Method, System, Device and Medium for Adaptive Generation of Power Communication Planning Indexes

Abstract

The present invention discloses a method, system, device and medium for adaptively generating power communication planning indicators, which recommend different indicator contents according to user types; explicitly and implicitly feedback indicators based on user behavior records, perform indicator fusion through indicator feedback to determine an initial scoring matrix of the user for the indicator template content; determine a similarity matrix between indicator templates according to the cosine similarity between indicator templates; and form a user scoring matrix after collaborative filtering based on the scoring matrix and the similarity matrix. According to the user scoring matrix and the important indicator frequency of the indicator template, determine the set of important indicators in the indicator template, use the important indicators in the set as the input of the network model, determine the associated indicator templates derived from the important indicators, and form an architecture model for generating power communication planning indicator templates that mutually cooperate among high-frequency indicator recommendation, collaborative filtering recommendation and important indicator recommendation.

Description

Technical Field

[0001] The present invention belongs to the technical field of power communication systems, and particularly relates to a method, a system, a device and a medium for adaptively generating power communication planning indicators. Background Art

[0002] With the advent of the era of "cloud, big data, Internet of Things, mobility, and intelligence", the current situation of the power system generating a large amount of data has long become the norm. All links of the power system can easily obtain high-quality information content through information processing technologies and tools. However, due to the huge amount of information, it has become very difficult to obtain the accurate information required by users by means of information processing and the like. And due to the different focuses and data selection methods of different systems, therefore, meeting the professional needs of specific power system links is a key problem that needs to be solved urgently.

[0003] In order to solve the selection of massive data indicators, for the intelligent planning of power communication, a recommendation strategy based on meeting user needs has emerged as the times require. The automatic generation strategy in the selection of planning indicators is a specific professional and application requirement solution with good compatibility and reliability. Information technologies such as big data, cloud computing, Internet of Things, and mobile Internet will become important management means. The power communication network faces a new form of rapid development and responds to constantly changing new requirements. Scientific planning is a necessary means for the sustainable and healthy development of the power communication network. The construction of the power communication network is a partial construction design and focuses more on the overall application design. Therefore, the key to the planning of the power communication network lies in the word "selection", selecting technologies, resources, network architectures, and making local construction adjustments. For the concerns and construction needs of different planning objects, by analyzing and mining the potential behaviors of users' planning indicator selections, appropriate planning indicators will be derived and pushed to users, so as to set a relatively precise demand solution from the whole to the part for them.

[0004] Given that there are many ways to recommend data or user content on the Internet. Among them, the collaborative filtering recommendation method is relatively widespread. Collaborative filtering is to evaluate the resource information by users and find a set of similar users or resources. However, the sparsity and cold start problems faced by the collaborative filtering method are still the key problems restricting its development. Summary of the Invention

[0005] Aiming at the sparsity and cold start problems existing in the traditional collaborative filtering method, the present invention provides a method, a system, a device and a medium for adaptively generating power communication planning indicators. By introducing high-frequency recommendation, the occurrence of the cold start problem can be reduced, and at the same time, through the method of important indicator recommendation, the reliability of the indicator recommendation system can be more accurately improved.

[0006] To achieve the above object, the present invention adopts the following technical solutions:

[0007] In a first aspect of the present invention, there is provided a method for adaptively generating power communication planning indicators based on user behavior, including:

[0008] Obtaining historical data information of users to determine user types, and recommending different indicator contents according to the user types to obtain user selection feedback;

[0009] Recording the selection of explicit indicators and implicit indicators according to the user selection feedback, fusing the explicit indicators and implicit indicators, and determining an initial scoring matrix of the user for the content of the indicator template; calculating the cosine similarity between the indicator templates to determine a similarity matrix between the indicator templates;

[0010] Determining a user scoring matrix after collaborative filtering according to the scoring matrix and the similarity matrix;

[0011] Determining a set of important indicators in the indicator template according to the important indicator frequency of the indicator template and the user scoring matrix; using the set of important indicators as input for network training to determine associated indicator templates for the important indicators; performing semantic fusion on the associated indicator templates to finally determine the recommended indicators.

