New energy learning content recommendation method and system based on adaptive strategy

Abstract

The invention provides a new energy learning content recommendation method and system based on an adaptive strategy, and relates to the technical field of data processing, and the method comprises the steps: obtaining learning resources related to new energy; carrying out dimensionality reduction on the learning resources in combination with an auto-encoder; performing self-clustering on the dimensionality-reduced learning resources in combination with a soft allocation algorithm and a k-means clustering algorithm to obtain multiple classes of sub-learning resources; generating theme labels of various types of sub-learning resources; multiple learning behaviors of the target user are obtained, and the learning behaviors comprise learning content of each learning, learning content completion degree, learning time and learning duration; calculating interestingness of the target user to different theme tags based on each learning behavior; according to the interestingness, the recommendation proportion of each type of sub-learning resources is calculated in a positive correlation mode with a Nash equilibrium factor; and recommending each sub-learning resource to the target user according to the corresponding recommendation proportion. Recommendation diversity and accuracy are ensured, and user learning efficiency is improved.

Description

Technical Field

[0001] The present invention relates to the field of data processing technology, and in particular to a new energy learning content recommendation method and system based on an adaptive strategy. Background Art

[0002] New energy learning content refers to knowledge resources around new energy technologies, applications and industry-related fields, including technical theories, engineering applications, policies and regulations, market analysis and learning materials of solar energy, wind energy, energy storage technology, electric vehicles, hydrogen energy, etc., providing learners with systematic professional learning support.

[0003] Because new energy technology is complex and developing rapidly, learners face a huge amount of information and it is difficult to filter out high-quality resources. Therefore, it is extremely important to recommend new energy learning content. Through intelligent recommendation, we can accurately match user interests and needs, improve learning efficiency, help users quickly master core knowledge, and guide them to focus on key areas, promoting the popularization of new energy technology and industrial development.

[0004] However, when recommending existing new energy learning content, it is often impossible to select learning content that meets user needs from the massive amount of new energy learning content. In addition, there is also the problem of too single recommended content, which leads to insufficient accuracy of content recommendation and affects user learning effects. Summary of the invention

[0005] In order to solve the technical problem that, when recommending new energy learning content, the prior art often fails to select learning content that meets user needs from a massive amount of new energy learning content, and there is also a problem that the recommended content is too single, which leads to insufficient accuracy of content recommendation and affects the user's learning effect, the present invention provides a new energy learning content recommendation method and system based on an adaptive strategy.

[0006] The technical solution provided by the embodiment of the present invention is as follows:

[0007] First aspect

[0008] An embodiment of the present invention provides a new energy learning content recommendation method based on an adaptive strategy, comprising:

[0009] S1: Obtain learning resources about new energy;

[0010] S2: Combine autoencoders to reduce the dimension of learning resources;

[0011] S3: Combine the soft assignment algorithm and the k-means clustering algorithm to self-cluster the reduced-dimensional learning resources to obtain multiple types of sub-learning resources;

[0012] S4: Generate topic labels for various sub-learning resources;

[0013] S5: Acquire multiple learning behaviors of the target user, wherein the learning behaviors include the learning content, completion degree of the learning content, learning time and learning duration of each learning;

[0014] S6: Calculate the interest of the target user in different topic tags based on each learning behavior;

[0015] S7: Calculate the recommended proportion of each type of sub-learning resources according to the interest level in a positive correlation manner with a Nash equilibrium factor;

[0016] S8: Recommend each sub-learning resource to the target user according to the corresponding recommendation ratio.

[0017] Second aspect

[0018] An embodiment of the present invention provides a new energy learning content recommendation system based on an adaptive strategy, comprising:

[0019] processor;

[0020] A memory having computer-readable instructions stored thereon, wherein when the computer-readable instructions are executed by the processor, the method for recommending new energy learning content based on an adaptive strategy as described in the first aspect is implemented.

[0021] The third aspect

[0022] An embodiment of the present invention provides a computer-readable storage medium, on which a computer program is stored. When the program is executed by a processor, the method for recommending new energy learning content based on an adaptive strategy as described in the first aspect is implemented.

[0023] The beneficial effects brought about by the technical solution provided by the embodiment of the present invention include at least:

[0024] In an embodiment of the present invention, the learning resources are reduced in dimension by an autoencoder, the redundancy of the learning resources is reduced, and the core semantic information is retained. Then, the learning resources after dimension reduction are self-clustered by combining the soft allocation algorithm and the k-means clustering algorithm, which can improve the discrimination of different cluster clusters, make the resource classification more accurate, and dynamically allocate the probability to strengthen the influence of high-confidence samples, avoid the limitations of hard clustering, and more accurately divide the learning content topics, help the recommendation system capture the user's real interest in various new energy technologies, improve the relevance and diversity of recommended content, and meet personalized learning needs. After that, the theme tags of each sub-learning resource obtained by clustering are generated, and the interest of the target user in different theme tags, i.e., sub-learning resources, is calculated based on the target user's multiple learning behaviors, and then the recommendation ratio of each sub-learning resource is calculated in combination with a positive correlation method with a Nash equilibrium factor, and finally different categories of sub-learning resource content are extracted and recommended to the target user according to the recommendation ratio. Ensure that high-interest content is recommended first and diversity is retained, and finally accurately match the user's learning needs. This method can more effectively capture the user's real interest in new energy technologies, improve the relevance and exploratory nature of recommendations, and dynamically adapt to changes in user interests, significantly improving user learning efficiency and experience. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0026] Figure 1 A schematic diagram of a flow chart of a new energy learning content recommendation method based on an adaptive strategy provided by an embodiment of the present invention;

[0027] Figure 2 A schematic diagram of the structure of a new energy learning content recommendation system based on an adaptive strategy provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0028] The technical solution of the present invention is described below in conjunction with the accompanying drawings.

