used in the embodiment description will be briefly introduced. Obviously, the drawings in the following description are only some embodiments of the application, and other drawings can be obtained by those skilled in the art without creative effort on the basis of these drawings.

[0059] Figure 1a is a scene schematic diagram of a user account pushing method provided by the embodiment of the application;

[0060] Figure 1b is a flow schematic diagram of a user account pushing method provided by the embodiment of the application;

[0061] Figure 1c is a structure schematic diagram of a recurrent neural network provided by the embodiment of the application;

[0062] Figure 1d is an internal structure schematic diagram of an LSTM provided by the embodiment of the application;

[0063] Figure 1e is a flow schematic diagram of calculating a violation probability provided by the embodiment of the application;

[0064] Figure 1f is a flow schematic diagram of pushing a user account provided by the embodiment of the application;

[0065] Figure 2a is a flow schematic diagram of a user account pushing method provided by another embodiment of the application;

[0066] Figure 2b is a whole architecture diagram of a user pushing method provided by the embodiment of the application;

[0067] Figure 3 is a structural schematic diagram of a user account pushing device provided by an embodiment of the present application.

[0068] Figure 4 is a structural schematic diagram of an electronic device provided by an embodiment of the present application. DETAILED DESCRIPTION

[0069] The technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments of the present application. Based on the embodiments in the present application, all other embodiments obtained by a person skilled in the art without creative work fall within the scope of protection of the present application.

[0070] The embodiments of the present application provide a user account pushing method and device, an electronic device and a storage medium.

[0071] The user account pushing device can be integrated in an electronic device, which can be a terminal, a server or the like. The terminal can be a mobile phone, a tablet computer, a smart Bluetooth device, a notebook computer or a personal computer (PC) or the like. The server can be a single server or a server cluster composed of multiple servers.

[0072] In some embodiments, the user account pushing device can also be integrated in multiple electronic devices, for example, the user account pushing device can be integrated in multiple servers, and the multiple servers can implement the user account pushing method of the present application.

[0073] In some embodiments, the server can also be implemented in the form of a terminal.

[0074] For example, as shown in Figure 1a , a schematic diagram of an application scenario of the user account pushing method provided by an embodiment of the present application is shown.

[0075] As shown in Figure 1a , the user terminal 101, the server 102 and the quality inspection platform 103 are located in a wireless network or a wired network, the user terminal 101 and the server 102 perform data interaction, and the server 102 can interact with the quality inspection platform 103. The server 102 can be a single server or a server cluster, and can be a local server or a cloud server. The quality inspection platform 103 refers to a platform for manually inspecting user accounts.

[0076] The user terminal 101 can be installed with a corresponding application to realize information exchange through information publishing on the network. When a user encounters a user publishing undesirable information during information exchange, the user can report the user account of the user publishing the undesirable information. For example, the user terminal 101 can mark the user account of the user publishing the undesirable information as a rule violation account, and send the identifier corresponding to the user account to the server 102. Thus, the server 102 can obtain the user account marked as a rule violation account by the user terminal.

[0077] After obtaining the user account marked as a rule violation account by the user terminal, the server 102 detects that the user account is reported by other users, and obtains operation time sequence data of the user account in a preset time period before being reported, for example, operation data generated by the user account when performing a specified operation in the preset time period, and arranges the operation data in chronological order.

[0078] The specified operation can refer to the operation of publishing specified information by the user account, and the specified information can refer to pictures, texts, network links, audios, videos, etc. The operation data can refer to operation data generated by the user corresponding to the user account when performing a specified operation in the application, for example, the number of times of publishing specified information, etc. The server 102 can obtain the operation time sequence data by user reporting or through SDK components in each application in the user terminal.

[0079] After obtaining the operation time sequence data of the user account in the preset time period before being reported, the server 102 can determine a change feature of the specified operation in the preset time period based on the operation time sequence data, wherein the change feature is used to indicate the change trend of the specified operation in the preset time period. Based on the change feature, the server 102 calculates a rule violation probability of the user account, and determines a rule violation level of the user account based on the rule violation probability. Finally, the server 102 pushes the user account to the quality inspection platform 103 based on the rule violation level, so as to manually review the user account to determine whether the user account is a rule violation account. The manual review can refer to manually checking the complained content and other account information of the user account to verify whether the user account is a rule violation account.

[0080] In some embodiments, if the user account is determined to be a rule violation account, the quality inspection platform 103 can feed back the review result to the server 102, and the server 102 feeds back the review result to the user terminal 101 marking the user account as a rule violation account.

[0081] It should be noted that in the specific embodiments of the present application, the data related to the user, such as user account information, operation data, and other data related to the user, such as the length of time the target user uses a certain application, the number of times of publishing information, and the like, when the embodiments of the present application apply these data to specific products or technologies, the user's permission or consent needs to be obtained, and the collection, use and processing of related data needs to comply with relevant laws, regulations and standards of the country and region, which will be described in detail below.

[0082] Artificial intelligence (AI) is a technology that uses digital computers to simulate human perception of the environment, acquire knowledge, and use knowledge, which can enable machines to have functions similar to human perception, reasoning, and decision-making. Artificial intelligence basic technologies generally include technologies such as sensors, special artificial intelligence chips, cloud computing, distributed storage, big data processing technology, operation / interaction system, mechatronics, etc. Artificial intelligence software technology mainly includes computer vision technology, speech processing technology, natural language processing technology, and machine learning / deep learning, autonomous driving, intelligent transportation, etc.

[0083] Machine learning (ML) is a multi-disciplinary subject that involves probability theory, statistics, approximation theory, convex analysis, algorithm complexity theory, etc. It is a subject that studies how computers simulate or implement human learning characteristics to acquire new knowledge or skills, and reorganize existing knowledge structure to continuously improve their performance. Machine learning is the core of artificial intelligence and the fundamental approach to enabling computers to have intelligence, and its applications are widespread in various fields of artificial intelligence. Machine learning and deep learning generally include artificial neural networks, belief networks, reinforcement learning, transfer learning, inductive learning, and teaching learning.

[0084] With the research and progress of artificial intelligence technology, artificial intelligence technology is being researched and applied in many fields, such as common smart home, smart wearable devices, virtual assistants, smart speakers, smart marketing, unmanned vehicles, autonomous vehicles, drones, robots, smart medical care, smart customer service, Internet of Vehicles, autonomous driving, intelligent transportation, etc. It is believed that with the development of technology, artificial intelligence technology will be applied in more fields and play an increasingly important role.

[0085] In the present embodiment, a user account pushing method based on artificial intelligence is provided, as shown in Figure 1b The specific process of the user account pushing method can be as follows:

[0086] S110, in the case where it is detected that a user account is reported, operation time sequence data of the user account in a preset time period before being reported is acquired.