[0012] As an optional embodiment of the present invention, the explicit indicators include the selection frequency C of the content of the indicator template ij and the expert score R of the importance of the indicator ij ; the implicit indicator is the usage duration T of the indicator system ij .

[0013] As an optional embodiment of the present invention, the selection frequency C of the content of the indicator template ij =n, where n represents the frequency of the user selecting the planning indicator; the expert score of the importance of the indicator m represents the satisfaction degree of the score, and m ∈ (1, 10).

[0014] As an optional embodiment of the present invention, the usage duration of the indicator system is the average speed of the user's historical browsing, and L represents the total length of the corresponding read indicator template.

[0015] As an optional embodiment of the present invention, the initial scoring matrix P:

[0016]

[0017] P ij is the comprehensive score after fusion of user i for indicator template j.

[0018] As an alternative embodiment of the present invention, the cosine similarity between the metric templates is calculated to determine the similarity matrix between the metric templates, and the user rating matrix after collaborative filtering is determined based on the scoring matrix and the similarity matrix, specifically as follows:

[0019] Similarity N(i) represents the total number of people for rating metric i, and N(j) represents the total number of people for rating metric j;

[0020] Similarity matrix

[0021] Based on the similarity between the metric templates and the user initial rating matrix, the attention rating P' of the user to any metric template is determined as P' = P * B.

[0022] As an alternative embodiment of the present invention, the specific method for determining the associated metric templates of important metrics is as follows:

[0023] Obtain the metric templates with the top p ratings as the important metric k;

[0024] Calculate the important metric frequency of the metric template where f k represents the number of times the important metric k appears, and ∑f is the total number of all important metrics that appear in the metric template;

[0025] The importance rating of the important metric k in the metric template j by user i is expressed as where P ij ' is the attention rating of user i to metric template j, and w k is the frequency of the important metric k appearing in metric template j;

[0026] According to the importance rating Sort in descending order, select the top m important metrics as the input for the neural language model for BP model training, and finally determine the associated metric content of the important metrics.

[0027] In the second aspect of the present invention, a system for the method of adaptively generating power communication planning metrics based on user behavior is provided, including:

[0028] A user selection feedback acquisition module, configured to acquire the historical data information of the user to determine the user type, and recommend different metric contents according to the user type to obtain the user selection feedback;

[0029] A first matrix acquisition module, configured to record the explicit metric and implicit metric selections according to the user selection feedback, perform metric fusion on the explicit metric and the implicit metric, and determine the initial rating matrix of the user for the metric template content; calculate the cosine similarity between the metric templates, and determine the similarity matrix between the metric templates;

[0030] A second matrix acquisition module, configured to determine a user rating matrix after collaborative filtering according to the scoring matrix and the similarity matrix;

[0031] A recommendation metric module, configured to determine a set of important metrics in a metric template according to the frequency of important metrics in the metric template and the user rating matrix; use the set of important metrics as input for network training to determine an associated metric template for the important metrics; perform semantic fusion on the associated metric template, and finally determine a recommendation metric.

[0032] In a third aspect of the present invention, there is provided a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the computer program, the power communication planning metric adaptive generation method based on user behavior as described above is implemented.

[0033] In a fourth aspect of the present invention, there is provided a computer-readable storage medium storing a computer program, wherein when the computer program is executed by a processor, the power communication planning metric adaptive generation method based on user behavior as described above is implemented.

[0034] The beneficial effects of the present invention are as follows:

[0035] 1) The power communication planning metric adaptive generation method provided by the embodiments of the present invention adopts a method of metric derivation and fusion, uses explicit and implicit metrics as evaluation criteria, and performs fusion by selecting characteristics such as frequency, expert scoring of metric importance, and system usage duration, so as to obtain the user's evaluation of metric selection. On the basis of the recommendation of the traditional collaborative filtering method, the recommendation of important metrics and the recommendation of high-frequency metrics are incorporated, effectively improving the accuracy and extensibility of the collaborative filtering algorithm, reducing the cold start problem of metric generation recommendation, and effectively improving the accuracy and extensibility of the metric recommendation system.