[0029] In the embodiments of the present invention, words such as "exemplarily" and "for example" are used to indicate examples, illustrations or explanations. Any embodiment or design described as "example" in the present invention should not be interpreted as being more preferred or more advantageous than other embodiments or designs. Specifically, the use of the word "example" is intended to present the concept in a specific way. In addition, in the embodiments of the present invention, the meaning expressed by "and / or" can be both, or it can be either of the two.

[0030] In order to make the technical problems, technical solutions and advantages to be solved by the present invention more clear, a detailed description will be given below with reference to the accompanying drawings and specific embodiments.

[0031] Reference Manual Attached Figure 1 , which shows a flow chart of a new energy learning content recommendation method based on an adaptive strategy provided in an embodiment of the present invention.

[0032] The embodiment of the present invention provides a method for recommending new energy learning content based on an adaptive strategy, which can be implemented by a device for recommending new energy learning content based on an adaptive strategy, and the device for recommending new energy learning content based on an adaptive strategy can be a terminal or a server. The processing flow of the method for recommending new energy learning content based on an adaptive strategy can include the following steps:

[0033] S1: Obtain learning resources about new energy.

[0034] It should be noted that by acquiring various forms of new energy learning resources (such as scientific research papers, video courses, industry reports, etc.), a diverse and comprehensive data foundation is provided for the recommendation system, covering new energy technology, industry and policy fields, ensuring the richness and applicability of the recommended content.

[0035] In one possible implementation, the learning resources include e-books, scientific research papers, academic journals, industry reports, case studies, and experimental tutorials.

[0036] Understandably, learning resources cover e-books, scientific research papers, academic journals, industry reports, case studies and experimental tutorials, providing comprehensive and diverse forms of knowledge. This approach ensures the coverage of different learning needs, meets users' learning needs in theory, application, practice and cases, and improves the applicability and practical value of recommended content.

[0037] S2: Combine autoencoders to reduce the dimension of learning resources.

[0038] Among them, the autoencoder is an artificial neural network model used for unsupervised learning. By compressing the input data into a low-dimensional latent space (encoding) and reconstructing it back into a high-dimensional space (decoding), the model learns the core features and key patterns of the data and reduces the redundancy of the data. By reducing the dimensionality of new energy learning resources through the autoencoder, it is possible to reduce the redundant information of the learning resources, extract the core semantic features, retain the key content, and reduce the complexity of the data, providing more accurate input for the subsequent clustering and recommendation process, thereby improving the relevance of the recommended content and the learning efficiency of users.

[0039] In a possible implementation, S2 specifically includes:

[0040] S201: Convert each text content in the learning resource into a low-dimensional continuous vector in a weighted average manner:

[0041]

[0042] Among them, v text Represents a low-dimensional continuous vector of each text content, W represents the word set of the text content, and v w represents the pre-trained word vector of word w, p(w) represents the recurrence frequency of word w in the text content to which it belongs, and a represents the smoothing factor used to control the weight of high-frequency words.

[0043] Among them, the pre-trained word vector can be a pre-trained word vector in Word2Vec or GloVe. The smoothing factor is a parameter used to dynamically adjust the weight of high-frequency words in text representation. Although high-frequency words appear frequently in text, their semantic information is usually less, so their influence needs to be appropriately reduced. Through such a smoothing factor, it is possible to effectively avoid high-frequency words dominating the text representation while making full use of the semantic value of low-frequency words. For short texts, 0.001 or 0.01 is often used to ensure that low-frequency words have a greater impact. For long texts, 0.1 or 1 is often used to appropriately reduce the weight of high-frequency words.

[0044] It should be noted that the weighted average method can dynamically adjust the importance of words, balance the contribution of high-frequency and low-frequency words to text representation, and generate low-dimensional continuous vectors that are both semantically accurate and efficient and compact. This method can significantly improve the model's ability to understand semantics in various natural language processing tasks.

[0045] It can be understood that representing text content as a dense low-dimensional continuous vector effectively solves the sparsity problem that may appear in the text content, while retaining the core semantic information, making short texts easier for subsequent clustering or classification tasks, and improving the final clustering accuracy.

[0046] S202: removing the first principal component of all low-dimensional continuous vectors from the low-dimensional continuous vectors of each text content to obtain an optimized low-dimensional continuous vector:

[0047] v final =v text -(v text ·u)·u

[0048] Among them, v final Represents the optimized low-dimensional continuous vectors of each text content obtained after removing the first principal component u from all low-dimensional continuous vectors.

[0049] The first principal component u in all low-dimensional continuous vectors is specifically the first principal component obtained by performing principal component analysis (PCA) on all low-dimensional continuous vectors. After the text embedding is processed by removing the principal component, the obtained vector is more focused on the unique semantic features of the text, rather than being disturbed by global frequent features. This improves the semantic resolution of the text embedding representation, making it more suitable for subsequent clustering or classification tasks.

[0050] It should be noted that through weighted averaging and principal component removal operations, the core semantic information of the text is retained while the redundancy of features is reduced.

[0051] S203: The optimized low-dimensional continuous vector is mapped to a potential representation through the encoder of the autoencoder to complete the dimensionality reduction of the learning resource.

[0052] Specifically, the encoder of the autoencoder compresses the input optimized low-dimensional continuous vector into a more compact latent space through a multi-layer neural network structure to capture the core semantic features of the data. Specifically, each layer of the encoder performs a linear transformation (weight matrix and bias term) on the input features and combines a nonlinear activation function to gradually extract higher-level abstract features, thereby generating the final latent representation, which condenses the key information of the learning resource.

