User interest mining method and device, electronic equipment and readable storage medium
By constructing long-term and short-term interest models and performing quantum state fusion, the interest accuracy problem caused by the overall modeling of medium- and long-term and short-term behavior data in the existing technology is solved, and more accurate user interest mining and advertising delivery are achieved.
Patent Information
- Application Number
- CN202510394401.8
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-03-31
- Publication Date
- 2025-08-29
AI Technical Summary
In the existing user interest mining system, the overall modeling of long-term behavior data and short-term behavior data leads to excessive attention to short-term dynamic changes, ignoring the stability of long-term interests, and affecting the accuracy of user interest mining and the accuracy of advertising delivery.
The user's long-term interest model and short-term interest model are respectively constructed, and the target interest quantum state is generated through quantum state encoding and quantum state superposition and fusion, and quantum measurements are performed to obtain the target interest characteristics that characterize the user's current real interests.
It improves the accuracy of exploring users' real interests, improves the accuracy of advertising delivery, and effectively combines long-term and short-term interests through quantum computing to capture the complex relationship between users.
Smart Images
Figure CN120563174A_ABST
Abstract
Description
Technical Field
[0001] The present application belongs to the field of artificial intelligence technology, and specifically relates to a user interest mining method, device, electronic device and readable storage medium. Background Art
[0002] In existing user interest mining systems, user interest modeling is a key step to achieve good user interest mining results. However, existing user interest mining systems mostly use a holistic modeling approach to obtain an overall user interest model and mine user interests based on the overall user interest model.
[0003] In the process of realizing this application, the inventor discovered that the existing technology exists, which models the user's long-term interests and short-term interests as a whole. This method ignores the fact that long-term behavioral data is usually preferences or hobbies, and usually has a strong time span and stability, while short-term behavioral data is often dynamically affected by the context. In this case, when long-term behavioral data and short-term behavioral data are modeled as a whole, the model may pay too much attention to the dynamic changes of short-term behavior and ignore the stability of long-term interests, thereby affecting the accuracy of mining the user's real interests, and thus reducing the accuracy of advertising to users. Summary of the Invention
[0004] The purpose of the embodiments of the present application is to provide a user interest mining method, device, electronic device and readable storage medium, which can solve the problem of overall modeling of long-term behavior data and short-term behavior data. The model may focus too much on the dynamic changes of short-term behavior and ignore the stability of long-term interests, thereby affecting the accuracy of mining the user's real interests.
[0005] In a first aspect, an embodiment of the present application provides a method for mining user interests, the method comprising:
[0006] Obtain users' long-term and short-term behavior data;
[0007] Constructing a long-term interest model of the user based on the long-term behavior data, and constructing a short-term interest model of the user based on the short-term behavior data;
[0008] Performing quantum state encoding on the features of the long-term interest model and the short-term interest model respectively to obtain a long-term interest quantum state and a short-term interest quantum state;
[0009] Based on quantum state superposition and quantum gates, the long-term quantum state of interest and the short-term quantum state of interest are merged to generate a target quantum state of interest;
[0010] A quantum measurement is performed on the target interest quantum state to obtain a target interest feature, where the target interest feature is used to characterize the current real interest of the user.
[0011] In one possible implementation, based on quantum state superposition and quantum gates, the long-term quantum state of interest and the short-term quantum state of interest are fused to generate a target quantum state of interest, including:
[0012] Based on the superposition of the quantum states, the long-term quantum state of interest and the short-term quantum state of interest are fused to obtain a fused quantum state;
[0013] Based on a quantum gate, the fused quantum state is adjusted to obtain the target quantum state of interest.
[0014] In one possible implementation, constructing the user's short-term interest model based on the short-term behavior data includes:
[0015] Aggregating the short-term behavior data into a short-term behavior sequence of the user based on a dynamic time window;
[0016] Based on the user's short-term behavior sequence, a multi-head self-attention mechanism is used to generate the user's short-term interest model.