[0036] 2) The power communication planning metric adaptive generation method provided by the embodiments of the present invention searches for new interest preferences of users by means of high-frequency metric template content. The derivative recommendation strategy adds high-frequency metric content to the system, selects the top n high-frequency metrics and adds them to the metric recommendation list, and alleviates the impact brought by the cold start problem by recommending high-frequency metrics to users. Description of the Drawings

[0037] The specification drawings constituting a part of this application are used to provide a further understanding of the present invention. The schematic embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute an improper limitation to the present invention. In the drawings:

[0038] Figure 1It is a flowchart of the method for adaptively generating power communication planning indicators in an embodiment of the present invention.

[0039] Figure 2 Simulation comparison graph of the recall rate of the collaborative filtering algorithm based on indicator fusion and the CF algorithm in an embodiment of the present invention.

[0040] Figure 3 Simulation comparison graph of the accuracy rate of the collaborative filtering algorithm based on indicator fusion and the CF algorithm in an embodiment of the present invention.

[0041] Figure 4 It is a comparison graph of the accuracy rate of the recommended association indicators under the neural network training in an embodiment of the present invention. Specific implementation manners

[0042] The present invention will be described in detail below with reference to the accompanying drawings and in conjunction with embodiments. It should be noted that, without conflict, the embodiments in the present application and the features in the embodiments can be combined with each other.

[0043] The following detailed descriptions are all exemplary descriptions, aiming to provide further detailed descriptions of the present invention. Unless otherwise specified, all technical terms adopted by the present invention have the same meaning as commonly understood by those of ordinary skill in the art to which the present application belongs. The terms used in the present invention are only for describing specific implementation manners, and are not intended to limit the exemplary embodiments according to the present invention.

[0044] As Figure 1 shown, in the first aspect of the embodiment of the present invention, a method for adaptively generating power communication planning indicators based on user behavior is provided, and the specific process is as follows:

[0045] S1: Determine the user type according to historical data information, and recommend different indicator contents according to the user type.

[0046] For new users, recommend the high-frequency indicators as the user's indicator selection;

[0047] For old users, obtain the user preferences through historical data information, and recommend the user preference indicators as the user's indicator selection.

[0048] S2: Explicitly and implicitly select indicators according to the user behavior record, perform indicator fusion through the indicator selection, and determine the initial score and the initial score matrix P of the user for the indicator template content.

[0049] S3: Determine the similarity matrix B between indicator templates according to the cosine similarity between indicator templates, and determine the user score matrix P' after collaborative filtering according to the score matrix and the similarity matrix.

[0050] S4: According to the user preference feedback and the change of focus, add a decay function to reduce the recommendation of low-interest metric templates.

[0051] S5: According to the user rating matrix and the frequency w of important metrics in the metric template k , determine the set of important metrics in the metric template, use the important metrics in the set as the input of the network model for model training, and determine the associated metric templates of the important metrics. Form an architecture model for generating power communication planning metric templates that synergistically combines high-frequency metric recommendations, collaborative filtering recommendations, and important metric recommendations.

[0052] S6: Perform semantic fusion according to the above-mentioned associated metric templates, determine the final recommended metrics and send them to the user.

[0053] To enable those skilled in the art to better understand the present invention, a more detailed embodiment is listed below. The embodiment of the present invention provides a method for adaptively generating power communication planning metrics based on user behavior, specifically as follows:

[0054] S100: To reduce the cold start phenomenon of traditional recommendation methods, judge the user type according to historical data information, and recommend different metric contents according to the user type to obtain user selection feedback. Among them, the user selection feedback includes display feedback and implicit feedback.

[0055] Applied to the embodiment of the present invention, recommending different metric contents includes: for new users, recommending high-frequency metrics as the user's metric selection; for old users, obtaining user preferences through historical data information and recommending user preference metrics as the user's metric selection.

[0056] As an example, explicit feedback: the behavior of a user clearly expressing their preference for an item. Implicit feedback: the behavior that cannot clearly reflect the user's preference. It is impossible to judge whether implicit feedback is dislike. For explicit feedback, it is obvious to distinguish whether it is like or dislike. The explicit feedback value represents the degree of preference, and the implicit feedback value represents the confidence level.

[0057] S200: According to the user selection feedback, fuse the display feedback and implicit feedback information through a metric fusion module to obtain the user's initial rating for the metric content and the initial rating matrix P.