[0053] It should be noted that through weighted averaging, autoencoders and PCA dimensionality reduction technology, this solution can reduce the redundancy of new energy learning resources, extract its core semantic features, and make the representation more compact and efficient. At the same time, the interference of frequent features is avoided by removing the principal component, ensuring the uniqueness of the potential representation. This process not only improves the semantic expression accuracy of learning resources, but also provides better input data for subsequent clustering and recommendation, significantly improving the relevance of recommended content and user learning efficiency.

[0054] In a possible implementation, the training method of the autoencoder is specifically as follows:

[0055] The autoencoder is trained with the goal of minimizing the reconstruction error of all low-dimensional continuous vectors after removing the first principal component. The reconstruction error is:

[0056]

[0057] Among them, v i Represents the optimized low-dimensional continuous vector of the i-th text content, i = 1, 2, ..., N, N represents the total number of text contents, Denotes the decoder pair v of the autoencoder i The reconstructed optimized low-dimensional continuous vector, L recon represents the reconstruction error.

[0058] It should be noted that during the training process, the network weights of the autoencoder are continuously optimized to minimize the reconstruction error, ensuring that the potential representation mapped by the encoder retains the original embedding, that is, the semantic features of the optimized low-dimensional continuous vector are retained to improve the accuracy of text classification. By minimizing the reconstruction error of the autoencoder, ensuring that the semantic features of the optimized low-dimensional continuous vector are fully retained, the potential representation is optimized, and the model can more accurately capture the core features of new energy learning content, thereby improving the accuracy of recommended content and classification efficiency, and meeting users' needs for diversified learning content.

[0059] S3: Combine the soft assignment algorithm and the k-means clustering algorithm to self-cluster the reduced-dimensional learning resources to obtain multiple categories of sub-learning resources.

[0060] Among them, the soft assignment algorithm is a clustering method. When classifying data, instead of assigning each sample to a cluster, it calculates the probability of each sample belonging to different clusters and generates flexible classification results, which better reflects the uncertainty of the data. K-means clustering is a commonly used hard clustering algorithm. Through iterative optimization, the data is divided into clusters, so that the Euclidean distance between the data points in each cluster and the center of the cluster is minimized, thereby achieving efficient classification of data. Sub-learning resources refer to a collection of categorized resources divided from the original learning resources by clustering methods. Each sub-resource represents the content of a specific topic or field, such as photovoltaic technology, wind energy storage, etc. The soft assignment algorithm and K-means clustering are used to self-cluster the reduced-dimensional learning resources, which can flexibly handle the uncertainty of resources and optimize the classification accuracy, and improve the distinction between different sub-learning resources. This solution makes the classification of new energy learning content more accurate, can effectively capture the real interest of users in different topics, ensure the diversity and relevance of recommended content, and meet the personalized learning needs of users.

[0061] In a possible implementation, S3 specifically includes:

[0062] S301: Cluster each potential representation once using a k-means clustering algorithm to obtain multiple category clusters.

[0063] Among them, the objective function of a clustering is specifically:

[0064]

[0065] Among them, z i represents the potential representation of the i-th text content, c i Indicates z i The label of the cluster to which it belongs, Represents the cluster label c i The corresponding cluster center, L k-means represents the objective function value, i.e., the primary clustering error. Indicates the calculation of z i and The square of the Euclidean distance between them.

[0066] It can be understood that the objective function of a clustering is to calculate the sum of the squares of the Euclidean distances between the potential representations of all text contents and their corresponding cluster centers, minimize the error and optimize the distribution of cluster centers, thereby making the text content within each cluster more compact, improving the accuracy of clustering and the representativeness of clustering results, and laying a good foundation for subsequent classification and recommendation.

[0067] S302: Calculate the probability that each potential representation belongs to a different category cluster through a soft assignment algorithm:

[0068]

[0069] Among them, q ij represents the probability that the potential representation of the i-th text content belongs to the j-th category cluster, j′ represents the cluster index of all category clusters, μ j and μ j′ They represent the center points of the j-th category cluster and the j′-th category cluster respectively.

[0070] It should be noted that by calculating the distance between each potential representation and the center of all category clusters, the probability distribution belonging to each cluster is generated. The higher the probability, the stronger the similarity between the potential representation and the corresponding cluster. This method avoids the rigid limitations of hard clustering, better reflects the uncertainty of data, and makes the clustering results more flexible and accurate, thereby improving the meticulousness of new energy learning resource classification and the flexibility of the recommendation system.

[0071] S303: Generate expected target probability about probability:

[0072]

[0073] Among them, p ij The expected target probability that the latent representation representing the i-th text content belongs to the j-th category cluster.

[0074] in, The current allocation q for cluster j ij Strengthen (square) and normalize to the inside of the cluster, Traverse all clusters and normalize the molecules to ensure that the total probability of all clusters is 1. The expected target probability p ij Prefer to strengthen those soft assignment probabilities q ijThe target probability is a self-supervisory signal that drives the model (such as encoder parameters and cluster centers) to continuously optimize by minimizing the KL divergence with the soft assignment. The target probability can avoid the interference of large clusters on the overall optimization and improve the clustering purity.

[0075] More specifically, the expected target probability can strengthen the clustering results with high confidence and weaken the clustering results with low confidence. For example, if a clustering q ij The result of clustering is 0.9, then the expected target probability will be enhanced to 0.95. If the result of a clustering is 0.1, then the expected target probability will be weakened to 0.01.

[0076] S304: Update the center position of the category cluster with the goal of minimizing the KL divergence between the probability and the expected target probability:

[0077]

[0078] Among them, L KL Indicates p ij With q ij The KL divergence between , K represents the total number of category clusters, and log represents the logarithmic function.

