User portrait classification method, device, equipment and medium based on deep learning

Abstract

The present invention relates to the field of artificial intelligence, and discloses a user portrait classification method, device, electronic device and storage medium based on deep learning, wherein the method comprises: obtaining user data of a user portrait to be classified, and preprocessing the user data to obtain target data; collecting similar user portraits of the user portrait to be classified, and extracting similar user features of the similar user portraits; inputting the target data and similar user features into a preset user classification model as model data, so as to detect the temporal correlation features of the model data using a temporal memory network in the user classification model; extracting the information semantic sequence of the temporal correlation features using an attention mechanism network in the user classification model, identifying the data category of the model data using a convolutional network in the user classification model according to the information semantic sequence, and determining the user category of the user portrait to be classified according to the data category. The present invention can improve the accuracy of user portrait classification.

Description

Technical Field

[0001] The present invention relates to the field of artificial intelligence, and in particular to a user portrait classification method, device, electronic device and storage medium based on deep learning. Background Art

[0002] User portraits can be understood as modeling of users in the real world. Through user portraits, some highly generalized and easy-to-understand features can be used to describe users, making it more convenient to describe users. However, user portraits usually involve a broader and more complex user dimension. Therefore, it is particularly important to accurately locate the category of user portraits and ensure accurate positioning and analysis of user portraits.

[0003] At present, user portrait classification is usually based on the portrait features in the user portrait, and an ensemble learning classifier is used to implement portrait category analysis. However, when generating a base classifier cluster through an ensemble learning classifier, some redundant individuals with similar user portrait classification results will be generated, resulting in a decrease in cluster differentiation and an imbalance in user portrait categories, which will affect the accuracy of user portrait classification. Summary of the invention

[0004] In order to solve the above problems, the present invention provides a user portrait classification method, device, electronic device and storage medium based on deep learning, which can improve the accuracy of user portrait classification.

[0005] In a first aspect, the present invention provides a user portrait classification method based on deep learning, comprising:

[0006] Obtaining user data of the user portrait to be classified, and preprocessing the user data to obtain target data;

[0007] Collecting similar user portraits of the user portrait to be classified, and extracting similar user features of the similar user portraits;

[0008] Inputting the target data and the similar user features as model data into a preset user classification model, so as to detect the temporal correlation features of the model data using a temporal memory network in the user classification model;

[0009] The attention mechanism network in the user classification model is used to extract the information semantic sequence of the temporal correlation features. According to the information semantic sequence, the convolutional network in the user classification model is used to identify the data category of the model data. According to the data category, the user category of the user portrait to be classified is determined.

[0010] In a possible implementation manner of the first aspect, preprocessing the user data to obtain target data includes:

[0011] Extracting a data field of each data in the user data, and identifying a data attribute of each data in the user data according to the data field;

[0012] The data attributes are matched with a preset business attribute table, and the data corresponding to the successfully matched data attributes are used as the target data.

[0013] In a possible implementation manner of the first aspect, collecting similar user portraits of the user portrait to be classified includes:

[0014] Acquire user attributes of the user portrait to be classified, filter the user attributes, and construct branch attributes of the filtered user attributes;

[0015] Collect historical user portraits of the branch attributes, and calculate similar user scores for each user portrait in the historical user portraits;

[0016] The user portrait whose similar user score is greater than the preset score is selected as the similar user portrait.

[0017] In a possible implementation manner of the first aspect, the detecting the temporal correlation features of the model data by using the temporal memory network in the user classification model includes:

[0018] Calculating the state value of the model data using the input gate in the temporal memory network, and reducing the dimension of the state value using the projection layer in the temporal memory network to obtain a reduced-dimensional state value;

[0019] Calculating the activation value of the model data using the forget gate in the temporal memory network, and reducing the dimension of the activation value of the dimensionality reduction state value using the projection layer in the temporal memory network to obtain a dimensionality reduction activation value;

[0020] Calculate the state update value of the model data according to the dimension reduction state value and the dimension reduction activation value;

[0021] The output gate in the temporal memory network is used to calculate the associated feature sequence of the state update value, and the projection layer in the temporal memory network is used to reduce the dimension of the associated feature sequence to obtain the temporal associated features of the model data.

[0022] In a possible implementation manner of the first aspect, the extracting the information semantic sequence of the temporal association feature by using the attention mechanism network in the user classification model includes:

[0023] Using the position encoding layer in the attention mechanism network to position encode the temporal association features to obtain encoded association features;

[0024] Using the encoder in the attention mechanism to perform feature encoding on the encoding-related features to obtain key encoding features;

[0025] The decoder in the attention mechanism is used to decode the semantic information of the key encoding features to obtain an information semantic sequence.