[0017] In one possible implementation, based on the user's short-term behavior sequence, a multi-head self-attention mechanism is used to generate the user's short-term interest model, including:
[0018] Using a multi-head self-attention mechanism to dynamically assign attention weights to the user's short-term behavior sequence;
[0019] A short-term interest model of the user is obtained according to the attention weight and the correlation between the short-term behaviors in the short-term behavior sequence of the user.
[0020] In a possible implementation, before aggregating the short-term behavior data into the short-term behavior sequence of the user based on the dynamic time window, the method includes:
[0021] Obtaining a log of the user, wherein the log of the user includes multiple time periods and the user activity corresponding to each time period;
[0022] The dynamic time window is determined according to the user activity.
[0023] In one possible implementation, building the long-term interest model of the user based on the long-term behavior data includes:
[0024] Clustering the long-term behavior data to obtain a first feature matrix;
[0025] Orthogonalization constraints are performed on the first feature matrix to obtain the long-term interest model.
[0026] In a possible implementation, clustering is performed on the long-term behavior data to obtain a first feature matrix, including:
[0027] Construct the interest cluster based on the user's long-term behavior data
[0028] Applying a sparsity constraint to the feature vector of each of the interest clusters;
[0029] Normalize the feature vectors of the interest cluster to obtain the first feature matrix.
[0030] In a second aspect, an embodiment of the present application provides a user interest mining device, the device comprising:
[0031] The acquisition module is used to obtain the user's long-term behavior data and short-term behavior data;
[0032] A construction module, configured to construct a long-term interest model of the user based on the long-term behavior data, and to construct a short-term interest model of the user based on the short-term behavior data;
[0033] An encoding module, configured to perform quantum state encoding on the features of the long-term interest model and the short-term interest model, respectively, to obtain a long-term interest quantum state and a short-term interest quantum state;
[0034] A first fusion module is configured to fuse the long-term quantum state of interest and the short-term quantum state of interest based on quantum state superposition and quantum gates to generate a target quantum state of interest;
[0035] The measurement module is used to perform quantum measurement on the target interest quantum state to obtain a target interest feature, where the target interest feature is used to characterize the user's current real interest.
[0036] In a third aspect, an embodiment of the present application provides an electronic device comprising a processor, a memory, and a program stored in the memory and executable on the processor, wherein the program, when executed by the processor, implements the steps of the method described in the first aspect.
[0037] In a fourth aspect, an embodiment of the present application provides a computer-readable storage medium having a computer program stored thereon, and when the computer program is executed by a processor, the steps of the method described in the first aspect are implemented.
[0038] In an embodiment of the present application, by modeling the user's long-term interests and short-term interests separately, the overall modeling of the user's long-term behavior data and short-term behavior data is avoided, and the dynamic changes of short-term behavior data affected by the context and the stability of long-term behavior data as hobbies are fully taken into account. The long-term interest features extracted from the long-term interest model and the short-term interest features extracted from the short-term interest model are fused to obtain the target interest features to represent the user's current real interests, thereby improving the accuracy of mining the user's real interests and further improving the accuracy of advertising to the user. BRIEF DESCRIPTION OF THE DRAWINGS
[0039] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the following briefly introduces the drawings required for use in the description of the embodiments of the present application. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without any creative work.
[0040] Figure 1 A flowchart of a method for mining user interests provided in an embodiment of the present application;
[0041] Figure 2 A schematic diagram of the structure of a user interest mining device provided in an embodiment of the present application;
[0042] Figure 3 A schematic diagram of the structure of an electronic device provided in an embodiment of the present application. DETAILED DESCRIPTION
[0043] The following will be combined with the accompanying drawings in the embodiments of this application to clearly describe the technical solutions in the embodiments of this application. Obviously, the embodiments described are part of the embodiments of this application, not all of the embodiments. Based on the embodiments in this application, all other embodiments obtained by ordinary technicians in this field are within the scope of protection of this application.
[0044] The terms "first," "second," and the like in the specification and claims of this application are used to distinguish similar objects, and are not used to describe a specific order or precedence. It should be understood that the terms used in this manner are interchangeable where appropriate, so that the embodiments of this application can be implemented in an order other than that illustrated or described herein, and that the objects distinguished by "first," "second," and the like are generally of the same type, and do not limit the number of objects; for example, the first object can be one or more. In addition, the term "and / or" in the specification and claims refers to at least one of the connected objects, and the character " / " generally indicates that the objects connected are in an "or" relationship.