[0079] Among them, KL divergence measures the difference between the soft assignment probability and the expected target probability. The goal is to minimize this difference, thereby optimizing the center position of the category cluster. By iteratively adjusting the cluster center, the actual distribution and the target distribution gradually converge, strengthening the influence of high-confidence samples, and improving the accuracy and stability of the clustering results. This optimization process ensures the rationality and robustness of the new energy learning content classification, providing a basis for accurate recommendation.

[0080] It should be noted that the results of the preliminary clustering are combined with the soft assignment method to generate probability distribution, and the model is optimized based on this, making the final clustering result more accurate and more robust. It is a dynamic, self-supervised optimization strategy that improves the clustering quality while avoiding the risk of premature convergence or model collapse.

[0081] In actual application, the encoder parameters and the center point position of the category cluster can be updated together:

[0082] The update formula of the encoder is as follows:

[0083]

[0084] Among them, θ represents the encoder parameters, which include the weight matrix that determines how each layer extracts features from the output of the previous layer and the bias term used to adjust the output value after the input passes through the weight matrix. represents the partial derivative, and ∝ represents a directly proportional relationship.

[0085] The update formula of the center position of the category cluster, i.e. the cluster center, is as follows:

[0086]

[0087] Specifically, this joint update scheme of encoder parameters and cluster center positions dynamically adjusts feature extraction and cluster centers by minimizing KL divergence, enabling the encoder to extract more accurate potential representations while optimizing the clustering structure. For new energy learning content recommendation, this scheme can more clearly divide different topic contents in the clustering results, improve classification accuracy and the model's adaptability to dynamic interests, thereby providing more relevant and efficient recommended content to meet users' personalized learning needs.

[0088] S305: Taking the center position of each updated category cluster as the cluster center point, perform secondary clustering on each potential representation through the k-means clustering algorithm to obtain multiple categories of sub-learning resources.

[0089] It can be understood that after the latent representations are clustered, multiple categories of sub-learning resources can be obtained by dividing the learning resources according to the clustering results.

[0090] In actual application, the primary clustering and soft allocation algorithms are combined to enhance the influence of high-confidence samples and dynamically optimize the location of cluster centers, thereby avoiding large cluster interference and premature convergence, thereby improving the purity and robustness of clustering. Secondary clustering further optimizes the accuracy of resource classification, ensuring that new energy learning content can be accurately classified and allocated according to user interests, effectively improving the relevance of recommendations and user learning efficiency, while enhancing the diversity and adaptability of recommendations to meet user personalized needs.

[0091] S4: Generate topic labels for various sub-learning resources.

[0092] It should be noted that by generating topic tags for various sub-learning resources, the core content of each type of learning resource can be clearly identified, the semantic expression ability of the resources can be improved, and the recommendation system can match user interests more accurately. At the same time, it is convenient for users to quickly understand and select new energy topics of interest, thereby improving learning efficiency and personalized recommendation experience.

[0093] In a possible implementation manner, S4 is specifically:

[0094] Topic labels for various sub-learning resources are generated through topic models, where the topic models include LDA topic models and NMF topic models.

[0095] It should be noted that by generating topic labels for various sub-learning resources through LDA and NMF topic models, the core semantic features of each type of resource can be extracted, so that the recommendation system can more accurately match user interests, while providing users with intuitive topic information for quick understanding and selection. This solution can effectively improve the recommendation accuracy, classification clarity and user learning efficiency of new energy learning content, and meet the diverse learning needs in complex fields.

[0096] S5: Obtain multiple learning behaviors of target users.

[0097] Among them, learning behavior includes the learning content, completion degree of learning content, learning time and learning duration of each learning session.

[0098] Among them, learning behavior refers to the specific data of users' interaction with learning resources during the learning process. By obtaining users' multiple learning behaviors, we can fully capture the dynamics of users' interests in different learning resources, combine learning content, completion and duration information, and more accurately characterize users' learning needs, providing data support for subsequent interest calculations and personalized recommendations, thereby improving the accuracy of recommended content and user satisfaction.

[0099] It should be noted that those skilled in the art can set the number of times of learning behavior collection according to actual needs, and the present invention does not limit this. The learning behavior acquired in step S5 is a plurality of learning behaviors acquired based on the current moment, that is, recent learning behavior.

[0100] S6: Calculate the target user's interest in different topic tags based on each learning behavior.

[0101] It should be noted that by analyzing users' multiple learning behaviors and dynamically calculating users' interest in different topic tags, we can accurately capture users' real needs and interest changes in new energy learning content, provide targeted data support for the recommendation system, and effectively improve the relevance and personalization of recommended content. At the same time, it ensures that recommendations can be dynamically adjusted according to user interests, thereby enhancing learning experience and efficiency.

[0102] In a possible implementation, S6 specifically includes:

[0103] S601: Extracting learning content topic labels of each learning content through a topic model.

[0104] S602: Calculate the cosine similarity between the subject label of each learning content and the subject label of each sub-learning resource.

[0105] The calculation method of cosine similarity is as follows:

[0106]

[0107] Among them, S represents the learning content topic label embedding vector V for each learning a Any of the topic labels in the learning resource is embedded in the vector V b The cosine similarity of , where the embedding vector can be generated by the pre-trained word vector model Word2Vec or GloVe.

[0108] It is understandable that in this way, the semantic association between learning content and learning resources can be effectively captured, ensuring that the recommended resources are highly matched with the user's learning content needs, thereby improving the accuracy and relevance of the recommendations.

[0109] S603: retaining the topic label of the sub-learning resource with the highest cosine similarity to the topic label of each learning content.