[0087] A user account refers to an account created by a user for using an application or a function, and is used to identify the identity of the user in the application. When a user account is reported, the user account becomes a reported user account, which refers to an account that has performed a violation operation and has been reported by other users, for example, an account that has been reported for publishing bad information, or an account that has been reported for illegal activities, such as an account that has been reported for publishing false information using a user account, or an account that has been reported for illegal activities of obtaining money from others using a user account.

[0088] A user can publish information through a user account, and can also view information published by other user accounts to achieve information exchange. In some embodiments, the application can provide a function of marking a violation account, so that a user can use the function of marking a violation account to report a user account. For example, during the process of information exchange through a user account, a user discovers that a user account a publishes bad information, and then the user can mark the user account a as a violation account through the function of marking a violation account provided by the application, so as to report the user account a and send the user account a to the server.

[0089] In some embodiments, after a user account is marked as a violation account, a preset violation mark can be generated accordingly, and the preset violation mark and the user account are sent to the server together, so that when the server receives the user account carrying the preset violation mark, it can be considered that the user account is reported.

[0090] When the server detects that a user account is reported, the operation time sequence data of the user account in a preset time period before being reported can be obtained, wherein the operation time sequence data can include operation data of the user account performing a specified operation, and the operation times are arranged in chronological order. In some embodiments, when the user account performs a specified operation, the server can obtain the operation data of the specified operation and store it, for example, the time when the operation data is generated and the corresponding operation data can be stored in chronological order. For example, the user account performs a specified operation at 12 o'clock to generate operation data 1, and the user account performs a specified operation at 14 o'clock to generate operation data 2, then the server can store the operation data and time in chronological order, 12 o'clock-operation data 1, 14 o'clock-operation data 2.

[0091] When it is detected that a user account is reported, the operation data of a preset time period before being reported can be extracted from these data according to time, so that the operation time sequence data can be obtained.

[0092] The preset time period can be set to any length of time, for example, 15 days, 7 days, etc., and can be set according to actual needs, which is not limited here.

[0093] The specified operation can refer to a pre-set user operation, which can refer to an operation frequently performed by the violation account. For example, the violation account can add a large number of friends in a short time, the login location of the user account frequently jumps, and at the same time sends messages to different user accounts, etc., so that the specified operation can include the operation of logging in the account, the publishing operation of the specified information, etc., wherein the specified information can refer to text, pictures, videos, audio, network links, etc. The operation data can refer to data generated when the specified operation is performed, or refer to the number of times, time, etc. of performing the specified operation, for example, the account login location and account login time, the number of times of publishing specified information, etc.

[0094] In some embodiments, when a user reports a user account, the user can also send the reason for reporting the user account, i.e., the marking information, to the server, and the marking information can be in the form of text, image, audio, network link, etc. After the server obtains the marking information, it can first audit the user account based on the marking information, for example, through text recognition, image recognition, etc. If the user account cannot be audited based on the marking information, the operation time sequence data of the user account in the preset time period before being reported is obtained.

[0095] It should be noted that the operation data described above is obtained with the authorization and permission of the user, for example, when the operation data needs to be obtained, a prompt can be given on the user side, and when the user's confirmation operation is received, it is considered that the user has authorized, and then the operation data is obtained.

[0096] S120, based on the operation time sequence data, determining the change feature of the specified operation in the preset time period, the change feature representing the change trend of the specified operation in the preset time period.

[0097] Based on the obtained operation time sequence data of the user account in the preset time period, i.e., the operation data of the user account performing the specified operation in the preset time period, the operation time sequence data can be processed to obtain the change feature of the specified operation in the preset time period, wherein the change feature is used to represent the change trend of the specified operation in the preset time period.

[0098] For example, the preset time period is 2 days, if the operation data is counted every day, the user account publishes specified information 3 times the day before, and the user account publishes specified information 6 times yesterday, it can be seen that the specified operation in the preset time period presents a growth trend.

[0099] In some embodiments, based on the operation time series data, the change feature of the specified operation in the preset time period can be determined by dividing the preset time period into multiple sub-time periods, then performing feature extraction on the operation time series data to obtain a feature vector sequence; and performing data fusion processing on the feature vector to obtain the change feature of the specified operation in the preset time period.

[0100] As an implementation, the sub-time period can be a pre-set time length, such as 1 day, 1 hour, etc. If the operation time series data in 7 days is obtained, according to the time length of the sub-time period being 1 day, the operation time series data can be divided into 7 parts. When performing feature extraction on the operation time series data to obtain a feature vector sequence, the operation time series data in each sub-time period can be extracted to obtain a feature vector corresponding to each sub-time period, and then all the feature vectors are arranged in time sequence to obtain the feature vector sequence.

[0101] In some embodiments, when performing feature extraction on the operation time series data to obtain a feature vector sequence, the operation time series data can be input into a fully connected (FC) layer, the operation time series data is encoded by the FC layer, and feature dimension reduction and selection are performed to obtain the feature vector sequence.

[0102] After performing feature extraction on the operation time series data to obtain a feature vector sequence, data fusion processing can be performed on the feature vector to obtain the change feature of the specified operation in the preset time period.

[0103] In some embodiments, when performing data fusion processing on the feature vector, a recurrent neural network (RNN) can be used to perform data fusion on the feature vector. The recurrent neural network is a recursive neural network with sequence data as input, which is recursively connected in the evolution direction of the sequence, and all the recurrent units are connected in a chain. The recurrent neural network can generally depict the relationship between the current output of a sequence and the previous information. In order to achieve this effect, the nodes between the hidden layers of the recurrent neural network are connected, and the input of the hidden layer includes not only the output of the output layer, but also the output of the previous hidden layer.

[0104] As an implementation, when performing data fusion processing on the feature vector, for the feature vector of each current sub-time period, the feature vector corresponding to the forward hidden layer state of the previous sub-time period can be obtained, wherein the previous sub-time period is a period before the current sub-time period. The forward hidden layer state is a hidden layer state considering the influence of the hidden layer state of the previous sub-time period on the current hidden layer state. For example, the sub-time period T1 is April 20, the sub-time period T2 is April 21, and the current sub-time period is T2 period, then the previous sub-time period is T1 period.

[0105] For example, you can refer to Figure 1c The diagram shows a schematic of the structure of a recurrent neural network. Figure 1c The recurrent neural network shown on the left is the basic structure of the RNN model, where x t Let A represent the input of the RNN model, and let h represent the hidden layer of the RNN model. t The hidden state of the RNN model is represented by W1, the weights between the input layer and the hidden layer are represented by W1, and the weights of the previous hidden layer value used as the current input are represented by U1. To facilitate understanding of the RNN model's operational mechanism, the RNN model on the left can be expanded over time to obtain... Figure 1c The model structure shown on the right is shown in the middle.