[0026] In a possible implementation of the first aspect, the using the position encoding layer in the attention mechanism network to position encode the temporal association features to obtain the encoded association features includes:

[0027] The following formula is used to position encode the temporal correlation feature:

[0028]

[0029]

[0030] Among them, E (p,2x) represents the even-numbered coded correlation feature in the time series correlation feature, pos represents the position coding index, x represents the i-th feature vector in the time series correlation feature, d ml represents the feature vector dimension, E (p,2x+1) Indicates the odd-numbered coded correlation features in the temporal correlation features.

[0031] In a possible implementation manner of the first aspect, identifying the data category of the model data by using a convolutional network in the user classification model according to the information semantic sequence includes:

[0032] According to the information semantic sequence, using the convolution layer in the convolution network to perform a convolution operation on the model data to obtain feature data;

[0033] Standardizing the feature data using a standard layer in the convolutional network to obtain standard data;

[0034] The fully connected layer in the convolutional network is used to calculate the data category probability of the standard data, and the output layer in the convolutional network is used to output the data category of the model data according to the data category probability.

[0035] In a second aspect, the present invention provides a user portrait classification device based on deep learning, the device comprising:

[0036] A data preprocessing module is used to obtain user data of the user portrait to be classified, and preprocess the user data to obtain target data;

[0037] A similar user collection module is used to collect similar user portraits of the user portrait to be classified, and extract similar user features of the similar user portraits;

[0038] A correlation feature detection module, used to input the target data and the similar user features as model data into a preset user classification model, so as to detect the temporal correlation features of the model data using the temporal memory network in the user classification model;

[0039] A user category identification module is used to utilize the attention mechanism network in the user classification model to extract the information semantic sequence of the temporal correlation features, and based on the information semantic sequence, utilize the convolutional network in the user classification model to identify the data category of the model data, and determine the user category of the user portrait to be classified based on the data category.

[0040] In a third aspect, the present invention provides an electronic device, comprising:

[0041] at least one processor; and a memory communicatively coupled to the at least one processor;

[0042] In which, the memory stores a computer program that can be executed by the at least one processor, so that the at least one processor can execute the user portrait classification method based on deep learning as described in any one of the first aspects above.

[0043] In a fourth aspect, the present invention provides a computer-readable storage medium storing a computer program, which, when executed by a processor, implements a user portrait classification method based on deep learning as described in any one of the first aspects above.

[0044] Compared with the prior art, the technical principle and beneficial effects of this solution are:

[0045] The embodiment of the present invention firstly pre-processes the user data of the user portrait to be classified to obtain the target data, so as to filter out the useless data in the user data, ensure the timeliness and speed of the subsequent data processing, and extract the similar user features after collecting the similar user portraits of the user portrait to be classified, so as to enrich the portrait data of the user portrait to be classified, expand the information semantics in the subsequent processing of the user portrait data to be classified, and ensure the classification accuracy of the user portrait to be classified; secondly, the embodiment of the present invention inputs the target data and the similar user features as the model data into the preset user classification model, so as to utilize the temporal memory network in the user classification model to detect the model data. The time-series correlation features can mine the long-time correlation features in the model data, avoid the phenomenon of imbalanced portrait categories in subsequent user portraits, and ensure the classification accuracy of subsequent user portraits to be classified; further, the embodiment of the present invention utilizes the attention mechanism network in the user classification model to extract the information semantic sequence of the time-series correlation features, and utilizes the convolutional network in the user classification model to identify the data category of the model data, and determines the user category of the user portrait to be classified according to the data category, so that the model data can contain richer semantic information when classifying, improve the accuracy of the model data, and thus ensure the classification accuracy of the user portrait to be classified. Therefore, the user portrait classification method, device, electronic device and storage medium based on deep learning proposed in the embodiment of the present invention can improve the accuracy of user portrait classification. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the invention and, together with the description, serve to explain the principles of the invention.

[0047] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, for ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative labor.

[0048] Figure 1 A flowchart of a user portrait classification method based on deep learning provided by one embodiment of the present invention;

[0049] Figure 2 In one embodiment of the present invention Figure 1 A flowchart of one of the steps in a user portrait classification method based on deep learning is provided;

[0050] Figure 3 In one embodiment of the present invention Figure 1A flowchart of another step of a user portrait classification method based on deep learning is provided;

[0051] Figure 4 A schematic diagram of a module of a user portrait classification device based on deep learning provided by an embodiment of the present invention;

[0052] Figure 5 A schematic diagram of the internal structure of an electronic device for implementing a user portrait classification method based on deep learning provided by one embodiment of the present invention. DETAILED DESCRIPTION

[0053] It should be understood that the specific embodiments described herein are only used to explain the present invention, and are not used to limit the present invention.