[0058] (1) Explicit metrics include the selection frequency C of the metric template content ij and the expert score R of the importance of the metric ij , defined as:

[0059] C ij =n

[0060] In the formula, n represents the frequency of the user's selection of planning metrics. The expert score R of the importance of the metricij It is divided into 1 to 10 rating levels, indicating the user's satisfaction with the browsing metrics. The expert rating of the importance of the metrics can be defined as:

[0061]

[0062] where m represents the rating satisfaction, m ∈ (1, 10). When n is 0, the user has not selected the browsing metric, and the expert rating is set to 0. There is only an expert rating of the importance of browsing the metric after the user selects the browsing metric.

[0063] (2) The implicit metric is the usage duration T of the metric system ij , the longer the usage duration of the metric system, the higher the user's attention to the content of the metric system. However, the usage time cannot be infinitely reduced or increased. The invalid usage duration of the metric system should be excluded, and upper and lower thresholds are set for the system usage duration, expressed as:

[0064]

[0065] where is the average speed of the user's historical browsing, and L represents the total length of the corresponding reading metric template. The comprehensive rating P of metric fusion ij is expressed as:

[0066]

[0067] where are the normalized values respectively. The metric fusion uses the critic assignment method to solve the weights, and α + β + γ = 1.

[0068] (3) Through the metric fusion model, an initial rating matrix P of multiple users and multiple metric templates is constructed. It is defined as:

[0069]

[0070] where P ij is the comprehensive rating after fusion of user i for metric template j.

[0071] S300. According to the cosine similarity calculation, the similarity B between metric template i and metric template j is obtained ij :

[0072]

[0073] where N(i) represents the total number of people rating metric i; N(j) represents the total number of people rating metric j.

[0074] Therefore, a similarity matrix B between multiple metric templates is constructed:

[0075]

[0076] S400. Evaluate the user's attention score P' for any index template based on the similarity between index templates and the user's initial scoring matrix.

[0077] P' = P × B

[0078] According to the user preference feedback and the change of focus, add a decay function to reduce the recommendation of index templates with low user interest. Therefore, the decay function of user i for index template j is expressed as:

[0079]

[0080] where λ is a constant and D is the number of recommendation cycle times. Therefore, after each recommendation cycle feedback, the recommendation system generates a new attention score P' for any index template. new It can be iteratively updated as:

[0081]

[0082] where P' represents the attention score matrix for the index template in the previous round, and α represents the decay function matrix of the user for the index template. Through iterative update, the preference prediction of the user for the index template can be dynamically adjusted.

[0083] Finally, conduct a simulation on the collaborative filtering method after index fusion. Select 20 users and 300 index templates for the simulation experiment respectively. Randomly select 10 users to compare their accuracy and recall rate, and use the average value of the users as the simulation data of the experiment.

[0084] As Figure 2 and Figure 3 shown, compared with the user-based collaborative filtering (CF(user)) algorithm, the proposed collaborative filtering algorithm based on index fusion in the present invention has better recommendation accuracy and recall rate than the CF(user) algorithm.

[0085] S500. Obtain the index templates with the top p scores according to the collaborative filtering method as the data source for selecting important indicators. For index templates, the user's preference degree for index templates is related to the frequency of the appearance of important interest indicators. Therefore, the important indicator frequency w of the index template k is expressed as:

[0086]

[0087] where, f k represents the number of times the important indicator k appears; ∑f is the total number of all important indicators that appear in the index template; then the importance score of the important indicator k of user i in index template j Expressed as:

[0088]

[0089] Where P ij ' is the attention score of user i for indicator template j, and w k is the frequency of occurrence of important indicator k in indicator template j. According to the importance score sort in descending order, select the top m important indicators as the input of the network language model for BP model training, and finally determine the associated indicator content of the important indicators.

[0090] As shown in the Figure 4 figure is a comparison chart of the recommendation association index accuracy under neural network training. By using different numbers of neurons and the number of hidden layers, the highest accuracy of the recommendation can reach more than 90%.

[0091] S600. Perform semantic fusion according to the indicator template, and finally determine the recommended indicators and send them to the user.