[0110] S604: Calculate the interest of the target user in each retained topic tag by combining the learning content completion, learning time and learning duration:

[0111]

[0112] Among them, I r represents the target user’s interest in the rth retained topic tag, C r,k represents the target user’s learning content completion degree for the sub-learning resource corresponding to the rth topic label in the kth learning behavior, T r,k represents the learning time of the target user for the sub-learning resource corresponding to the rth topic label in the kth learning behavior, K represents the total number of learning behaviors obtained, and t now Indicates the current timestamp, t k represents the timestamp of the kth learning behavior, ε represents a constant, M r represents the number of times the target user selects the rth topic tag in the historical learning behavior, M represents the total number of historical learning behaviors, and P r represents the target user’s preference value for the rth topic tag, α and β represent the first weight and the second weight, respectively.

[0113] Among them, the criterion for determining whether a historical learning behavior is a historical learning behavior in the total number of historical learning behaviors is also based on the current moment, that is, all learning behaviors generated before the current moment are historical learning behaviors. The user learning content labels are extracted through the topic model, and the topic labels most relevant to the sub-learning resources are retained. At the same time, the interest degree is dynamically calculated using the time decay weight, taking into account recent learning behaviors and long-term interest preferences. This method ensures that the recommended content accurately matches user needs, which not only improves the relevance of new energy learning content recommendations, but also enhances the system's adaptability to dynamic changes in user interests, thereby improving the personalization of recommendations and learning efficiency.

[0114] Among them, the time decay weight is introduced to calculate the interest of the target user for each retained topic tag, ensuring that the recent learning behavior has a greater impact on the interest. ε specifically represents a small constant to avoid the denominator being zero. The introduction of Pr ensures that the user's long-term interest preferences are taken into account in the calculation process of interest.

[0115] In actual application, the weights for measuring recent learning behavior and long-term learning behavior, ie, the first weight and the second weight, can be determined based on the user's recent learning behavior and the degree of completion of each learning content in the user's historical learning behavior.

[0116]

[0117] in, represents the Laplace transform, C recent (t) and C historical (t) represents the correlation function between the learning content completion of the user’s recent learning behavior and the user’s historical learning behavior and time t, that is, a learning content completion time series data composed of the learning content completion of the learning behavior corresponding to the timestamp t, α(s) and β(s) represent the first weight and the second weight under the Laplace complex frequency variable s, respectively.

[0118] Among them, the Laplace complex frequency variable is the independent variable in the Laplace transform. By introducing the Laplace complex frequency variable to calculate the weight, the time trend of recent and historical behaviors can be considered at the same time, and the weight ratio can be dynamically allocated. This method can not only capture the rapid changes in user behavior, but also smoothly process the long-term behavior impact through frequency domain analysis, thereby improving the model's responsiveness to changes in interest and the accuracy of weight allocation.

[0119] It should be noted that the Laplace transform maps the learning behavior distribution in the time domain to the frequency domain, which can capture the long-term trend and short-term fluctuation of the user's learning behavior. The balance between recent and historical weights can be more reasonably adjusted in the frequency domain to more accurately measure the importance of recent and historical learning behaviors, so that the final recommended content can better fit the preferences of the target user. Optionally, when there is not enough learning behavior data, α and β can also be set to specific values ​​0.5 and 0.5 respectively to balance the weights of recent and long-term learning behaviors.

[0120] S7: Calculate the recommended proportion of each type of sub-learning resources according to the interest level in a positive correlation manner with a Nash equilibrium factor.

[0121] Among them, the Nash equilibrium factor is a weight adjustment factor introduced from the perspective of game theory, which is used to optimize the recommendation ratio so that the recommendation system can achieve a dynamic balance between different topics of user interest. The weight of the recommendation ratio is adjusted by calculating the degree of deviation between the interest level of each topic and the average interest level. If the interest level of a topic is significantly higher or lower than the average, the corresponding Nash equilibrium factor will increase, strengthening the recommendation ratio, making users pay more attention to high-interest content or being guided to low-interest but important content.

[0122] It should be noted that the proportional recommendation method can dynamically adjust the recommended proportion of various sub-learning resources to meet users' interest needs in different topics and avoid a decline in learning experience caused by too single recommended content. By broadening the scope of recommendations, the system can not only cover areas of potential interest to users, but also guide users to discover new knowledge points, improve the comprehensiveness and exploratory nature of learning, and effectively improve the diversity and generalization ability of recommendations, thereby enhancing user stickiness and satisfaction.

[0123] In a possible implementation manner, the recommended ratio is calculated as follows:

[0124]

[0125] Among them, G r represents the recommended proportion of sub-learning resources corresponding to the rth topic label, represents the average interest of the target user in all retained topic tags, Δ r I represents the target user's interest in the rth retained topic tag r and The difference between them, R represents the total number of topic tags, θ r represents the Nash equilibrium factor for the rth topic label.

[0126] It should be noted that by introducing the Nash equilibrium factor to dynamically adjust the recommendation ratio, a balanced optimization is performed based on the difference between the user's interest in different topic tags and the average interest, so that high-interest topics receive a higher recommendation weight, while guiding users to pay attention to low-interest but potentially important topics. This method ensures the dynamic and balanced nature of the recommendation, which can not only meet the user's core learning needs, but also expand their exploration areas, and improve the diversity, accuracy and user learning efficiency of the recommendation.

[0127] S8: Recommend each sub-learning resource to the target user according to the corresponding recommendation ratio.