[0054] The embodiment of the present invention provides a user portrait classification method based on deep learning, and the execution subject of the user portrait classification method based on deep learning includes but is not limited to at least one of the electronic devices such as a server and a terminal that can be configured to execute the method provided by the embodiment of the present invention. In other words, the user portrait classification method based on deep learning can be executed by software or hardware installed on a terminal device or a server device, and the software can be a blockchain platform. The server includes but is not limited to: a single server, a server cluster, a cloud server or a cloud server cluster, etc. The server can be an independent server, or it can be a cloud server that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communications, middleware services, domain name services, security services, content delivery networks (Content Delivery Network, CDN), and big data and artificial intelligence platforms.

[0055] See also Figure 1 FIG. 1 is a flow chart of a user portrait classification method based on deep learning provided by an embodiment of the present invention. Figure 1 The deep learning-based user portrait classification methods described in include:

[0056] S1. Obtain user data of the user portrait to be classified, and preprocess the user data to obtain target data.

[0057] In an embodiment of the present invention, the user portrait to be classified refers to a portrait that classifies a labeled user model abstracted from user information. By classifying the user portrait, the user can be accurately located in the user category to which he belongs, thereby facilitating the precise analysis of subsequent users and achieving precise recommendation of user data. The user data can be understood as dimensional information used to describe the user corresponding to the user portrait to be classified, which includes basic user information and user behavior information. In detail, the basic user information includes name, age, etc., and the user behavior information includes browsing history, favorite history, and other information.

[0058] Furthermore, the embodiment of the present invention pre-processes the user data to filter out useless data in the user data, thereby ensuring the timeliness and speed of subsequent data processing.

[0059] As an embodiment of the present invention, the preprocessing of the user data to obtain the target data includes: extracting the data field of each data in the user data, identifying the data attributes of each data in the user data according to the data field, matching the data attributes with a preset business attribute table, and using the data corresponding to the successfully matched data attributes as the target data.

[0060] Among them, the data field refers to the information dimension field used to characterize each data in the user data, the data attribute refers to the information feature used to describe each data in the user data, and the preset business attribute table is set based on different business scenarios, which is used to filter out the attributes required in the corresponding business environment. For example, attributes such as date and behavior can be set as business attributes in the business attribute table.

[0061] Furthermore, in an optional embodiment of the present invention, the extraction of the data field can be achieved through a preset automation script, the preset automation script can be compiled through the js scripting language, the data attributes can be identified by querying the field object features corresponding to the data field, and the data attributes and the preset business attribute table can be matched through a matching algorithm, such as a similarity algorithm, a distance formula, etc.

[0062] S2. Collect similar user portraits of the user portrait to be classified, and extract similar user features of the similar user portraits.

[0063] The embodiment of the present invention collects similar user portraits of the user portrait to be classified to enrich the portrait data of the user portrait to be classified, thereby expanding the information semantics in the subsequent processing of the user portrait data to be classified and ensuring the classification accuracy of the user portrait to be classified. The similar user portrait refers to a portrait with the same dimensional information as the user portrait to be classified.

[0064] As an embodiment of the present invention, see Figure 2 As shown, the collecting of similar user portraits of the user portrait to be classified includes:

[0065] S201, obtaining user attributes of the user portrait to be classified, screening the user attributes, and constructing branch attributes of the screened user attributes;

[0066] S202: Collect historical user portraits of the branch attributes, and calculate similar user scores for each user portrait in the historical user portraits;

[0067] S203: Select the user portrait whose similar user score is greater than a preset score as the similar user portrait.

[0068] Among them, the user attribute refers to the user information dimension used to express the user portrait to be classified, which can be obtained by querying according to the above-mentioned data attributes. The attribute branch refers to the attribute used to refine the attributes existing in the user attribute. For example, if the user attribute is gender, it can be a branch attribute of male / female.

[0069] Furthermore, in an optional embodiment of the present invention, the branch attributes of the user attributes are set based on different user needs, the similar user portraits can be collected through point-of-sale technology, and the similar user scores can be calculated in a weighted form. For example, the number of branch attributes that each user portrait in the similar user portraits has the same as the user portrait to be classified can be counted to realize the calculation of the similar user scores. The preset score can be set to 80, and can also be set according to the actual business scenario.

[0070] Furthermore, the embodiment of the present invention extracts similar user features of the similar user portraits to filter out useless data of the similar user portraits, thereby improving subsequent data processing efficiency.

[0071] As an embodiment of the present invention, the extracting of similar user features of the similar user portraits includes: querying similar user attributes in the similar user portraits, calculating the attribute weight of each user attribute in the similar user attributes, and taking the user attribute corresponding to the attribute weight greater than the preset weight as the similar user feature.

[0072] Among them, the similar user attributes refer to the dimensional information used to characterize the similar users in the similar user portraits, and the attribute weight refers to the weight ratio used to describe the user attribute in the similar user attributes.

[0073] Furthermore, in an optional embodiment of the present invention, the similar user attributes can be queried through an SQL query statement, the attribute weight can be calculated through an information entropy algorithm, and the preset weight can be set to 0.8, or can be set according to an actual business scenario.