[0106] The structure of the RNN model after time expansion shows that the RNN model has an input at each time step, and then provides an output h based on the current hidden state of the RNN model. t In an RNN model, each hidden state A is determined by the hidden state of the previous time step and the current input. For the feature vector sequence, which includes the feature vector corresponding to each sub-time period, inputting this feature vector sequence into the RNN model allows us to obtain the forward hidden state corresponding to the feature vector of the sub-time period preceding the current sub-time period for each current sub-time period's feature vector. Then, based on the forward hidden state and the feature vector, we obtain the changing features.

[0107] For example, the feature vector sequence is a sequence of length T {x1, x2, ... x}. T}, where x i =(x i,1 ,x i,2 ,…,x i,n Let be an n-dimensional vector, and its hidden state can be calculated using the following formula:

[0108] h t =f(U1h t-1 +W1x t +b1);

[0109] Among them, h t h represents the current hidden state. t-1 b0 represents the hidden state at the previous time step, and b1 represents the bias value. During the calculation process, h0 is usually set to a vector of all zeros.

[0110] In some implementations, data fusion processing of the feature vectors can be performed by forward encoding the feature vectors of each current sub-time period to obtain the forward hidden state of each sub-time period; backward encoding the feature vectors of each current sub-time period to obtain the backward hidden state of each sub-time period; and concatenating the forward and backward hidden states of each sub-time period to obtain the change features of the specified operation over a preset time period. The backward hidden state is a hidden state obtained by considering the influence of the hidden states of sub-time periods following the current sub-time period on the current hidden state.

[0111] As one implementation method, a Long Short-Term Memory (LSTM) neural network can be used to perform data fusion processing on the feature vectors. The LSTM model is a special type of recurrent neural network. LSTM allows useful information to be passed to subsequent time steps by forgetting existing information and remembering new information, while discarding useless information. At each time step, it outputs the hidden state. Forgetting, remembering, and output are controlled by a forget gate, an input gate, and an output gate, calculated using the hidden state from the previous time step and the current input.

[0112] For details, please refer to [link / reference]. Figure 1d The diagram shows the internal structure of an LSTM, where the forget gate is the output h of the previous unit. t-1 and the input x of this unit t The input sigmoid function, i.e. σ1 in the figure, is C. t-1 Each item in the array generates a value in the range [0,1] to control the degree to which the state of the previous unit is forgotten.

[0113] The input gate σ2 and a tanh function (tanh 1 in the diagram) work together to control what new information is added. The tanh function generates a new candidate vector, and the input gate generates a value in [0,1] for each item in the candidate vector, controlling how much new information is added.

[0114] The output gate controls how much of the current cell state is filtered out. First, the cell state is activated, and then the output gate generates a value in the range [0,1] for each item, controlling the degree to which the control state is filtered. Figure 1d Using tanh 2 and σ3, the hidden state h is calculated. t And the current state C t .

[0115] Before using the LSTM model, it can be trained with a large amount of training data in advance so that the LSTM model can fuse feature vectors to obtain changing features.

[0116] In some embodiments, when encoding the feature vector of each current sub-time period forward, the feature vector sequence can be input into the LSTM in the forward direction to obtain the forward hidden layer state corresponding to each feature vector. When encoding the feature vector backward, the feature vector sequence can be input into the LSTM in the reverse direction to obtain the backward hidden layer state corresponding to each feature vector.

[0117] In some embodiments, the change feature can be directly obtained by using a bidirectional long short-term memory network (Bi-directional Long Short-Term Memory, BiLSTM). The BiLSTM is composed of a forward LSTM and a backward LSTM. The feature vector sequence is input into the BiLSTM. The feature vector sequence is input into the forward LSTM for forward encoding to obtain the forward hidden layer state corresponding to each feature vector. The feature vector sequence is input into the backward LSTM for backward encoding to obtain the backward hidden layer state corresponding to each feature vector.

[0118] After obtaining the forward hidden layer state and the backward hidden layer state corresponding to each feature vector, the forward hidden layer state and the backward hidden layer state can be spliced to obtain the change feature corresponding to each sub-time period. The change features of all sub-time periods are taken as the change feature of the specified operation in the preset time period. For example, the forward hidden layer state is {hL0, hL1, hL2}, the backward hidden layer state is {hR0, hR1, hR2}, and splicing can obtain {[hL0, hR2], [hL1, hR1], [hL2, hR0]} to obtain the change feature.

[0119] In some embodiments, the change feature can be directly input into the FC layer for classification to obtain the violation probability. In order to further improve the accuracy of the violation probability, the mutual influence between each sub-change feature in the change feature can be considered, and different weights can be assigned to each sub-change feature in the change feature to realize more accurate encoding of the change feature. For example, the change feature can be linearly transformed based on a preset weight parameter to obtain an attention weight corresponding to the change feature; an attention feature corresponding to the change feature is calculated based on the attention weight and the preset weight parameter; and the violation probability value of the user account is obtained according to the attention feature.

[0120] After obtaining the change feature, the change feature can represent the change trend of the specified operation in the preset time period, and the violation probability of the user account can be calculated according to the change feature, where the violation probability refers to the probability that the user account is a violation account. The change feature can include a plurality of sub-change features, each of which corresponds to a sub-time period.

[0121] In some embodiments, the change feature can be directly input into the FC layer for classification to obtain the violation probability. In order to further improve the accuracy of the violation probability, the mutual influence between each sub-change feature in the change feature can be considered, and different weights can be assigned to each sub-change feature in the change feature to realize more accurate encoding of the change feature. For example, the change feature can be linearly transformed based on a preset weight parameter to obtain an attention weight corresponding to the change feature; an attention feature corresponding to the change feature is calculated based on the attention weight and the preset weight parameter; and the violation probability value of the user account is obtained according to the attention feature.

[0122] The attention feature corresponding to the change feature can be calculated using an attention mechanism. The attention mechanism is a problem-solving method that simulates human attention. It can quickly filter out high-value information from a large amount of information. Algorithms related to the attention mechanism can include Soft Attention, Hard Attention, static Attention, dynamic Attention, and Self-Attention. Any one or more of the above can be used when calculating the attention feature corresponding to the change feature. The following will be described in detail with Self-Attention as an example.

[0123] Before calculating the attention feature, the attention model can be trained to obtain a preset weight parameter. The preset weight parameter can include W Q , W K , and W V . W Q can be used as the first weight parameter, W K as the second weight parameter, and W V as the third weight parameter.