[0128] Specifically, by combining the calculation results of the user's interest in each topic label and the Nash equilibrium factor, the recommendation ratio of each sub-learning resource is dynamically adjusted; then, according to the recommendation ratio, the sub-learning resources of the topic label with the corresponding area ratio or database ratio are displayed on each recommendation page according to this recommendation ratio. The extraction method in the sub-learning resources can be random. This ensures that the recommended content can not only give priority to the user's high-interest topics, but also retain the diversity and exploratory nature of the recommended content, and finally accurately recommend it to the user. For example, when recommending new energy learning resources, assuming that the user has a high interest in "photovoltaic technology", the recommendation ratio is 50%, and the recommendation ratio for "energy storage technology" and "electric vehicle technology" is 30% and 20% respectively. On the recommendation page, 5 photovoltaic-related contents, 3 energy storage contents and 2 electric vehicle contents are displayed according to the ratio, and the specific sub-learning resources displayed are randomly extracted from each category. This method not only gives priority to meeting the user's learning needs for photovoltaic technology, but also retains the content related to energy storage and electric vehicles, and improves the diversity and exploratory nature of the recommendation.

[0129] In the actual application process, by acquiring comprehensive and diverse new energy learning resources, combining autoencoder dimensionality reduction to reduce redundancy and retain core features, and accurately dividing sub-learning resources through soft allocation algorithm and K-means clustering, clear topic labels are generated to improve semantic expression capabilities. By collecting user learning behaviors to dynamically calculate interest, the Nash equilibrium factor is used to optimize the recommendation ratio to achieve dynamic balance, and finally recommend content in proportion to improve the accuracy, diversity and learning efficiency of recommendations, and fully meet the personalized needs of users. It effectively avoids the problem that current recommendation schemes often cannot clearly distinguish the categories of learning content, resulting in large deviations in recommended content. This solution can capture user interests and recommend learning content based on interests, rather than recommending based on inherent patterns, reducing the difficulty of information acquisition for target users and improving learning efficiency.

[0130] The beneficial effects brought about by the technical solution provided by the embodiment of the present invention include at least:

[0131] In an embodiment of the present invention, the learning resources are reduced in dimension by an autoencoder, the redundancy of the learning resources is reduced, and the core semantic information is retained. Then, the learning resources after dimension reduction are self-clustered by combining the soft allocation algorithm and the k-means clustering algorithm, which can improve the discrimination of different cluster clusters, make the resource classification more accurate, and dynamically allocate the probability to strengthen the influence of high-confidence samples, avoid the limitations of hard clustering, and more accurately divide the learning content topics, help the recommendation system capture the user's real interest in various new energy technologies, improve the relevance and diversity of recommended content, and meet personalized learning needs. After that, the theme tags of each sub-learning resource obtained by clustering are generated, and the interest of the target user in different theme tags, i.e., sub-learning resources, is calculated based on the target user's multiple learning behaviors, and then the recommendation ratio of each sub-learning resource is calculated in combination with a positive correlation method with a Nash equilibrium factor, and finally different categories of sub-learning resource content are extracted and recommended to the target user according to the recommendation ratio. Ensure that high-interest content is recommended first and diversity is retained, and finally accurately match the user's learning needs. This method can more effectively capture the user's real interest in new energy technologies, improve the relevance and exploratory nature of recommendations, and dynamically adapt to changes in user interests, significantly improving user learning efficiency and experience.

[0132] Reference Manual Attached Figure 2 , showing a structural schematic diagram of a new energy learning content recommendation system based on an adaptive strategy provided by the present invention.

[0133] The present invention further provides a new energy learning content recommendation system 20 based on an adaptive strategy, which is applied to the above-mentioned new energy learning content recommendation method based on an adaptive strategy, and comprises:

[0134] Processor 201.

[0135] The memory 202 stores computer-readable instructions. When the computer-readable instructions are executed by the processor 201, the method for recommending new energy learning content based on an adaptive strategy as in the method embodiment is implemented.

[0136] The new energy learning content recommendation system 20 based on adaptive strategy provided by the present invention can execute the above-mentioned new energy learning content recommendation method based on adaptive strategy and achieve the same or similar technical effects. To avoid repetition, the present invention will not go into details.

[0137] The beneficial effects brought about by the technical solution provided by the embodiment of the present invention include at least:

[0138] In an embodiment of the present invention, the learning resources are reduced in dimension by an autoencoder, the redundancy of the learning resources is reduced, and the core semantic information is retained. Then, the learning resources after dimension reduction are self-clustered by combining the soft allocation algorithm and the k-means clustering algorithm, which can improve the discrimination of different cluster clusters, make the resource classification more accurate, and dynamically allocate the probability to strengthen the influence of high-confidence samples, avoid the limitations of hard clustering, and more accurately divide the learning content topics, help the recommendation system capture the user's real interest in various new energy technologies, improve the relevance and diversity of recommended content, and meet personalized learning needs. After that, the theme tags of each sub-learning resource obtained by clustering are generated, and the interest of the target user in different theme tags, i.e., sub-learning resources, is calculated based on the target user's multiple learning behaviors, and then the recommendation ratio of each sub-learning resource is calculated in combination with a positive correlation method with a Nash equilibrium factor, and finally different categories of sub-learning resource content are extracted and recommended to the target user according to the recommendation ratio. Ensure that high-interest content is recommended first and diversity is retained, and finally accurately match the user's learning needs. This method can more effectively capture the user's real interest in new energy technologies, improve the relevance and exploratory nature of recommendations, and dynamically adapt to changes in user interests, significantly improving user learning efficiency and experience.

[0139] It should be understood that the processor in the embodiment of the present invention may be a central processing unit (CPU), and the processor may also be other general-purpose processors, digital signal processors (DSP), application specific integrated circuits (ASIC), field programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor may be a microprocessor or the processor may also be any conventional processor, etc.