[0074] S3. Input the target data and the similar user features as model data into a preset user classification model, so as to use the temporal memory network in the user classification model to detect the temporal correlation features of the model data.

[0075] The embodiment of the present invention inputs the target data and the similar user features as model data into a preset user classification model to intelligently detect the user category of the user portrait to be classified. It should be noted that, in the embodiment of the present invention, the preset user classification model is constructed using an end-to-end neural network to reduce human participation in actions and realize automatic classification of the user portrait to be classified. Furthermore, the preset user classification model includes a temporal memory network, an attention mechanism network and a convolutional network. The temporal memory network includes an LSTM network, the attention mechanism network includes a Transform network, and the convolutional network includes a CNN network.

[0076] Furthermore, before the embodiment of the present invention inputs the target data and the similar user features as model data into a preset user classification model, it also includes: converting the target data and the similar user features into data in vector form to ensure the data processing premise of the subsequent model. Optionally, the vector conversion of the target data and the similar user features is implemented by a vector conversion algorithm, such as word2vec, one-hot and other algorithms.

[0077] Furthermore, the embodiment of the present invention utilizes the temporal memory network in the user classification model to detect the temporal correlation features of the model data, so as to mine the long-term temporal correlation features in the model data and ensure the classification accuracy of the subsequent user portraits to be classified.

[0078] As an embodiment of the present invention, the use of the temporal memory network in the user classification model to detect the temporal association characteristics of the model data includes: using the input gate in the temporal memory network to calculate the state value of the model data, and using the projection layer in the temporal memory network to reduce the dimension of the state value to obtain a reduced dimension state value; using the forgetting gate in the temporal memory network to calculate the activation value of the model data, and using the projection layer in the temporal memory network to reduce the dimension of the reduced dimension state value to obtain a reduced dimension activation value; calculating the state update value of the model data according to the reduced dimension state value and the reduced dimension activation value; using the output gate in the temporal memory network to calculate the associated feature sequence of the state update value, and using the projection layer in the temporal memory network to reduce the dimension of the associated feature sequence to obtain the temporal association characteristics of the model data.

[0079] Among them, the input gate is used to selectively record new information in the model data into the cell state, the forget gate is used to selectively forget the information in the cell state, the output gate is used to save the selected information sequence, and the projection layer is used to reduce the model parameters of the temporal memory network and improve its computational efficiency.

[0080] S4. Utilize the attention mechanism network in the user classification model to extract the information semantic sequence of the temporal correlation features, and based on the information semantic sequence, utilize the convolutional network in the user classification model to identify the data category of the model data, and determine the user category of the user portrait to be classified based on the data category.

[0081] The embodiment of the present invention utilizes the attention mechanism network in the user classification model to extract the information semantic sequence of the temporal correlation features to obtain the semantic information carried in the temporal correlation features, thereby further ensuring the subsequent classification accuracy of the user portraits to be classified.

[0082] As an embodiment of the present invention, see Figure 3 As shown, the method of extracting the information semantic sequence of the temporal correlation feature by using the attention mechanism network in the user classification model includes:

[0083] S301, using the position encoding layer in the attention mechanism network to position encode the temporal association features to obtain encoded association features;

[0084] S302, using the encoder in the attention mechanism to perform feature encoding on the encoding-related features to obtain key encoding features;

[0085] S303: Use the decoder in the attention mechanism to decode the semantic information of the key coding features to obtain an information semantic sequence.

[0086] Among them, the position encoding layer is used to determine the position sequence information of the time series correlation features to make up for the lack of position information of the subsequent time series correlation features in the feature encoding process; the encoder is used to compress the time series correlation features after position encoding into a latent space representation, thereby extracting the corresponding information speech features; the decoder is used to reconstruct the encoded time series correlation features into an information semantic sequence.

[0087] Furthermore, in an optional embodiment of the present invention, the following formula is used to perform position encoding on the temporal correlation feature:

[0088]

[0089]

[0090] Among them, E (p,2x) represents the even-numbered coded correlation feature in the time series correlation feature, pos represents the position coding index, x represents the i-th feature vector in the time series correlation feature, d ml represents the feature vector dimension, E (p,2x+1) Indicates the odd-numbered coded correlation features in the temporal correlation features.

[0091] Furthermore, in an optional embodiment of the present invention, the feature encoding of the encoding association feature can be implemented by a self-attention module in the encoder, and the semantic information decoding of the key encoding feature can be implemented by a self-attention module in the decoder.

[0092] Furthermore, an embodiment of the present invention utilizes a convolutional network in the user classification model to identify the data category of the model data according to the information semantic sequence, so that the model data can contain richer semantic information when classifying, thereby improving the accuracy of the model data and ensuring the classification accuracy of subsequent user portraits to be classified.