[0124] The change feature is input into the attention model, which can calculate the attention weight and output the attention feature corresponding to the change feature based on the attention weight. When the change feature is linearly changed based on the preset weight parameter to obtain the attention weight corresponding to the change feature, the product of the change feature and the first weight parameter can be calculated to obtain the first attention weight. The product of the change feature and the second weight parameter is calculated to obtain the second attention weight. The attention weight corresponding to the change feature is obtained according to the first attention weight and the second attention weight.

[0125] For example, if the change feature is denoted as X, the first attention weight is denoted as Q, and the second attention weight is denoted as K, then Q = X * W Q and K = X * W K .

[0126] Then, when calculating the attention weight corresponding to the change feature, the transpose of the first attention weight can be calculated, and the dot product of the second attention weight is calculated. Then, the dot product is divided by the square root of the dimension of the K vector, and then normalized by softmax to obtain the attention weight. The specific calculation process can be referred to the following formula:

[0127]

[0128] where K T represents the transpose of K, and d k is the number of columns of Q and K, i.e., the dimension of the vector.

[0129] After the attention weight is calculated, the attention feature corresponding to the change feature can be calculated based on the attention weight and a preset weight parameter. For example, the change feature can be linearly transformed based on a third weight parameter to obtain an adjusted change feature; and the product of the attention weight and the adjusted change feature is calculated to obtain the attention feature corresponding to the change feature.

[0130] As described in the foregoing example, the adjusted change feature can be represented as: V = X * W V Then, the attention weight calculated in the foregoing and the adjusted change feature are multiplied to retain the position feature.

[0131] Then, the attention weight and the adjusted change feature are multiplied to obtain the attention feature corresponding to the change feature, which can be calculated according to the following formula:

[0132]

[0133] wherein Z represents the attention feature, represents the attention weight, and V represents the adjusted change feature. Through the attention model, when each sub-change feature in the change feature is encoded, the importance score of other sub-change features to the current encoded sub-change feature is considered, and the weight of the change feature with a higher contribution degree is enhanced, and the dimension of the sub-change feature is not changed.

[0134] After the attention feature is calculated, the violation probability of the user account can be obtained based on the attention feature. In some embodiments, the obtained attention feature can be input into a fully connected layer, and the violation probability of the user account is output through the calculation of the fully connected layer. Each node of the fully connected layer is connected to all nodes of the previous layer, and is used to integrate the extracted attention feature to be converted into a probability.

[0135] As an implementation, the target parameter and the target bias can be obtained; the attention feature is weighted calculated using the target parameter to obtain a weighted attention feature; the sum of the weighted attention feature and the target bias is calculated to obtain a target feature vector; and the target feature vector is normalized to obtain the violation probability of the user account.

[0136] The target parameter and the target bias are obtained through training, and the attention feature and the target parameter are multiplied to obtain a weighted attention feature, and then the weighted attention feature and the target bias are added to obtain a target feature vector, and then the target feature vector is normalized to obtain the violation probability. For example, the target feature vector can be calculated according to the following formula:

[0137]

[0138] wherein, represents the violation probability, W2 represents the target parameter, b2 represents the target bias, Z represents the attention feature, and the sigmoid function is also called the Logistic function, and the output thereof is between 0 and 1.

[0139] In order to more clearly describe the process of calculating the violation probability, refer to Figure 1e , a flowchart for calculating the violation probability is shown. In Figure 1e , the operation time series data of the preset time period obtained is input into a feature extraction module to obtain a feature vector sequence, and then the feature vector is input into a data fusion module to obtain a change feature, and then the change feature is input into an attention mechanism module to obtain an attention feature, and then the attention feature is input into a probability calculation module to obtain the violation probability. The feature extraction module can be an FC layer; the data fusion module can be an RNN, an LSTM, a BiLSTM, a GRU, or the like, for example, Figure 1e , the data fusion module shown in is a BiLSTM; the attention mechanism module can be a Soft Attention, a Hard Attention, a static Attention, a dynamic Attention, a Self-Attention, or the like; and the probability calculation module can be an FC layer.

[0140] In some embodiments, the FC layer, the data fusion module, and the attention mechanism module can be referred to as a probability calculation model. Before using the probability calculation model, the model can be trained in advance using training data. The training data can be operation time series data of a user account that has been audited and labeled as a violation account in a preset time period. Data corresponding to the user account whose audit result is a violation account is used as a positive sample, and data corresponding to the user account whose audit result is a normal account is used as a negative sample. Then the model is trained based on the positive sample and the negative sample until the model converges, and the probability calculation model is obtained.

[0141] In some embodiments, other classification or clustering models can also be used to simply determine the violation probability of the user account, for example, linear classification, decision tree, K-Nearest Neighbor (KNN), Support Vector Machine (SVM), or the like.

[0142] S140, determining a violation level of the user account based on the violation probability.

[0143] S150, push the user account according to the violation level. The higher the value of the violation probability, the greater the possibility that the user account is a violation account. Of course, the higher the value of the violation probability, the lower the possibility that the user account is a violation account. The specific setting can be made according to actual needs. In the embodiment of the application, only the case that the higher the value of the violation probability, the greater the possibility that the user account is a violation account is taken as an example for description.

[0144] After obtaining the violation probability corresponding to the user account, the violation level of the user account can be determined according to the violation probability, and the user account can be pushed according to the violation level.

[0145] As an implementation manner, when the violation level of the user account is determined based on the violation probability, it can be determined whether the violation probability is greater than or equal to a first probability. If the violation probability is greater than or equal to the first probability, the violation level of the user account can be determined as a first violation level. If the violation probability is less than the first probability, it can be determined whether the violation probability is greater than or equal to a second probability. If the violation probability is greater than or equal to the second probability, the violation level of the user account can be determined as a second violation level. If the violation probability is less than the second probability, the violation level of the user account can be determined as a third violation level. The first violation level is higher than the second violation level, and the second violation level is higher than the third violation level.

[0146] For example, the violation probability of a certain user account is y, the first probability is y1, and the second probability is y2, where y1>y>y2. Thus, the violation level of the user account can be determined as the second violation level.

[0147] For reference Figure 1f , a flowchart for pushing the user account is shown. As an implementation manner, when the user account is pushed according to the violation level, the user account of the first violation level can be pushed first, then the user account of the second violation level can be pushed, and finally the user account of the third violation level can be pushed. As another implementation manner, only the user account of the first violation level can be pushed.

[0148] In some embodiments, if there are multiple reported user accounts, the violation probability corresponding to each user account can be calculated, and the violation level of the user account can be determined based on the violation probability corresponding to all user accounts. For example, the multiple user accounts can be sorted in descending order of violation probability, and the violation level decreases in the order of sorting.