[0045] The method provided in the embodiment of the present application is described in detail below through specific embodiments and their application scenarios in conjunction with the accompanying drawings.
[0046] The embodiments of the present application are applicable to the field of advertising delivery.
[0047] like Figure 1 As shown, an embodiment of the present application provides a user interest mining method, which includes:
[0048] Step 101: Obtain the user's long-term behavior data and short-term behavior data;
[0049] Step 102: constructing a long-term interest model of the user based on the long-term behavior data, and constructing a short-term interest model of the user based on the short-term behavior data;
[0050] Step 103: Perform quantum state encoding on the features of the long-term interest model and the short-term interest model respectively to obtain a long-term interest quantum state and a short-term interest quantum state;
[0051] Step 104: Based on quantum state superposition and quantum gates, the long-term quantum state of interest and the short-term quantum state of interest are merged to generate a target quantum state of interest;
[0052] Step 105: Perform quantum measurement on the target interest quantum state to obtain target interest features, where the target interest features are used to characterize the user's current real interest.
[0053] It should be understood that long-term behavior data refers to a set of records of historical behaviors generated by users over a longer time span, which has temporal, sparse and diverse characteristics. For example, the longer time span may be greater than 6 months.
[0054] It should be understood that short-term behavior data refers to a set of records of historical behaviors generated by users within a relatively short time span, which has temporal, sparse and diverse characteristics. For example, the relatively short time span may be less than 6 months.
[0055] The specific implementation process of the embodiment of this application is as follows:
[0056] Assume that the long-term interest characteristics and short-term interest characteristics are: long-term interest characteristics: [0.8, 0.6]; short-term interest characteristics: [0.4, 0.9]. It should be understood that in the long-term interest characteristics: [0.8, 0.6], 0.8 refers to the interest value for electronic products, which is relatively high, while 0.6 refers to the interest value for furniture users, which is slightly high. In the short-term interest characteristics: [0.4, 0.9], 0.4 refers to the interest value for maternal and child products, which is slightly low, and 0.9 refers to the interest value for clothing, which is extremely high.
[0057] Furthermore, the features of the long-term interest model and the short-term interest model are respectively quantum-encoded to obtain the long-term interest quantum state and the short-term interest quantum state as follows:
[0058] Long-term quantum state of interest: |ψlong>=0.8|0>+0.6|1>;
[0059] Short-term quantum state of interest: |ψshort>=0.4|0>+0.9|1>;
[0060] Furthermore, the long-term interest and the short-term interest are fused by quantum state superposition: |ψfused>=α|ψlong>+β|ψshort>; α is the weight coefficient of the long-term interest quantum state, and β is the weight coefficient of the short-term interest quantum state;
[0061] Assuming α = 0.7, β = 0.3, then: |ψfused> = 0.7(0.8|0> + 0.6|1>) + 0.3(0.4|0> + 0.9|1>); ψfused is the fused quantum state of long-term interest and short-term interest; after expansion: |ψfused> = 0.68|0> + 0.69|1>;
[0062] Furthermore, quantum gate operations are applied to the fused quantum state |ψfused>.
[0063] Exemplarily, a Hadamard gate H is used:
[0064] H|0>=12(|0>+|1>), H|1>=12(|0>-|1>) Apply Hadamard gate to |ψfused>|ψfused>: H|ψfused>=0.6812(|0>+|1>)+0.6912(|0>-|1>)H|ψfused>=0.682 1(|0>+|1>)+0.692 1(|0>-|1>)
[0065] After expansion: H|ψfused>=(0.68+0.69)|0>+12(0.68-0.69)|1>;
[0066] H|ψfused>=0.97|0>-0.01|1>;
[0067] Furthermore, the final optimized quantum state |ψoptimized> is measured to obtain the classical feature vector = [0.97, -0.01], that is, the target interest feature = [0.97, -0.01];
[0068] It should be understood that the target interest feature = [0.97, -0.01] here means that electronic product clothing can be recommended to the user, such as self-heating jackets, which the user may be extremely interested in; and avoid pushing maternal and child furniture products to the user, which the user may be extremely disgusted with, such as cribs.