[0140] It should also be understood that the memory in the embodiments of the present invention may be a volatile memory or a non-volatile memory, or may include both volatile and non-volatile memories. Among them, the non-volatile memory may be a read-only memory (ROM), a programmable read-only memory (PROM), an erasable programmable read-only memory (EPROM), an electrically erasable programmable read-only memory (EEPROM), or a flash memory. The volatile memory may be a random access memory (RAM), which is used as an external cache. By way of example and not limitation, many forms of random access memory (RAM) are available, such as static RAM (SRAM), dynamic random access memory (DRAM), synchronous DRAM (SDRAM), double data rate synchronous dynamic random access memory (DDR SDRAM), enhanced synchronous dynamic random access memory (ESDRAM), synchronous link DRAM (SLDRAM), and direct RAM bus random access memory (DR RAM).

[0141] The above embodiments can be implemented in whole or in part by software, hardware (such as circuits), firmware or any other combination. When implemented by software, the above embodiments can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, the process or function described in the embodiment of the present invention is generated in whole or in part. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium, or transmitted from one computer-readable storage medium to another computer-readable storage medium. For example, the computer instructions can be transmitted from one website, computer, server or data center to another website, computer, server or data center by wired (such as infrared, wireless, microwave, etc.). The computer-readable storage medium can be any available medium that can be accessed by a computer or a data storage device such as a server or data center that contains one or more available media sets. The available medium can be a magnetic medium (for example, a floppy disk, a hard disk, a tape), an optical medium (for example, a DVD), or a semiconductor medium. The semiconductor medium can be a solid-state hard disk.

[0142] It should be understood that the term "and / or" in this article is only a description of the association relationship of associated objects, indicating that there can be three relationships. For example, A and / or B can represent: A exists alone, A and B exist at the same time, and B exists alone. A and B can be singular or plural. In addition, the character " / " in this article generally indicates that the associated objects before and after are in an "or" relationship, but it may also indicate an "and / or" relationship. Please refer to the context for specific understanding.

[0143] In the present invention, "at least one" means one or more, and "plurality" means two or more. "At least one of the following" or similar expressions refers to any combination of these items, including any combination of single items or plural items. For example, at least one of a, b, or c can be represented by: a, b, c, ab, ac, bc, or abc, where a, b, c can be single or multiple.

[0144] It should be understood that in various embodiments of the present invention, the size of the serial numbers of the above-mentioned processes does not mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present invention.

[0145] Those skilled in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of the present invention.

[0146] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the above-described equipment, devices and units can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0147] 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 units is only a logical function division. There may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another device, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms.

[0148] The units described as separate components may or may not be physically separated, and the components shown as units 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 units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0149] In addition, each functional unit in each embodiment of the present invention may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit.

[0150] If the functions are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention can be essentially or partly embodied in the form of a software product that contributes to the prior art. The computer software product is stored in a storage medium and includes several instructions for enabling a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the methods described in various embodiments of the present invention. The aforementioned storage media include: U disk, mobile hard disk, read-only memory (ROM), random access memory (RAM), magnetic disk or optical disk, etc., various media that can store program codes.

[0151] An embodiment of the present invention provides a computer-readable storage medium having a computer program stored thereon. When the program is executed by a processor, the method for recommending new energy learning content based on an adaptive strategy as described in the method embodiment is implemented.

[0152] A computer-readable storage medium provided by the present invention can implement the steps and effects of the new energy learning content recommendation method based on the adaptive strategy of the above method embodiment. To avoid repetition, the present invention will not go into details.

[0153] The beneficial effects brought about by the technical solution provided by the embodiment of the present invention include at least:

[0154] In an embodiment of the present invention, the learning resources are reduced in dimension by an autoencoder, the redundancy of the learning resources is reduced, and the core semantic information is retained. Then, the learning resources after dimension reduction are self-clustered by combining the soft allocation algorithm and the k-means clustering algorithm, which can improve the discrimination of different cluster clusters, make the resource classification more accurate, and dynamically allocate the probability to strengthen the influence of high-confidence samples, avoid the limitations of hard clustering, and more accurately divide the learning content topics, help the recommendation system capture the user's real interest in various new energy technologies, improve the relevance and diversity of recommended content, and meet personalized learning needs. After that, the theme tags of each sub-learning resource obtained by clustering are generated, and the interest of the target user in different theme tags, i.e., sub-learning resources, is calculated based on the target user's multiple learning behaviors, and then the recommendation ratio of each sub-learning resource is calculated in combination with a positive correlation method with a Nash equilibrium factor, and finally different categories of sub-learning resource content are extracted and recommended to the target user according to the recommendation ratio. Ensure that high-interest content is recommended first and diversity is retained, and finally accurately match the user's learning needs. This method can more effectively capture the user's real interest in new energy technologies, improve the relevance and exploratory nature of recommendations, and dynamically adapt to changes in user interests, significantly improving user learning efficiency and experience.

[0155] The above is only a specific embodiment of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art can easily think of changes or substitutions within the technical scope disclosed by the present invention, which should be included in the protection scope of the present invention. Therefore, the protection scope of the present invention should be based on the protection scope of the claims.

[0156] There are a few points to note:

[0157] (1) The drawings of the embodiments of the present invention only relate to the structures related to the embodiments of the present invention, and other structures may refer to the general design.

[0158] (2) For the sake of clarity, in the drawings used to describe the embodiments of the present invention, the thickness of the layers or regions is exaggerated or reduced, that is, these drawings are not drawn according to the actual scale. It is understood that when an element such as a layer, film, region or substrate is referred to as being "on" or "under" another element, the element may be "directly" "on" or "under" the other element or there may be intermediate elements.

[0159] (3) In the absence of conflict, the embodiments of the present invention and the features therein may be combined with each other to obtain new embodiments.

[0160] The above are only specific embodiments of the present invention, but the protection scope of the present invention is not limited thereto. The protection scope of the present invention shall be based on the protection scope of the claims.