[0149] In an implementation, when calculating the violation level of the user account, the violation level of the user account can also be determined based on the number of times the user account is reported and the violation probability. For example, the number of times the user account is reported within a preset time period can be obtained; a first preset weight and the number of times are multiplied to obtain a first parameter value, a second preset weight and the violation probability are multiplied to obtain a second parameter value; the first parameter value and the second parameter value are summed to obtain a target parameter value; and the violation level of the user account is determined based on the target parameter value.

[0150] In some implementations, the server can record the marking time of the user account being marked as a violation account, i.e., the time of being reported, and determine the number of times the user account is marked as a violation account within a preset time period based on the marking time, wherein the more times the user account is marked as a violation account, the more likely it is that the user account is a violation account or is maliciously marked. Therefore, when determining the violation level of the user account, the number of times the user account is reported and the violation probability can be considered comprehensively to determine the violation level of the user account more accurately.

[0151] wherein the first preset weight is the weight corresponding to the number of times, the second preset weight is the weight corresponding to the violation probability, the first preset weight is less than the second preset weight, the first preset weight and the number of times are multiplied to obtain a first parameter value, the second preset weight and the violation probability are multiplied to obtain a second parameter value, then the first parameter value and the second parameter value are added to obtain a target parameter value, and finally the violation level is determined based on the target parameter value. When determining the violation level, the influence of the number of times being marked can also be considered comprehensively to improve the accuracy of determining the violation level. When determining the violation level based on the target parameter value, if the target parameter value is greater than or equal to a first threshold, the violation level is determined to be a first violation level; if the target parameter value is less than the first threshold and greater than or equal to a second threshold, the violation level is determined to be a second violation level; and if the target parameter value is less than the second threshold, the violation level is determined to be a third violation level, wherein the first threshold is greater than the second threshold.

[0152] When pushing the user account, the user account can be pushed to the quality inspection platform according to the violation level, so that the user account can be audited by a human to determine whether the user account is a violation account.

[0153] In some implementations, to improve the accuracy of the foregoing probability calculation model, the audit result can be used for iteration of the probability calculation model. For example, the audit result of the user account audited by a human can be obtained once at a preset time interval, and the operation time sequence data of the user account and the audit result are used as training data to continue updating and iterating the probability calculation model to ensure the adaptability of the probability calculation model and continuously improve the accuracy of violation probability calculation, thereby improving the detection rate of violation accounts.

[0154] After the user account is audited by the human, the audit result corresponding to the user account can be obtained. In some embodiments, the audit result can indicate whether the user account is a violation account, and if so, the violation account can be punished. As an implementation, when the violation account is punished, a preset punishment strategy can be obtained, and the violation account can be punished directly by using the preset punishment strategy.

[0155] As another implementation, when the audit result indicates that the user account is a violation account, a punishment strategy input by a staff can be obtained, and the violation account can be punished according to the punishment strategy. The preset trigger strategy and the punishment strategy can include prohibiting the violation account from publishing information within a certain time period, recycling the violation account, and the like, which can be set according to actual needs. If the audit result indicates that the user account is not a violation account, the process ends. By punishing the violation account, the violation account can be prevented from publishing bad information again, and the security of the network environment or the application program can be enhanced.

[0156] The user account pushing solution provided by the embodiments of the present application can be applied in various scenarios requiring human auditing. For example, taking the human auditing of whether a user account is a violation account as an example, the solution provided by the embodiments of the present application can be used to obtain operation time sequence data of a user account in a preset time period when the user account is marked as a violation account, and the variation characteristics of a specified operation in the preset time period can be obtained after the operation time sequence data is processed. Then, the violation probability of the user account is calculated by using the variation characteristics, and finally, the user account is pushed based on the violation level determined by the violation probability, so as to push the user account with a higher violation level to a quality inspection platform. Compared with the case of randomly selecting a user account, the waste of manpower and time can be reduced under limited auditing capacity, and the detection rate of violation accounts can be improved.

[0157] According to the method described in the above embodiments, the following will be further described in detail.

[0158] In this embodiment, the pushing of a violation account will be taken as an example to describe the method of the embodiments of the present application in detail.

[0159] As shown in FIG. 1, a violation account pushing method includes the following specific process: Figure 2a

[0160] S210, operation time sequence data of a user account in a preset time period before the user account is reported is pulled from a data table, and the operation time sequence data includes operation data of a specified operation performed by the user account.

[0161] S220, the operation time sequence data is input into a violation probability calculation model, and a violation probability of the user account is output. ​

[0162] S230, determine the violation level of the user account based on the violation probability of the user account, and sequentially push the user account to the quality inspection platform based on the violation level.

[0163] S240, obtain the audit result of the user account, and process the user account according to the audit result.

[0164] In order to clearly describe the overall process of pushing the user account, the following can be referred to Figure 2b , which shows the overall architecture diagram of the user account pushing method.

[0165] First, the operation data of the user account can be stored in the data table. Specifically, when storing the operation data of the specified operation, the asynchronous queue can be used to filter the operation data of the specified operation from the actual data generated by the user account, that is, the business data, and store the operation time sequence data in the data table. Through the asynchronous queue, the influence of the data storage process on the actual operation of the user account can be avoided, and the normal information publishing of the user account can be ensured. The specified operation in the embodiment of the application can be a specified information publishing operation, an account login operation, etc. The specified information can be friend adding information, group adding information, etc.

[0166] Suppose a large number of user accounts marked as violation accounts can be detected every day, and the operation time sequence data of these user accounts in a preset time period can be pulled from the data table. Then the operation time sequence data of the user account is output to the probability calculation model to obtain the violation probability corresponding to the user account, wherein the probability calculation model specifically calculates the violation probability in the following steps:

[0167] 1. Encode the operation time sequence data in the preset time period through the FC layer to realize feature dimension reduction and selection, and obtain a feature vector sequence;

[0168] 2. Input the obtained feature vector sequence into the recurrent neural network for fusion processing to obtain a change feature, wherein the change feature represents the change trend of the specified operation in the preset time period;

[0169] 3. Then the change feature is input into the self-attention layer for attention calculation, and different weights are given to the change features of different time periods, and then the attention feature is obtained;

[0170] 4. Input the attention feature into the full connection layer to realize classification, and finally output the violation probability.

[0171] Then, the violation level of the user account is determined according to the violation probability, and the user account is pushed to the quality inspection platform according to the violation level.

[0172] As an implementation manner, the user account can be determined according to a violation level according to the violation probability, wherein the higher the violation probability is, the higher the violation level is, and the user account is pushed in sequence according to the order from high to low of the violation level. For example, the violation probability of the user account a is y1, the violation probability of the user account b is y2, the violation probability of the user account c is y3, and the violation probability of the user account d is y4, wherein y2>y4>y1>y3, and the violation level of the user account from high to low is: the user account b, the user account d, the user account a, and the user account c, and then the user account can be pushed in the following order: the user account b, the user account d, the user account a, and the user account c.