[0092] In some other embodiments, the present invention also provides an indicator derivation recommendation strategy: according to the explicit and implicit feedback of users, perform recommendation scoring on the collected data required for power communication planning using indicators, and fuse according to the selection frequency and browsing times of indicators, the expert scoring of the importance of indicators, the system usage duration and other characteristics to obtain the preference ranking of user indicator selection. Then, perform indicator derivation according to the indicator generation algorithm to obtain new selectable indicators. It mainly includes three parts: semantic fusion of collaborative filtering recommendation, important indicator recommendation, and high-frequency indicator recommendation. By analyzing the functions and roles of each part, the accuracy and extensibility of the indicator recommendation system can be effectively improved. The indicator derivation recommendation strategy is divided into four major modules, specifically as follows:

[0093] (1) Scoring indicator fusion module, used to perform indicator fusion on explicit indicators and implicit indicators to obtain the value score of the user for the indicator content. Explicit indicators include the selection frequency (or browsing times) C ij of the indicator content and the expert scoring R ij of the importance of the indicator, defined as: C ij =n; where n represents the frequency of the user selecting the planning indicator. For the expert scoring R ij of the importance of the indicator, it is divided into 1-10 scoring levels, indicating the satisfaction degree of the user with the browsed indicator. The expert scoring of the importance of the indicator is defined as:

[0094] where m represents the satisfaction degree of the score, m ∈ (1, 10). When n is 0, it means that the user has not selected the browsed indicator, so the expert score is set to 0. There is only an expert score for the importance of browsing the indicator after the user selects the browsed indicator.

[0095] The implicit indicators include the usage duration T of the indicator system ij , the longer the usage duration of the indicator system, the higher the user's attention to the content of the indicator system. However, the usage time cannot be infinitely reduced or increased, and the usage duration of invalid indicator systems needs to be excluded. Therefore, upper and lower thresholds for the system usage duration need to be set. Thus, the system usage time T ij is expressed as: represents the average speed of the user's historical browsing of the indicator system, and L represents the total amount of the corresponding indicator system content. The comprehensive score P through the fusion of critic indicators ij is expressed as: where are the values after normalization of the three indicators respectively. The indicator fusion uses the critic assignment method to solve the weights, and α + β + γ = 1.

[0096] (2) The collaborative filtering recommendation module is used to recommend similar indicator content to similar users based on the user's preference degree for the indicator system content or the similarity of users, and can better recommend indicators for users. First, the evaluation indicator fusion module scores the indicator template content to construct a scoring matrix P between multiple users and multiple indicator templates:

[0097]

[0098] where P ij is the comprehensive score after fusion of user i for indicator template j. According to the cosine similarity calculation, the similarity B between indicator template i and indicator template j can be obtained ij . where N(i) represents the total number of people scoring indicator i; N(j) represents the total number of people scoring indicator j.

[0099] Construct a similarity matrix B between multiple indicator templates:

[0100]

[0101] Through the similarity between indicator contents and the user scoring matrix, the attention score P' of the user to any indicator template can be predicted, P' = P × B.

[0102] Through the attention score P', sort and select the top p scored indicator templates preferred by user i as the user indicator recommendation. However, the indicator recommendation content generated by the collaborative filtering method may cause changes in the indicator focus in the communication plan over time, resulting in an unselected state in the recommendation list. Therefore, the collaborative filtering method needs to appropriately increase a decay function to reduce the user's score for such indicator templates. The decay function is expressed as:

[0103]

[0104] where λ is a constant and D is the number of recommended cycle times. Update P' for the preference of each round of recommendation scores new = α × P', thus ensuring the dynamic variability of the user's interest preference.

[0105] (3) Important index recommendation module. This module performs index derivation on important indexes through an index generation algorithm to obtain selectable new indexes. Using important index recommendation can make up for the extensibility problem of the collaborative filtering recommendation module and improve the performance of the recommendation system. The important index recommendation in the present invention adopts Word2vec in the neural language model. Word2vec is a model that converts text into word vectors. It can train word vectors and predict the central word (context) through the context (central word), so as to obtain the correlation relationship between words in semantics.

[0106] For important index recommendation, first, it is necessary to construct a basic important index word library. Through the analysis of the content, select an important index set that can represent all the content of the index template and record the correlation relationship between words. Secondly, use Word2vec in the neural language model to perform BP neural network training on important indexes. The BP neural network calculates the loss function through forward propagation and updates the model parameters through backpropagation according to the loss function, thereby deriving the associated indexes related to each important index.