[0093] As an embodiment of the present invention, the method of identifying the data category of the model data by using the convolutional network in the user classification model according to the information semantic sequence includes: performing a convolution operation on the model data by using the convolutional layer in the convolutional network according to the information semantic sequence to obtain feature data, standardizing the feature data by using the standard layer in the convolutional network to obtain standard data, calculating the data category probability of the standard data by using the fully connected layer in the convolutional network, and outputting the data category of the model data by using the output layer in the convolutional network according to the data category probability.

[0094] Among them, the convolution layer realizes the feature extraction of the model data by constructing convolution branches, the standard layer is used to normalize the feature data to reduce the calculation complexity of subsequent data, and the fully connected layer is used to fuse all data features in the standard data to realize the category probability calculation of the model data.

[0095] Furthermore, in an optional embodiment of the present invention, the convolution operation of the model data is implemented by the convolution kernel in the convolution layer, the standardization of the feature data is implemented by the normalization function in the standard layer, and the data category probability of the standard data is calculated by the activation function in the fully connected layer, such as the softmax function.

[0096] Furthermore, in another optional embodiment of the present invention, the normalization function includes:

[0097]

[0098] Among them, P(x) represents standard data, x represents characteristic data, μ(x) represents the mean function of standard data, and σ(x) represents the standard deviation function of standard data.

[0099] Furthermore, the embodiment of the present invention determines the user category of the user portrait to be classified according to the data category, so as to clarify the category to which the user of the user portrait to be classified belongs, thereby facilitating subsequent analysis and mining of user information.

[0100] It can be seen that the embodiment of the present invention firstly obtains the target data by preprocessing the user data of the user portrait to be classified, so as to filter out the useless data in the user data, ensure the timeliness and speed of subsequent data processing, and extracts the similar user features after collecting the similar user portraits of the user portrait to be classified, which can enrich the portrait data of the user portrait to be classified, expand the information semantics in the subsequent processing of the user portrait data to be classified, and ensure the classification accuracy of the user portrait to be classified; secondly, the embodiment of the present invention inputs the target data and the similar user features as the model data into the preset user classification model, so as to utilize the temporal memory network detection model in the user classification model The temporal correlation features of the data can mine the long-term temporal correlation features in the model data, avoid the phenomenon of imbalanced portrait categories in subsequent user portraits, and ensure the classification accuracy of subsequent user portraits to be classified; further, the embodiment of the present invention utilizes the attention mechanism network in the user classification model to extract the information semantic sequence of the temporal correlation features, and utilizes the convolutional network in the user classification model to identify the data category of the model data, and determines the user category of the user portrait to be classified according to the data category, so that the model data can contain richer semantic information when classifying, improve the accuracy of the model data, and thus ensure the classification accuracy of the user portrait to be classified. Therefore, a user portrait classification method based on deep learning proposed in an embodiment of the present invention can improve the accuracy of user portrait classification.

[0101] like Figure 4 The figure shows a functional module diagram of the user portrait classification device based on deep learning of the present invention.

[0102] The user portrait classification device 400 based on deep learning of the present invention can be installed in an electronic device. According to the functions implemented, the user portrait classification device based on deep learning can include a data preprocessing module 401, a similar user acquisition module 402, an associated feature detection module 403 and a user category identification module 404. The module of the present invention can also be called a unit, which refers to a series of computer program segments that can be executed by an electronic device processor and can complete fixed functions, which are stored in the memory of the electronic device.

[0103] In the embodiment of the present invention, the functions of each module / unit are as follows:

[0104] The data preprocessing module 401 is used to obtain user data of the user portrait to be classified, and preprocess the user data to obtain target data;

[0105] The similar user collection module 402 is used to collect similar user portraits of the user portrait to be classified, and extract similar user features of the similar user portraits;

[0106] The correlation feature detection module 403 is used to input the target data and the similar user features as model data into a preset user classification model, so as to detect the temporal correlation features of the model data using the temporal memory network in the user classification model;

[0107] The user category identification module 404 is used to utilize the attention mechanism network in the user classification model to extract the information semantic sequence of the temporal correlation features, and based on the information semantic sequence, utilize the convolutional network in the user classification model to identify the data category of the model data, and determine the user category of the user portrait to be classified based on the data category.

[0108] In detail, each module in the user portrait classification device 400 based on deep learning in the embodiment of the present invention is used in the same manner as above. Figures 1 to 3 The user portrait classification method based on deep learning described in the previous section has the same technical means and can produce the same technical effects, so I will not go into details here.

[0109] like Figure 5 The figure is a schematic diagram of the structure of an electronic device that implements the user portrait classification method based on deep learning in the present invention.

[0110] The electronic device may include a processor 50, a memory 51, a communication bus 52 and a communication interface 53, and may also include a computer program stored in the memory 51 and executable on the processor 50, such as a user portrait classification program based on deep learning.