[0173] The staff can manually audit the user account b, the user account d, the user account a, and the user account c in sequence, determine whether the user account is a violation account, if the user account is a violation account, the user account can be punished, and then the process ends, if the user account is not a violation account, the process can directly end. Therefore, in the case that the manual auditing capacity is limited, the user account with a high violation level can be audited first, so as to improve the detection rate of the violation account.

[0174] As can be seen from the above, the operation time sequence data of the user account in the preset time period before being reported can be pulled from the data table, the operation time sequence data of the user account is input into the probability calculation model, the violation probability of the user account is obtained, the violation level of the user account is determined based on the violation probability, and the user account is pushed to the quality inspection platform based on the violation level, so as to manually audit whether the user account is a violation account, if the user account is a violation account, the user account is punished. The user account is pushed based on the violation level determined by the violation probability, so as to push the user account with a high violation level to the quality inspection platform. Compared with the case that the user account is randomly selected, in the case that the auditing capacity is limited, the waste of manpower and time can be reduced, and the detection rate of the violation account is improved.

[0175] In order to better implement the above method, the embodiment of the present application also provides a user account pushing device, which can be integrated in an electronic device. The electronic device can be a terminal, a server, etc. The terminal can be a mobile phone, a tablet computer, a smart Bluetooth device, a notebook computer, a personal computer, etc. The server can be a single server or a server cluster composed of multiple servers.

[0176] For example, in the embodiment, the user account pushing device is integrated in the server, and the method of the embodiment of the present application is described in detail.

[0177] For example, as Figure 3As shown, the user account pushing apparatus 300 can include an acquisition module 310, a feature determination module 320, a probability calculation module 330, a level determination module 340, and a pushing module 350.

[0178] The acquisition module 310 is configured to, in a case where it is detected that a user account is reported, acquire operation time sequence data of the user account in a preset time period before the user account is reported, the operation time sequence data including operation data of the user account performing a specified operation arranged in time sequence.

[0179] The feature determination module 320 is configured to determine a variation feature of the specified operation in the preset time period based on the operation time sequence data, the variation feature representing a variation trend of the specified operation in the preset time period.

[0180] The probability calculation module 330 is configured to calculate a violation probability of the user account based on the variation feature.

[0181] The level determination module 340 is configured to determine a violation level of the user account based on the violation probability.

[0182] The pushing module 350 is configured to push the user account according to the violation level.

[0183] In some embodiments, the preset time period includes a plurality of sub-time periods, and the feature determination module 320 further includes:

[0184] A feature extraction unit is configured to perform feature extraction on the operation time sequence data to obtain a feature vector sequence, the feature vector sequence including a feature vector corresponding to each sub-time period.

[0185] A fusion unit is configured to perform data fusion processing on the feature vectors to obtain the variation feature of the specified operation in the preset time period.

[0186] In some embodiments, the fusion unit is further configured to:

[0187] For each feature vector of a current sub-time period, acquire a forward hidden layer state corresponding to a feature vector of a previous sub-time period, the previous sub-time period being a period before the current sub-time period.

[0188] According to the forward hidden layer state and each feature vector, the variation feature of the specified operation in the preset time period is calculated.

[0189] In some embodiments, the fusion unit is further configured to:

[0190] Forward encode each feature vector of a current sub-time period to obtain a forward hidden layer state of each sub-time period.

[0191] backwards encode the feature vector of each current sub-time period to obtain a backwards hidden layer state of each sub-time period;

[0192] splice the forwards hidden layer state and the backwards hidden layer state of each sub-time period to obtain a change feature of the specified operation in a preset time period.

[0193] In some embodiments, the probability calculation module 330 further includes:

[0194] an attention weight calculation unit configured to perform linear transformation on the change feature based on a preset weight parameter to obtain an attention weight corresponding to the change feature;

[0195] an attention feature calculation unit configured to calculate an attention feature corresponding to the change feature based on the attention weight and the preset weight parameter;

[0196] a probability calculation unit configured to obtain a violation probability of the user account according to the attention feature.

[0197] In some embodiments, the preset weight parameter includes a first weight parameter and a second weight parameter, and the attention weight calculation unit is further configured to:

[0198] calculate a product of the change feature and the first weight parameter to obtain a first attention weight;

[0199] calculate a product of the change feature and the second weight parameter to obtain a second attention weight;

[0200] obtain the attention weight corresponding to the change feature according to the first attention weight and the second attention weight.

[0201] In some embodiments, the preset weight parameter further includes a third weight parameter, and the attention feature calculation unit is further configured to:

[0202] perform linear transformation on the change feature based on the third weight parameter to obtain an adjusted change feature;

[0203] calculate a product of the attention weight and the adjusted change feature to obtain the attention feature corresponding to the change feature.

[0204] In some embodiments, the probability calculation unit is further configured to:

[0205] obtain a target parameter and a target bias;

[0206] perform weighted calculation on the attention feature using the target parameter to obtain a weighted attention feature;

[0207] compute a sum of the weighted attention feature and the target bias to obtain a target feature vector;

[0208] normalize the target feature vector to obtain a violation probability of the user account.

[0209] In some embodiments, the level determination module 340 is further configured to:

[0210] obtain a number of times that the user account is reported within the preset time period;

[0211] compute a product of a first preset weight and the number of times to obtain a first parameter value;

[0212] compute a product of a second preset weight and the violation probability to obtain a second parameter value, the first preset weight being less than the second preset weight;

[0213] sum the first parameter value and the second parameter value to obtain a target parameter value;

[0214] determine a violation level of the user account based on the target parameter value.

[0215] In some embodiments, the level determination module 340 is further configured to:

[0216] if the violation probability is greater than or equal to a first probability, determine that the violation level of the user account is a first violation level;

[0217] if the violation probability is less than the first probability, determine a size relationship between the violation probability and a second probability, the second probability being lower than the violation probability indicated by the first probability;

[0218] if the violation probability is greater than or equal to the second probability, determine that the violation level of the user account is a second violation level;

[0219] if the violation probability is less than the second probability, determine that the violation level of the user account is a third violation level.

[0220] In some embodiments, the specified operation includes a publishing operation of specified information by the user account, and the operation data includes a number of times of publishing the specified information.

[0221] In implementation, each of the above units or modules can be implemented as an independent entity, or can be combined as the same or several entities for implementation, and the specific implementation of each of the above units or modules can be referred to the method embodiments above, which will not be described here.