[0107] Use the content of the index template with the top p scores obtained by the collaborative filtering method based on indexes as the content of the index template for important index selection. For the index template, the preference degree of the user for the index template is often related to the frequency of the appearance of the preference important indexes. Therefore, the important index frequency w of the index template k can be expressed as: where, f k represents the number of times the important index k appears; ∑f is the total number of all important indexes that appear in the index template; then the importance score of the important index k of the user i for the index template j is expressed as: where P ij ' is the attention score of the user i for the index template j, and w k is the frequency of the appearance of the important index k in the index template j.

[0108] By sorting the importance scores in descending order, select the top m important indexes as the input layer of Word2vec, and the output layer is the associated index vector derived from the indexes. Through the training of the BP neural network, the index generation method is made more accurate and is more conducive to setting a relatively accurate personalized precise demand solution from the whole to the local.

[0109] (4) High-frequency index recommendation module, which is mainly used to solve the cold start problem of collaborative filtering method. The high-frequency recommendation method finds users' new interest preferences by leveraging high-frequency index template content. This derivative recommendation strategy adds high-frequency index content to the system, selects the top n high-frequency indexes and adds them to the index recommendation list, and alleviates the impact brought by the cold start problem by recommending high-frequency indexes to users. Perform Top-k recommendation selection on the content of collaborative filtering recommendation, important index recommendation, and high-frequency index recommendation, and recommend the final recommendation list to users.

[0110] In the second aspect of the present invention, there is provided a system for the method of adaptively generating power communication planning indexes based on user behavior, including:

[0111] User selection feedback acquisition module, which is used to acquire the historical data information of users to judge the user type, and recommend different index contents according to the user type to obtain user selection feedback;

[0112] First matrix acquisition module, which is used to record the explicit index and implicit index selection according to the user selection feedback, fuse the explicit index and implicit index, and determine the initial scoring matrix of the user for the index template content; calculate the cosine similarity between the index templates, and determine the similarity matrix between the index templates;

[0113] Second matrix acquisition module, which is used to determine the user scoring matrix after collaborative filtering according to the scoring matrix and the similarity matrix;

[0114] Recommended index module, which is used to determine the set of important indexes in the index template according to the important index frequency of the index template and the user scoring matrix; use the set of important indexes as the input for network training to determine the associated index template of the important indexes; perform semantic fusion on the associated index template, and finally determine the recommended indexes.

[0115] In the third aspect of the present invention, there is provided a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements the method of adaptively generating power communication planning indexes based on user behavior.

[0116] In the fourth aspect of the present invention, there is provided a computer-readable storage medium, which stores a computer program. When the computer program is executed by a processor, it implements the method of adaptively generating power communication planning indexes based on user behavior.

[0117] Those skilled in the art should understand that the embodiments of the present application can be provided as methods, systems, or computer program products. Therefore, the present application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Moreover, the present application can take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) that contain computer-usable program code.

[0118] The present application is described with reference to the flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each flow and / or block in the flowchart and / or block diagram, and the combination of flows and / or blocks in the flowchart and / or block diagram, can be realized by computer program instructions. These computer program instructions can be provided to the processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing devices to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing devices generate a device for realizing the functions specified in Figure 1 one or more of the flows Figure 1 or blocks or the combination of multiple flows and / or blocks.

[0119] These computer program instructions can also be stored in a computer-readable memory that can direct a computer or other programmable data processing device to work in a specific manner, so that the instructions stored in the computer-readable memory generate a manufactured article including an instruction device, and the instruction device realizes the functions specified in Figure 1 one or more of the flows Figure 1 or blocks or the combination of multiple flows and / or blocks.

[0120] These computer program instructions can also be loaded onto a computer or other programmable data processing device, so that a series of operation steps are executed on the computer or other programmable device to generate a computer-implemented process. Thus, the instructions executed on the computer or other programmable device provide steps for realizing the functions specified in Figure 1 one or more of the flows Figure 1 or blocks or the combination of multiple flows and / or blocks.