[0069] In this embodiment, quantum state superposition is the first step, used to initially fuse long-term and short-term interests. Quantum gate operations are the second step, used to optimize or nonlinearly transform the fused state. Thus, quantum state superposition provides linear fusion capabilities, while quantum gate operations add nonlinear transformation capabilities. The combination of these two makes feature fusion more flexible and efficient. This sequential design can fully leverage the advantages of quantum computing to capture the complex relationship between long-term and short-term interests, thereby improving the accuracy of user interest mining.
[0070] In an embodiment of the present application, by modeling the user's long-term interests and short-term interests separately, the overall modeling of the user's long-term behavior data and short-term behavior data is avoided, and the dynamic changes of short-term behavior data affected by the context and the stability of long-term behavior data as hobbies are fully taken into account. The long-term interest features extracted from the long-term interest model and the short-term interest features extracted from the short-term interest model are fused to obtain the target interest features to represent the user's current real interests, thereby improving the accuracy of mining the user's real interests and further improving the accuracy of advertising to the user.
[0071] Optionally, in some embodiments, based on quantum state superposition and quantum gates, the long-term quantum state of interest and the short-term quantum state of interest are fused to generate a target quantum state of interest, including:
[0072] Based on the superposition of the quantum states, the long-term quantum state of interest and the short-term quantum state of interest are fused to obtain a fused quantum state;
[0073] Based on a quantum gate, the fused quantum state is adjusted to obtain the target quantum state of interest.
[0074] In an embodiment of the present application, the features are initially fused through quantum state superposition, and the fused quantum states are further optimized through quantum gates, so as to capture the complex interactive relationship between long-term interests and short-term interests, and thus more accurately capture the user's current real interests.
[0075] Optionally, in some embodiments, constructing the user's short-term interest model based on the short-term behavior data includes:
[0076] Aggregating the short-term behavior data into a short-term behavior sequence of the user based on a dynamic time window;
[0077] Based on the user's short-term behavior sequence, a multi-head self-attention mechanism is used to generate the user's short-term interest model.
[0078] In the embodiments of the present application, the user's short-term behavior characteristics are effectively captured through a dynamic time window method and a multi-head self-attention mechanism to generate a high-quality short-term interest model, which has the advantages of flexibility, accuracy, efficiency and personalization. It can be widely used in scenarios that require short-term real-time user interest modeling, thereby improving the intelligence level of the advertising recommendation system and user experience.
[0079] Optionally, in some embodiments, based on the user's short-term behavior sequence, a multi-head self-attention mechanism is used to generate the user's short-term interest model, including:
[0080] Using a multi-head self-attention mechanism to dynamically assign attention weights to the user's short-term behavior sequence;
[0081] A short-term interest model of the user is obtained according to the attention weight and the correlation between the short-term behaviors in the short-term behavior sequence of the user.
[0082] In the embodiment of the present application, through the above-mentioned method, due to the use of a multi-head self-attention mechanism and support for parallel computing, the modeling efficiency is improved. Through the combination of attention weight and relevance, a short-term interest model that is more accurately in line with user interests can be obtained, thereby improving the user experience.
[0083] Optionally, in some embodiments, before aggregating the short-term behavior data into the user's short-term behavior sequence based on the dynamic time window, the process includes:
[0084] Obtaining a log of the user, wherein the log of the user includes multiple time periods and the user activity corresponding to each time period;
[0085] The dynamic time window is determined according to the user activity.
[0086] It should be understood that the formula for determining the dynamic time window according to the user activity is as follows:
[0087]
[0088] The W adapt Used to characterize the dynamic time window, W base Used to characterize the preset basic time window length, N 7d Used to represent the user's activity within 7 seconds, N avg Used to represent the average activity of users across the entire site.