[0111] In some embodiments, the processor 50 may be composed of an integrated circuit, for example, a single packaged integrated circuit, or a plurality of packaged integrated circuits with the same or different functions, including one or more central processing units (CPUs), microprocessors, digital processing chips, graphics processors, and combinations of various control chips. The processor 50 is the control core (ControlUnit) of the electronic device, and uses various interfaces and lines to connect the various components of the entire electronic device, and executes or executes programs or modules stored in the memory 51 (for example, executing a user portrait classification program based on deep learning, etc.), and calls the data stored in the memory 51 to execute various functions of the electronic device and process data.

[0112] The memory 51 includes at least one type of readable storage medium, and the readable storage medium includes a flash memory, a mobile hard disk, a multimedia card, a card-type memory (for example, an SD or DX memory, etc.), a magnetic memory, a disk, an optical disk, etc. The memory 51 may be an internal storage unit of an electronic device in some embodiments, such as a mobile hard disk of the electronic device. The memory 51 may also be an external storage device of an electronic device in other embodiments, such as a plug-in mobile hard disk, a smart memory card (Smart Media Card, SMC), a secure digital (Secure Digital, SD) card, a flash card (Flash Card), etc. equipped on the electronic device. Further, the memory 51 may also include both an internal storage unit of the electronic device and an external storage device. The memory 51 may not only be used to store application software and various types of data installed in the electronic device, such as the code of the database configuration connection program, but also be used to temporarily store data that has been output or is to be output.

[0113] The communication bus 52 may be a peripheral component interconnect (PCI) bus or an extended industry standard architecture (EISA) bus, etc. The bus may be divided into an address bus, a data bus, a control bus, etc. The bus is configured to realize connection and communication between the memory 51 and at least one processor 50, etc.

[0114] The communication interface 53 is used for communication between the above-mentioned electronic device 5 and other devices, including a network interface and a user interface. Optionally, the network interface may include a wired interface and / or a wireless interface (such as a WI-FI interface, a Bluetooth interface, etc.), which is generally used to establish a communication connection between the electronic device and other electronic devices. The user interface may be a display (Display), an input unit (such as a keyboard (Keyboard)), and optionally, the user interface may also be a standard wired interface, a wireless interface. Optionally, in some embodiments, the display may be an LED display, a liquid crystal display, a touch-sensitive liquid crystal display, and an OLED (Organic Light-Emitting Diode, organic light-emitting diode) touch device, etc. Among them, the display may also be appropriately referred to as a display screen or a display unit, which is used to display information processed in the electronic device and to display a visual user interface.

[0115] Figure 5 Only an electronic device with components is shown, and those skilled in the art will understand that Figure 5The structure shown does not constitute a limitation on the electronic device, and may include fewer or more components than shown in the figure, or combine certain components, or arrange the components differently.

[0116] For example, although not shown, the electronic device may also include a power source (such as a battery) for supplying power to various components. Preferably, the power source may be logically connected to the at least one processor 50 through a power management device, so that the power management device can realize functions such as charging management, discharging management, and power consumption management. The power source may also include any components such as one or more DC or AC power sources, recharging devices, power failure detection circuits, power converters or inverters, power status indicators, etc. The electronic device may also include a variety of sensors, Bluetooth modules, Wi-Fi modules, etc., which will not be repeated here.

[0117] It should be understood that the embodiment is for illustrative purposes only and the scope of the patent invention is not limited to this structure.

[0118] The database configuration connection program stored in the memory 51 in the electronic device is a combination of multiple computer programs. When running in the processor 50, it can achieve:

[0119] Obtaining user data of the user portrait to be classified, and preprocessing the user data to obtain target data;

[0120] Collecting similar user portraits of the user portrait to be classified, and extracting similar user features of the similar user portraits;

[0121] Inputting the target data and the similar user features as model data into a preset user classification model, so as to detect the temporal correlation features of the model data using a temporal memory network in the user classification model;

[0122] The attention mechanism network in the user classification model is used to extract the information semantic sequence of the temporal correlation features. According to the information semantic sequence, the convolutional network in the user classification model is used to identify the data category of the model data. According to the data category, the user category of the user portrait to be classified is determined.

[0123] Specifically, the specific implementation method of the processor 50 for the above computer program can refer to Figure 1 The description of the relevant steps in the corresponding embodiments will not be repeated here.

[0124] Furthermore, if the module / unit integrated in the electronic device is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a non-volatile computer-readable storage medium. The storage medium can be volatile or non-volatile. For example, the computer-readable medium can include: any entity or device capable of carrying the computer program code, a recording medium, a USB flash drive, a mobile hard disk, a magnetic disk, an optical disk, a computer memory, and a read-only memory (ROM).