[0222] As can be seen from the above, the user account push device in this embodiment can obtain the operation time sequence data of a user account within a preset time period when the user account is marked as a violation account. After processing the operation time sequence data, the change characteristics of the specified operation within the preset time period can be obtained. Then, the violation probability of the user account can be calculated using the change characteristics. Finally, the user account can be pushed based on the violation level determined by the violation probability, so as to push user accounts with a high violation probability to the quality inspection platform. Compared with the case of randomly selecting user accounts, under limited review capabilities, it can reduce the waste of manpower and time and improve the detection rate of violation accounts.

[0223] This application also provides an electronic device, which can be a terminal, a server, or other similar device. The terminal can be a mobile phone, tablet computer, smart Bluetooth device, laptop computer, personal computer, etc.; the server can be a single server or a server cluster composed of multiple servers, etc.

[0224] In some embodiments, the user account push device can also be integrated into multiple electronic devices. For example, the user account push device can be integrated into multiple servers, and the user account push method of this application can be implemented by multiple servers.

[0225] In this embodiment, a server will be used as an example for detailed description. For example, ... Figure 4 As shown, it illustrates a structural schematic diagram of the electronic device involved in the embodiments of this application, specifically:

[0226] The electronic device may include components such as a processor 401 with one or more processing cores, a memory 402 with one or more computer-readable storage media, a power supply 403, an input module 404, and a communication module 405. Those skilled in the art will understand that... Figure 4 The electronic device structure shown does not constitute a limitation on the electronic device and may include more or fewer components than shown, or combine certain components, or have different component arrangements. Wherein:

[0227] The processor 401 is the control center of the electronic device, connecting various parts of the device via various interfaces and lines. It performs various functions and processes data by running or executing software programs and / or modules stored in the memory 402, and by calling data stored in the memory 402. In some embodiments, the processor 401 may include one or more processing cores; in some embodiments, the processor 401 may integrate an application processor and a modem processor, wherein the application processor mainly handles the operating system, user interface, and applications, and the modem processor mainly handles wireless communication. It is understood that the modem processor may also not be integrated into the processor 401.

[0228] The memory 402 can be used to store software programs and modules, and the processor 401 can execute various function applications and data processing by running the software programs and modules stored in the memory 402. The memory 402 can mainly include a program storage area and a data storage area, wherein the program storage area can store an operating system, application programs required by at least one function (such as a sound playing function, an image playing function, etc.), and the like; and the data storage area can store data created according to the use of the electronic device, etc. In addition, the memory 402 can include a high-speed random access memory, and can also include a non-volatile memory such as at least one magnetic disk storage device, a flash memory device, or other volatile solid-state memory device. Accordingly, the memory 402 can also include a memory controller to provide the processor 401 with access to the memory 402.

[0229] The electronic device also includes a power supply 403 for powering the various components, and in some embodiments, the power supply 403 can be logically connected to the processor 401 through a power management system, so that the power management system can realize functions such as management of charging, discharging, and power consumption management. The power supply 403 can also include one or more direct current or alternating current power supplies, a recharging system, a power supply failure detection circuit, a power supply converter or inverter, a power supply status indicator, and the like.

[0230] The electronic device can also include an input module 404, which can be used to receive input digital or character information, and generate keyboard, mouse, joystick, optical or trackball signal inputs related to user settings and function controls.

[0231] The electronic device can also include a communication module 405, which in some embodiments can include a wireless module, and the electronic device can perform short-range wireless transmission through the wireless module of the communication module 405, thereby providing the user with wireless broadband Internet access. For example, the communication module 405 can be used to help the user send and receive emails, browse web pages, and access streaming media, etc.

[0232] Although not shown, the electronic device can also include a display unit, etc., which will not be described here. Specifically, in the present embodiment, the processor 401 in the electronic device will load the executable file corresponding to the process of one or more application programs into the memory 402 according to the following instructions, and run the application programs stored in the memory 402 by the processor 401, thereby realizing various functions, as follows:

[0233] In the case where it is detected that the user account is reported, operation time sequence data of the user account in a preset time period before being reported is obtained, and the operation time sequence data includes operation data of the user account performing a specified operation in chronological order.

[0234] determine a change feature of the specified operation in the preset time period based on the operation time series data, the change feature representing a change trend of the specified operation in the preset time period;

[0235] calculate a violation probability of the user account based on the change feature;

[0236] determine a violation level of the user account based on the violation probability;

[0237] push the user account according to the violation level.

[0238] The specific implementation of each operation can be referred to the foregoing embodiments, which will not be described here.

[0239] As can be seen from the above, the electronic device provided in the embodiments of the present application can obtain operation time series data of a user account in a preset time period before the user account is reported, and after processing the operation time series data, a change feature of a specified operation in the preset time period can be obtained. Then, the violation probability of the user account is calculated by using the change feature, and finally, the user account is pushed based on the violation level determined by the violation probability, so as to realize pushing the user account with a higher violation level to the quality inspection platform. Compared with the case of randomly selecting a user account, under the limited audit capability, the waste of manpower and time can be reduced, and the detection rate of violation accounts can be improved.

[0240] Those skilled in the art can understand that all or part of the steps of the various methods in the above embodiments can be completed by instructions, or by instructions controlling related hardware, which can be stored in a computer readable storage medium and loaded and executed by a processor.

[0241] Therefore, the embodiments of the present application provide a computer readable storage medium, which stores a plurality of instructions. The instructions can be loaded by a processor to execute the steps in any of the user account pushing methods provided by the embodiments of the present application. For example, the instructions can execute the following steps:

[0242] In the case of detecting that a user account is reported, operation time series data of the user account in a preset time period before the user account is reported is obtained, the operation time series data including operation data of the user account performing a specified operation in a time sequence;

[0243] determine a change feature of the specified operation in the preset time period based on the operation time series data, the change feature representing a change trend of the specified operation in the preset time period;

[0244] calculate a violation probability of the user account based on the change feature;

[0245] determine a violation level of the user account based on the violation probability;

[0246] push the user account according to the violation level.

[0247] The storage medium can include a read-only memory (ROM), a random access memory (RAM), a magnetic disk, an optical disk, or the like.

[0248] According to an aspect of the present application, a computer program product or computer program is provided, which includes computer instructions stored in a computer readable storage medium. A processor of a computer device reads the computer instructions from the computer readable storage medium, and the processor executes the computer instructions to cause the computer device to perform the method provided in any of the various optional implementation manners of the model training aspect or the user account pushing aspect provided in the above embodiments.

[0249] Due to the instructions stored in the storage medium, the steps of any of the user account pushing methods provided in the embodiments of the present application can be performed, and thus the beneficial effects of any of the user account pushing methods provided in the embodiments of the present application can be achieved. Details are described in the above embodiments, and thus will not be described here again.