[0089] It should be understood that user activity refers to the total number of user behaviors in a certain period of time in the past. These behaviors may include key behaviors such as clicking on products and adding products to shopping carts; the average activity of all users on the entire site refers to the average daily number of behaviors of all users on the site in a certain period of time in the past (cross-user statistics).
[0090] Application scenario examples:
[0091] User Type N7D Navg Wbase Wadapt Highly active users 300 50 1h 1.8h Ordinary users 80 50 1h 1.2h Silent User 5 50 1h 0.3h
[0092] In an embodiment of the present application, through the above method, the window can be enlarged for active users and the window can be shrunk for low-activity users, so as to dynamically adjust the size of the time window to meet the user behavior data with a certain amount of data in each time window, thereby facilitating the aggregation of user behavior data.
[0093] Optionally, in some embodiments, constructing the user's long-term interest model based on the long-term behavior data includes:
[0094] Clustering the long-term behavior data to obtain a first feature matrix;
[0095] Orthogonalization constraints are performed on the first feature matrix to obtain the long-term interest model.
[0096] In an embodiment of the present application, the solution constructs a long-term interest model of the user by clustering and orthogonalizing the long-term behavioral data. In this way, clustering can extract core interest features, orthogonalization constraints can reduce redundancy, and the generated long-term interest model is more accurate; at the same time, clustering dimensionality reduction and orthogonalization constraints improve computing efficiency, which is suitable for processing large-scale data; and the clustering results naturally form interest categories, and the model has stronger interpretability; and orthogonalization constraints enhance the stability and generalization ability of the model, which is suitable for a variety of scenarios.
[0097] Optionally, in some embodiments, clustering the long-term behavior data to obtain a first feature matrix includes:
[0098] Construct the interest cluster based on the user's long-term behavior data
[0099] Applying a sparsity constraint to the feature vector of each of the interest clusters;
[0100] Normalize the feature vectors of the interest cluster to obtain the first feature matrix.
[0101] In the embodiment of the present application, accuracy can be achieved through the above-mentioned methods: clustering extracts core interest categories, sparsity constraints highlight key features, normalization eliminates feature scale differences, and the generated feature matrix is more accurate.
[0102] like Figure 2As shown, the embodiment of the present application further provides a user interest mining device, and the user interest mining device 200 includes:
[0103] Acquisition module 201, used to acquire the user's long-term behavior data and short-term behavior data;
[0104] A construction module 202 is configured to construct a long-term interest model of the user based on the long-term behavior data, and to construct a short-term interest model of the user based on the short-term behavior data;
[0105] An encoding module 203 is configured to perform quantum state encoding on the features of the long-term interest model and the short-term interest model, respectively, to obtain a long-term interest quantum state and a short-term interest quantum state;
[0106] A first fusion module 204 is configured to fuse the long-term quantum state of interest and the short-term quantum state of interest based on quantum state superposition and quantum gates to generate a target quantum state of interest;
[0107] The measurement module 205 is configured to perform quantum measurement on the target interest quantum state to obtain a target interest feature, where the target interest feature is used to represent the current real interest of the user.
[0108] Optionally, the first fusion module 204 further includes:
[0109] A first fusion submodule is configured to fuse the long-term quantum state of interest and the short-term quantum state of interest based on the superposition of the quantum states to obtain a fused quantum state;
[0110] The adjustment submodule adjusts the fused quantum state based on the quantum gate to obtain the target quantum state of interest.
[0111] Optionally, the building module 202 further includes:
[0112] an aggregation submodule, configured to aggregate the short-term behavior data into a short-term behavior sequence of the user based on a dynamic time window;
[0113] The first generation submodule is used to generate a short-term interest model of the user based on the user's short-term behavior sequence using a multi-head self-attention mechanism.
[0114] Optionally, the first generation submodule further includes:
[0115] an allocation unit, configured to dynamically allocate attention weights to the user's short-term behavior sequence using a multi-head self-attention mechanism;
[0116] The first obtaining unit is used to obtain the short-term interest model of the user according to the correlation between the attention weight and the short-term behavior in the short-term behavior sequence of the user.