[0125] The present invention further provides a storage medium, wherein the readable storage medium stores a computer program, and when the computer program is executed by a processor of an electronic device, the computer program can implement:

[0126] Obtaining user data of the user portrait to be classified, and preprocessing the user data to obtain target data;

[0127] Collecting similar user portraits of the user portrait to be classified, and extracting similar user features of the similar user portraits;

[0128] Inputting the target data and the similar user features as model data into a preset user classification model, so as to detect the temporal correlation features of the model data using a temporal memory network in the user classification model;

[0129] The attention mechanism network in the user classification model is used to extract the information semantic sequence of the temporal correlation features. According to the information semantic sequence, the convolutional network in the user classification model is used to identify the data category of the model data. According to the data category, the user category of the user portrait to be classified is determined.

[0130] In the several embodiments provided by the present invention, it should be understood that the disclosed devices, apparatuses and methods can be implemented in other ways. For example, the device embodiments described above are only schematic, for example, the division of the modules is only a logical function division, and there may be other division methods in actual implementation.

[0131] The modules described as separate components may or may not be physically separated, and the components shown as modules may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0132] In addition, each functional module in each embodiment of the present invention may be integrated into one processing unit, each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of hardware plus software functional modules.

[0133] It is obvious to those skilled in the art that the present invention is not limited to the details of the above exemplary embodiments, and that the present invention can be implemented in other specific forms without departing from the spirit or essential characteristics of the present invention.

[0134] Therefore, no matter from which point of view, the embodiments should be regarded as illustrative and non-restrictive, and the scope of the present invention is limited by the appended claims rather than the above description, so it is intended that all changes falling within the meaning and scope of the equivalent elements of the claims are included in the present invention. Any attached figure mark in the claims should not be regarded as limiting the claims involved.

[0135] It should be noted that, in this article, relational terms such as "first" and "second" are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, the elements defined by the sentence "comprise a ..." do not exclude the existence of other identical elements in the process, method, article or device including the elements.

[0136] The above description is only a specific embodiment of the present invention, so that those skilled in the art can understand or implement the present invention. Various modifications to these embodiments will be apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to the embodiments shown herein, but should conform to the widest scope consistent with the principles and novel features invented herein.

Claims

1. A user portrait classification method based on deep learning, characterized in that: The method comprises: Obtaining user data of the user portrait to be classified, and preprocessing the user data to obtain target data; Collecting similar user portraits of the user portrait to be classified, and extracting similar user features of the similar user portraits; Inputting the target data and the similar user features as model data into a preset user classification model, so as to detect the temporal correlation features of the model data using a temporal memory network in the user classification model; Utilizing the attention mechanism network in the user classification model to extract the information semantic sequence of the temporal association features, and according to the information semantic sequence, utilizing the convolutional network in the user classification model to identify the data category of the model data, and determining the user category of the user portrait to be classified according to the data category; The preprocessing of the user data to obtain target data includes: Extracting a data field of each data in the user data, and identifying a data attribute of each data in the user data according to the data field; Matching the data attributes with a preset business attribute table, and taking the data corresponding to the successfully matched data attributes as the target data; The collecting of similar user portraits of the user portrait to be classified includes: Acquire user attributes of the user portrait to be classified, filter the user attributes, and construct branch attributes of the filtered user attributes; Collect historical user portraits of the branch attributes, and calculate similar user scores for each user portrait in the historical user portraits; Selecting a user portrait whose similar user score is greater than a preset score as the similar user portrait; The detecting the temporal correlation features of the model data by using the temporal memory network in the user classification model includes: Calculating the state value of the model data using the input gate in the temporal memory network, and reducing the dimension of the state value using the projection layer in the temporal memory network to obtain a reduced-dimensional state value; Calculating the activation value of the model data using the forget gate in the temporal memory network, and reducing the dimension of the activation value of the dimensionality reduction state value using the projection layer in the temporal memory network to obtain a dimensionality reduction activation value; Calculate the state update value of the model data according to the dimension reduction state value and the dimension reduction activation value; Calculating the associated feature sequence of the state update value using the output gate in the temporal memory network, and reducing the dimension of the associated feature sequence using the projection layer in the temporal memory network to obtain the temporal associated features of the model data; The extracting the information semantic sequence of the temporal association feature by using the attention mechanism network in the user classification model includes: Using the position encoding layer in the attention mechanism network to position encode the temporal association features to obtain encoded association features; Using the encoder in the attention mechanism to perform feature encoding on the encoding-related features to obtain key encoding features; Using the decoder in the attention mechanism to decode the key encoding features to obtain an information semantic sequence; The step of using the position encoding layer in the attention mechanism network to position encode the temporal association features to obtain the encoded association features includes: The following formula is used to position encode the temporal correlation feature: Among them, E (p,2x) represents the even-numbered coded correlation feature in the time series correlation feature, pos represents the position coding index, x represents the i-th feature vector in the time series correlation feature, d ml represents the feature vector dimension, E (p,2x+1) Indicates the odd-numbered coded correlation features in the temporal correlation features; The step of identifying the data category of the model data by using a convolutional network in the user classification model according to the information semantic sequence includes: According to the information semantic sequence, using the convolution layer in the convolution network to perform a convolution operation on the model data to obtain feature data; Standardizing the feature data using a standard layer in the convolutional network to obtain standard data; Calculating the data category probability of the standard data using the fully connected layer in the convolutional network, and outputting the data category of the model data using the output layer in the convolutional network according to the data category probability; The similar user portrait refers to a portrait having the same dimensional information as the user portrait to be classified; the user attribute refers to the user information dimension used to describe the user portrait to be classified, which is obtained by querying according to data attributes; the attribute branch refers to an attribute used to refine the user attributes.