Claims

1. A new energy learning content recommendation method based on adaptive strategy, characterized in that: Methods include: S1: Obtain learning resources about new energy; S2: reducing the dimension of the learning resource by combining the autoencoder; S3: Combine the soft assignment algorithm and the k-means clustering algorithm to self-cluster the reduced-dimensional learning resources to obtain multiple types of sub-learning resources; S4: Generate topic labels for various sub-learning resources; S5: Acquire multiple learning behaviors of the target user, wherein the learning behaviors include the learning content, completion degree of the learning content, learning time and learning duration of each learning; S6: Calculating the interest of the target user in different topic tags based on each learning behavior; S7: Calculating the recommended proportion of each type of sub-learning resource in a positive correlation manner with a Nash equilibrium factor according to the interest degree; S8: Recommend each sub-learning resource to the target user according to a corresponding recommendation ratio.

2. The method for recommending new energy learning content based on adaptive strategy according to claim 1, characterized in that: The learning resources include e-books, scientific research papers, academic journals, industry reports, case studies and experimental tutorials.

3. The new energy learning content recommendation method based on adaptive strategy according to claim 1 is characterized in that: The S2 specifically includes: S201: Convert each text content in the learning resource into a low-dimensional continuous vector in a weighted average manner: Among them, v text Represents a low-dimensional continuous vector of each text content, W represents the word set of the text content, and v w represents the pre-trained word vector of word w, p(w) represents the recurrence frequency of word w in the text content to which it belongs, and a represents the smoothing factor used to control the weight of high-frequency words; S202: removing the first principal component of all low-dimensional continuous vectors from the low-dimensional continuous vectors of each text content to obtain an optimized low-dimensional continuous vector: v final =v text -(v text ·u)·u Among them, v final represents the optimized low-dimensional continuous vector of each text content obtained after removing the first principal component u from all low-dimensional continuous vectors; S203: Mapping the optimized low-dimensional continuous vector to a potential representation through an encoder of an autoencoder to complete dimensionality reduction of the learning resource.

4. The new energy learning content recommendation method based on adaptive strategy according to claim 1 is characterized in that: The training method of the autoencoder is specifically as follows: The autoencoder is trained with the goal of minimizing the reconstruction error of all low-dimensional continuous vectors after removing the first principal component. The reconstruction error is specifically: Among them, v i Represents the optimized low-dimensional continuous vector of the i-th text content, i = 1, 2, ..., N, N represents the total number of text contents, Denotes the decoder pair v of the autoencoder i The reconstructed optimized low-dimensional continuous vector, L recon represents the reconstruction error.

5. The new energy learning content recommendation method based on adaptive strategy according to claim 3 is characterized in that: The S3 specifically includes: S301: clustering each potential representation once using the k-means clustering algorithm to obtain multiple category clusters; S302: Calculate the probability that each potential representation belongs to a different category cluster through a soft assignment algorithm: Among them, q ij represents the probability that the potential representation of the i-th text content belongs to the j-th category cluster, j′ represents the cluster index of all category clusters, μ j and μ j′ Represent the center points of the j-th category cluster and the j′-th category cluster respectively; S303: Generate an expected target probability about the probability: Among them, p ij The expected target probability that the latent representation of the i-th text content belongs to the j-th category cluster; S304: With the goal of minimizing the KL divergence between the probability and the expected target probability, update the center position of the category cluster: Among them, L KL Indicates p ij With q ij The KL divergence between , K represents the total number of category clusters, and log represents the logarithmic function; S305: Taking the center position of each updated category cluster as the cluster center point, perform secondary clustering on each potential representation through the k-means clustering algorithm to obtain multiple categories of sub-learning resources.

6. The method for recommending new energy learning content based on adaptive strategy according to claim 1, characterized in that: The S4 is specifically: The topic labels of various sub-learning resources are generated through a topic model, wherein the topic model includes an LDA topic model and an NMF topic model.

7. The method for recommending new energy learning content based on adaptive strategy according to claim 1, characterized in that: The S6 specifically includes: S601: extracting learning content topic labels of each learning content through a topic model; S602: Calculate the cosine similarity between the subject labels of each learning content and the subject labels of each sub-learning resource; S603: retaining the topic label of the sub-learning resource with the highest cosine similarity to the topic label of each learning content; S604: Calculate the interest of the target user in each retained topic tag by combining the completion degree of the learning content, the learning time and the learning duration: Among them, I r represents the target user’s interest in the rth retained topic tag, C r，k represents the target user’s learning content completion degree for the sub-learning resource corresponding to the rth topic label in the kth learning behavior, T r，k represents the learning time of the target user for the sub-learning resource corresponding to the rth topic label in the kth learning behavior, K represents the total number of learning behaviors obtained, and t now Indicates the current timestamp, t k represents the timestamp of the kth learning behavior, ε represents a constant, M r represents the number of times the target user selects the rth topic tag in the historical learning behavior, M represents the total number of historical learning behaviors, and P r represents the target user’s preference value for the rth topic tag, α and β represent the first weight and the second weight, respectively.

8. The method for recommending new energy learning content based on adaptive strategy according to claim 7, characterized in that: The recommended ratio is calculated as follows: Among them, G r represents the recommended proportion of sub-learning resources corresponding to the rth topic label, represents the average interest of the target user in all retained topic tags, Δ r I represents the target user's interest in the rth retained topic tag r and The difference between them, R represents the total number of topic tags, θ r represents the Nash equilibrium factor for the rth topic label.

9. A new energy learning content recommendation system based on adaptive strategy, characterized in that: include: processor; A memory having computer-readable instructions stored thereon, wherein when the computer-readable instructions are executed by the processor, the method for recommending new energy learning content based on an adaptive strategy as described in any one of claims 1 to 8 is implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the method for recommending new energy learning content based on an adaptive strategy as described in any one of claims 1 to 8 is implemented.

Images