[0117] Optionally, the building module 202 further includes:
[0118] An acquisition submodule, configured to acquire the user's log, wherein the user's log includes multiple time periods and the user activity corresponding to each time period;
[0119] A determination submodule is configured to determine the dynamic time window according to the user activity.
[0120] Optionally, the building block 202 further includes:
[0121] A clustering submodule, configured to cluster the long-term behavior data to obtain a first feature matrix;
[0122] The constraint submodule is used to perform orthogonalization constraints on the first feature matrix to obtain the long-term interest model.
[0123] Optionally, the clustering submodule includes:
[0124] A construction unit, configured to construct the interest cluster based on the user's long-term behavior data
[0125] a constraint unit, configured to impose a sparsity constraint on the feature vector of each of the interest clusters;
[0126] The second obtaining unit is configured to perform normalization processing on the feature vector of the interest cluster to obtain the first feature matrix.
[0127] It should be noted that the user interest mining device provided in the embodiment of the present application can achieve the following Figure 1 The entire technical process of the user interest mining method shown in the embodiment achieves the same technical effect, and will not be described again here to avoid repetition.
[0128] The embodiment of the present application also provides an electronic device, comprising: a processor, a memory, and a program stored in the memory and capable of running on the processor, wherein the program is executed by the processor to achieve the above Figure 1 The various processes of the method embodiment shown can achieve the same technical effect, and to avoid repetition, they will not be described here.
[0129] For details, see Figure 3 As shown, an embodiment of the present application further provides an electronic device, including a bus 301 , a transceiver 302 , an antenna 303 , a bus interface 304 , a processor 305 and a memory 306 .
[0130] In this embodiment, the electronic device further includes: a computer program stored in the memory 306 and executable on the processor 305. When the computer program is executed by the processor 305, the following can be achieved: Figure 1 The various processes of the user interest mining method shown in the embodiment can achieve the same technical effect, and will not be described again here to avoid repetition.
[0131] exist Figure 3 In the figure, the bus architecture (represented by bus 301) can include any number of interconnected buses and bridges. Bus 301 links various circuits including one or more processors represented by processor 305 and memory represented by memory 306. Bus 301 can also link various other circuits such as peripheral devices, voltage regulators, and power management circuits, which are well known in the art and are therefore not described further herein. Bus interface 304 provides an interface between bus 301 and transceiver 302. Transceiver 302 can be one element or multiple elements, such as multiple receivers and transmitters, providing a unit for communicating with various other devices on a transmission medium. Data processed by processor 305 is transmitted on a wireless medium via antenna 303. Furthermore, antenna 303 also receives data and transmits the data to processor 305.
[0132] The processor 305 is responsible for managing the bus 301 and general processing, and may also provide various functions, including timing, peripheral interfaces, voltage regulation, power management, and other control functions. The memory 306 may be used to store data used by the processor 305 when performing operations.
[0133] Optionally, the processor 305 may be a CPU, an ASIC, an FPGA, or a CPLD.
[0134] The present application also provides a computer-readable storage medium having a computer program stored thereon. When executed by a processor, the computer program implements the various processes of the above-described user interest mining method embodiment and achieves the same technical effects. To avoid repetition, the description is omitted here. The computer-readable storage medium may be, for example, ROM, RAM, a magnetic disk, or an optical disk.
[0135] It should be noted that, in this document, the terms "comprises," "includes," or any other variations thereof are intended to encompass non-exclusive inclusion, such that a process, method, article, or apparatus comprising a series of elements includes not only those elements but also other elements not explicitly listed, or elements inherent to such process, method, article, or apparatus. In the absence of further limitations, an element defined by the phrase "comprising a ..." does not exclude the presence of other identical elements in the process, method, article, or apparatus comprising the element.
[0136] Through the description of the above implementation methods, those skilled in the art can clearly understand that the above-mentioned embodiment methods can be implemented by means of software plus the necessary general hardware platform, and of course can also be implemented by hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of the present application, or the part that contributes to the prior art, can be embodied in the form of a software product, which is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk), and includes a number of instructions for enabling a terminal (which can be a mobile phone, computer, server, air conditioner, or network device, etc.) to execute the methods described in each embodiment of the present application.