2. A user portrait classification method and device based on deep learning, characterized in that: The device comprises: A data preprocessing module is used to obtain user data of the user portrait to be classified, and preprocess the user data to obtain target data; A similar user collection module is used to collect similar user portraits of the user portrait to be classified, and extract similar user features of the similar user portraits; A correlation feature detection module, used to input the target data and the similar user features as model data into a preset user classification model, so as to detect the temporal correlation features of the model data using the temporal memory network in the user classification model; A user category identification module, configured to extract the information semantic sequence of the temporal association feature by using the attention mechanism network in the user classification model, identify the data category of the model data by using the convolutional network in the user classification model according to the information semantic sequence, and determine the user category of the user portrait to be classified according to the data category; The preprocessing of the user data to obtain target data includes: Extracting a data field of each data in the user data, and identifying a data attribute of each data in the user data according to the data field; Matching the data attributes with a preset business attribute table, and taking the data corresponding to the successfully matched data attributes as the target data; The collecting of similar user portraits of the user portrait to be classified includes: Acquire user attributes of the user portrait to be classified, filter the user attributes, and construct branch attributes of the filtered user attributes; Collect historical user portraits of the branch attributes, and calculate similar user scores for each user portrait in the historical user portraits; Selecting a user portrait whose similar user score is greater than a preset score as the similar user portrait; The detecting the temporal correlation features of the model data by using the temporal memory network in the user classification model includes: Calculating the state value of the model data using the input gate in the temporal memory network, and reducing the dimension of the state value using the projection layer in the temporal memory network to obtain a reduced-dimensional state value; Calculating the activation value of the model data using the forget gate in the temporal memory network, and reducing the dimension of the activation value of the dimensionality reduction state value using the projection layer in the temporal memory network to obtain a dimensionality reduction activation value; Calculate the state update value of the model data according to the dimension reduction state value and the dimension reduction activation value; Calculating the associated feature sequence of the state update value using the output gate in the temporal memory network, and reducing the dimension of the associated feature sequence using the projection layer in the temporal memory network to obtain the temporal associated features of the model data; The extracting the information semantic sequence of the temporal association feature by using the attention mechanism network in the user classification model includes: Using the position encoding layer in the attention mechanism network to position encode the temporal association features to obtain encoded association features; Using the encoder in the attention mechanism to perform feature encoding on the encoding-related features to obtain key encoding features; Using the decoder in the attention mechanism to decode the key encoding features to obtain an information semantic sequence; The step of using the position encoding layer in the attention mechanism network to position encode the temporal association features to obtain the encoded association features includes: The following formula is used to position encode the temporal correlation feature: Among them, E (p,2x) represents the even-numbered coded correlation feature in the time series correlation feature, pos represents the position coding index, x represents the i-th feature vector in the time series correlation feature, d ml represents the feature vector dimension, E (p,2x+1) Indicates the odd-numbered coded correlation features in the temporal correlation features; The step of identifying the data category of the model data by using a convolutional network in the user classification model according to the information semantic sequence includes: According to the information semantic sequence, using the convolution layer in the convolution network to perform a convolution operation on the model data to obtain feature data; Standardizing the feature data using a standard layer in the convolutional network to obtain standard data; Calculating the data category probability of the standard data using the fully connected layer in the convolutional network, and outputting the data category of the model data using the output layer in the convolutional network according to the data category probability; The similar user portrait refers to a portrait having the same dimensional information as the user portrait to be classified; the user attribute refers to the user information dimension used to describe the user portrait to be classified, which is obtained by querying according to data attributes; the attribute branch refers to an attribute used to refine the user attributes.

3. An electronic device, characterized in that: The electronic device comprises: at least one processor; and, a memory communicatively connected to the at least one processor; wherein, The memory stores a computer program that can be executed by the at least one processor, and the computer program is executed by the at least one processor so that the at least one processor can execute the user portrait classification method based on deep learning as described in claim 1.

4. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the user portrait classification method based on deep learning as described in claim 1 is implemented.

Images