[0121] As is known by common technical knowledge, the present invention can be implemented by other embodiments that do not depart from its spiritual essence or essential features. Therefore, the above-disclosed embodiments are illustrative in all aspects and are not exclusive. All changes within the scope of the present invention or within the scope equivalent to the present invention are encompassed by the present invention.

Claims

1. An adaptive generation method for power communication planning indicators based on user behavior, characterized in that Including: Obtain the historical data information of the user to determine the user type, and recommend different index contents according to the user type to obtain user selection feedback; Record the explicit index and implicit index selections according to the user selection feedback, fuse the explicit index and implicit index, and determine the initial scoring matrix of the user for the index template content; Calculate the cosine similarity between the index templates to determine the similarity matrix between the index templates; Determine the user scoring matrix after collaborative filtering according to the scoring matrix and the similarity matrix; Determine the set of important indexes in the index template according to the important index frequency of the index template and the user scoring matrix; Use the set of important indexes as input for network training to determine the associated index templates of the important indexes; Perform semantic fusion on the associated index templates to finally determine the recommended indexes.

2. The method for adaptively generating power communication planning indicators based on user behavior according to claim 1, wherein The explicit metrics include the selection frequency C of the content of the metric template ij and the expert score R of the importance of the metric ij ; the implicit metric is the usage duration T of the metric system ij .

3. The method for adaptively generating power communication planning indicators based on user behavior according to claim 2, wherein The frequency C of selecting the content of the index template ij = n, where n represents the frequency of the user selecting the planning index; the expert score of the importance of the index m represents the degree of satisfaction of the score, and m ∈ (1, 10).

4. The method for adaptively generating power communication planning indicators based on user behavior according to claim 3, wherein, The usage duration of the said indicator system is the average speed of the user's historical browsing, and L represents the total length of the corresponding reading indicator template.

5. The method for adaptively generating power communication planning indicators based on user behavior according to claim 4, wherein, The initial scoring matrix P: P ij It is the comprehensive score after fusing the index template j for user i.

6. The method for adaptively generating power communication planning indicators based on user behavior according to claim 5, wherein Calculate the cosine similarity between the index templates to determine the similarity matrix between the index templates, and determine the user scoring matrix after collaborative filtering according to the scoring matrix and the similarity matrix, specifically as follows: Similarity N(i) represents the total number of people for rating index i, and N(j) represents the total number of people for rating index j; Similarity matrix Determine the attention score P' of the user for any index template according to the similarity between the index templates and the user initial scoring matrix, P' = P * B.

7. The method for adaptively generating power communication planning indicators based on user behavior according to claim 6, characterized in that, The specific method for determining the associated index templates of the important indexes is: Obtain the index templates with the top p scores as the important index k; Calculation Index Template - Frequency of Important Indicators where f k represents the number of occurrences of important indicator k, and ∑f is the total number of all important indicators in the indicator template; The importance score given by user i to important indicator k in indicator template j is denoted as where P ij ' is the attention score of user i to indicator template j, and w k is the frequency of occurrence of important indicator k in indicator template j; Score according to importance Sort in descending order, select the top m important indicators as the input of the network language model for BP model training, and finally determine the content of the associated indicators of the important indicators.

8. A system for the method of adaptively generating power communication planning metrics based on user behavior as described in claim 1, characterized in that, Including: A user selection feedback acquisition module, configured to obtain the historical data information of the user to determine the user type, and recommend different index contents according to the user type to obtain user selection feedback; A first matrix acquisition module, configured to record the explicit index and implicit index selections according to the user selection feedback, fuse the explicit index and implicit index, and determine the initial scoring matrix of the user for the index template content; Calculate the cosine similarity between the index templates to determine the similarity matrix between the index templates; A second matrix acquisition module, configured to determine the user scoring matrix after collaborative filtering according to the scoring matrix and the similarity matrix; A recommended index module, configured to determine the set of important indexes in the index template according to the important index frequency of the index template and the user scoring matrix; Use the set of important indexes as input for network training to determine the associated index templates of the important indexes; Perform semantic fusion on the associated index templates to finally determine the recommended indexes.

9. A computer device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the method for adaptively generating power communication planning indexes based on user behavior according to any one of claims 1 to 7.

10. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by the processor, it implements the method for adaptively generating power communication planning indexes based on user behavior according to any one of claims 1 to 7.

Images