[0250] The above describes in detail a user account pushing method, device, electronic device and computer readable storage medium provided in the embodiments of the present application. The principles and implementation manners of the present application are described by applying specific examples in this paper. The above embodiment descriptions are only used to help understand the method of the present application and its core idea. Meanwhile, for those skilled in the art, according to the idea of the present application, the specific implementation manner and application range will be changed. In summary, the content of the present description should not be understood as a limitation of the present application.

Claims

1. A user account pushing method, characterized in that, The method comprises: In the case of detecting that a user account is reported, operation time sequence data of the user account in a preset time period before being reported is acquired, the operation time sequence data comprising operation data of the user account performing a specified operation arranged in time sequence; Based on the operation time sequence data, a change feature of the specified operation in the preset time period is determined, comprising: feature extraction is performed on the operation time sequence data to obtain a feature vector sequence, the feature vector sequence comprising a feature vector corresponding to each sub-time period; data fusion processing is performed on the feature vectors to obtain the change feature of the specified operation in the preset time period; the change feature represents a change trend of the specified operation in the preset time period; the preset time period comprises a plurality of sub-time periods; wherein the data fusion processing on the feature vectors to obtain the change feature of the specified operation in the preset time period comprises: for the feature vector of each current sub-time period, a forward hidden layer state corresponding to the feature vector of a previous sub-time period is acquired, the previous sub-time period being a period before the current sub-time period; based on the forward hidden layer state and each feature vector, the change feature of the specified operation in the preset time period is calculated; Based on the change feature, a violation probability of the user account is calculated; Based on the violation probability, a violation level of the user account is determined; The user account is pushed according to the violation level.

2. The method of claim 1, wherein, The data fusion processing on the feature vectors to obtain the change feature of the specified operation in the preset time period further comprises: The feature vector of each current sub-time period is forward encoded to obtain a forward hidden layer state of each sub-time period; The feature vector of each current sub-time period is backward encoded to obtain a backward hidden layer state of each sub-time period; The forward hidden layer state and the backward hidden layer state of each sub-time period are spliced to obtain the change feature of the specified operation in the preset time period.

3. The method of claim 1, wherein, The calculation of the violation probability of the user account based on the change feature comprises: Based on a preset weight parameter, a linear transformation is performed on the change feature to obtain an attention weight corresponding to the change feature; Based on the attention weight and the preset weight parameter, an attention feature corresponding to the change feature is calculated; The violation probability of the user account is obtained according to the attention feature.

4. The method of claim 3, wherein, The preset weight parameter comprises a first weight parameter and a second weight parameter, and the linear transformation performed on the change feature based on the preset weight parameter to obtain the attention weight corresponding to the change feature comprises: A product of the change feature and the first weight parameter is calculated to obtain a first attention weight; A product of the change feature and the second weight parameter is calculated to obtain a second attention weight; The attention weight corresponding to the change feature is obtained according to the first attention weight and the second attention weight.

5. The method of claim 3, wherein, The preset weight parameter further comprises a third weight parameter, and the calculation of the attention feature corresponding to the change feature based on the attention weight and the preset weight parameter comprises: linearly transform the change feature based on the third weight parameter to obtain an adjusted change feature; calculate a product of the attention weight and the adjusted change feature to obtain an attention feature corresponding to the change feature.

6. The method of claim 3, wherein, The obtaining of the violation probability of the user account according to the attention feature comprises: obtaining a target parameter and a target bias; performing weighted calculation on the attention feature using the target parameter to obtain a weighted attention feature; calculating a sum of the weighted attention feature and the target bias to obtain a target feature vector; performing normalization processing on the target feature vector to obtain the violation probability of the user account.

7. The method of claim 1, wherein, The pushing of the user account according to the violation level comprises: obtaining a number of times that the user account is reported within the preset time period; calculating a product of a first preset weight and the number of times to obtain a first parameter value; calculating a product of a second preset weight and the violation probability to obtain a second parameter value, the first preset weight being smaller than the second preset weight; summing the first parameter value and the second parameter value to obtain a target parameter value; determining the violation level of the user account based on the target parameter value.

8. The method of claim 1, wherein, The determining of the violation level of the user account based on the violation probability comprises: if the violation probability is greater than or equal to a first probability, determining that the violation level of the user account is a first violation level; if the violation probability is less than the first probability, determining a size relationship between the violation probability and a second probability, the second probability being lower than the violation probability indicated by the first probability; if the violation probability is greater than or equal to the second probability, determining that the violation level of the user account is a second violation level; if the violation probability is less than the second probability, determining that the violation level of the user account is a third violation level.

9. The method according to any one of claims 1 to 8, characterized in that, The specified operation comprises a publishing operation of specified information by the user account, and the operation data comprises a number of times of publishing the specified information.

10. A user account pushing apparatus, characterized by comprising: The apparatus comprises: an obtaining module configured to, in a case where a user account is reported, obtain operation time sequence data of the user account in a preset time period before being reported, the operation time sequence data comprising operation data of a specified operation of the user account arranged in time sequence; The feature determination module is configured to determine a change feature of the specified operation in the preset time period based on the operation time series data, including: performing feature extraction on the operation time series data to obtain a feature vector sequence, the feature vector sequence including a feature vector corresponding to each sub-time period; performing data fusion processing on the feature vectors to obtain the change feature of the specified operation in the preset time period; the change feature representing a change trend of the specified operation in the preset time period; the preset time period including a plurality of sub-time periods; wherein the data fusion processing on the feature vectors to obtain the change feature of the specified operation in the preset time period includes: for each feature vector of a current sub-time period, obtaining a forward hidden layer state corresponding to a feature vector of a previous sub-time period, the previous sub-time period being a period before the current sub-time period; and calculating the change feature of the specified operation in the preset time period based on the forward hidden layer state and each feature vector. The probability calculation module is configured to calculate a violation probability of the user account based on the change feature. The level determination module is configured to determine a violation level of the user account based on the violation probability. The pushing module is configured to push the user account according to the violation level.

11. An electronic device, comprising: The computer readable storage medium stores a plurality of instructions, and the instructions are suitable for being loaded by the processor to execute the steps in the user account pushing method according to any one of claims 1-9.

12. A computer-readable storage medium, characterized in that, The computer readable storage medium stores a plurality of instructions, and the instructions are suitable for being loaded by the processor to execute the steps in the user account pushing method according to any one of claims 1-9.

13. A computer program product, characterised in that, The computer readable storage medium stores a plurality of instructions, and the instructions are suitable for being loaded by the processor to execute the steps in the user account pushing method according to any one of claims 1-9.

Images