[0137] The embodiments of the present application are described above in conjunction with the accompanying drawings, but the present application is not limited to the above-mentioned specific implementation methods. The above-mentioned specific implementation methods are merely illustrative and not restrictive. Under the guidance of this application, ordinary technicians in this field can also make many forms without departing from the purpose of this application and the scope of protection of the claims, all of which are within the protection of this application.
Claims
1. A user interest mining method, characterized in that: The method comprises: Obtain users' long-term and short-term behavior data; Constructing a long-term interest model of the user based on the long-term behavior data, and constructing a short-term interest model of the user based on the short-term behavior data; Performing quantum state encoding on the features of the long-term interest model and the short-term interest model respectively to obtain a long-term interest quantum state and a short-term interest quantum state; Based on quantum state superposition and quantum gates, the long-term quantum state of interest and the short-term quantum state of interest are merged to generate a target quantum state of interest; A quantum measurement is performed on the target interest quantum state to obtain a target interest feature, where the target interest feature is used to characterize the current real interest of the user.
2. The method according to claim 1, characterized in that Based on quantum state superposition and quantum gates, the long-term quantum state of interest and the short-term quantum state of interest are merged to generate a target quantum state of interest, including: Based on the superposition of the quantum states, the long-term quantum state of interest and the short-term quantum state of interest are fused to obtain a fused quantum state; Based on a quantum gate, the fused quantum state is adjusted to obtain the target quantum state of interest.
3. The method according to claim 1, characterized in that Constructing a short-term interest model of the user based on the short-term behavior data includes: Aggregating the short-term behavior data into a short-term behavior sequence of the user based on a dynamic time window; Based on the user's short-term behavior sequence, a multi-head self-attention mechanism is used to generate the user's short-term interest model.
4. The method according to claim 3, characterized in that Based on the user's short-term behavior sequence, a multi-head self-attention mechanism is used to generate the user's short-term interest model, including: Using a multi-head self-attention mechanism to dynamically assign attention weights to the user's short-term behavior sequence; A short-term interest model of the user is obtained according to the attention weight and the correlation between the short-term behaviors in the short-term behavior sequence of the user.
5. The method according to claim 3, characterized in that Before aggregating the short-term behavior data into the short-term behavior sequence of the user based on the dynamic time window, the method includes: Obtaining a log of the user, wherein the log of the user includes multiple time periods and the user activity corresponding to each time period; The dynamic time window is determined according to the user activity.
6. The method according to claim 1, characterized in that Constructing a long-term interest model of the user based on the long-term behavior data includes: Clustering the long-term behavior data to obtain a first feature matrix; Orthogonalization constraints are performed on the first feature matrix to obtain the long-term interest model.
7. The method according to claim 6, characterized in that Clustering the long-term behavior data to obtain a first feature matrix, including: Construct the interest cluster based on the user's long-term behavior data Applying a sparsity constraint to the feature vector of each of the interest clusters; Normalize the feature vectors of the interest cluster to obtain the first feature matrix.
8. A user interest mining device, characterized in that: The device comprises: The acquisition module is used to obtain the user's long-term behavior data and short-term behavior data; A construction module, configured to construct a long-term interest model of the user based on the long-term behavior data, and to construct a short-term interest model of the user based on the short-term behavior data; An encoding module, configured to perform quantum state encoding on the features of the long-term interest model and the short-term interest model, respectively, to obtain a long-term interest quantum state and a short-term interest quantum state; A first fusion module is configured to fuse the long-term quantum state of interest and the short-term quantum state of interest based on quantum state superposition and quantum gates to generate a target quantum state of interest; The measurement module is used to perform quantum measurement on the target interest quantum state to obtain a target interest feature, where the target interest feature is used to characterize the user's current real interest.
9. An electronic device, characterized in that: include: A processor, a memory, and a program stored in the memory and executable on the processor, wherein when the program is executed by the processor, the steps of the user interest mining method according to any one of claims 1 to 7 are implemented.
10. A computer-readable storage medium, characterized in that The computer-readable storage medium stores a computer program, which, when executed by a processor, implements the steps of the user interest mining method according to any one of claims 1 to 7.