Online learning resource recommendation method and system based on improved badger population algorithm

Abstract

The invention discloses an online learning resource recommendation method and system based on an improved badger population algorithm. The method comprises the steps of obtaining a user portrait and learning resource characteristics of a learner; the user portrait and the learning resource features are input into a pre-trained learning resource recommendation model, and the learning resource recommendation model constructs a multi-dimensional matching fitness function between the learner and the learning resources according to the user portrait and the learning resource features of the learner, and solving the multi-dimensional matching fitness function through an improved badger population algorithm and outputting recommended learning resources. According to the online learning resource recommendation method and system based on the improved badger population algorithm, by analyzing multi-dimensional features of learners and learning resources, an online learning resource recommendation model for deep analysis of multi-dimensional feature differences of the learners and the learning resources is constructed, and by solving the improved badger population algorithm, the online learning resource recommendation method and system based on the improved badger population algorithm are obtained. And more personalized learning paths and resource recommendation meeting requirements are provided for learners.

Description

Technical Field

[0001] The present invention belongs to the field of intelligent learning applications, and particularly relates to a personalized online learning resource recommendation method and system based on an improved honey badger algorithm. Background Art

[0002] With the rapid development of artificial intelligence technologies such as knowledge graphs and deep learning, society is stepping into a new era of "Artificial Intelligence +" from the "Internet +" era. Against the backdrop of this intelligent era, the education field is facing unprecedented opportunities and challenges. How to rely on intelligent technologies to innovate teaching methods, reshape the learning process, transform the classroom structure, and then build a networked, digital, personalized, and lifelong education system has become the core issue of the current era. "China Education Modernization 2035" clearly states that it is necessary to use modern technologies to accelerate the reform of the talent cultivation model and achieve the organic integration of large-scale education and personalized cultivation. At the same time, the "Action Plan for Education Informatization 2.0" also emphasizes that in the context of "Internet +", it is necessary to build a new talent cultivation model, develop a new model of Internet-based education services, and explore new paths for education governance in the information age. The implementation of intelligent learning is the key to building a new talent cultivation model, and the intelligent learning model provides important support for the realization of intelligent learning. Building an intelligent learning model with complete structure, clear characteristics, and clear logic is a major challenge in implementing intelligent learning. With the continuous progress of technology, especially the increasing maturity of artificial intelligence technologies such as knowledge graphs, deep learning, and pattern recognition, it has laid a solid foundation for the construction of the intelligent learning model. Currently, how to make full use of artificial intelligence technologies to accurately describe the characteristics, constituent elements, and operation mechanism of intelligent learning has become an important topic in the field of education research. The application of information technology in education and teaching shows two levels: Level 1 focuses on technology supporting the original teaching methods and means; Level 2 is committed to technology changing the teaching process, reshaping the teaching structure, and supporting the innovation of teaching methods. However, in primary and secondary school classroom teaching, most teachers still stay at the application level of Level 1, that is, according to the original teaching ideas, select appropriate teaching resources, tools, and systems, and use information technology means for lesson preparation, teaching, examination, and evaluation. Although such practices can improve the efficiency of classroom teaching, it is difficult to significantly improve teaching quality. To truly achieve the deep integration of information technology and education and teaching, it is necessary to apply information technology in the way of Level 2, that is, use technology to solve the bottleneck problems in education and teaching, change the teaching process, reshape the teaching structure, innovate teaching methods, guide students to carry out intelligent learning, and cultivate their core disciplinary qualities, so as to greatly improve teaching quality. In this process, front-line teachers need to be able to use information technology means to help students carry out intelligent learning activities and promote their diverse and comprehensive development of wisdom. However, currently, many teachers still face challenges and it is difficult to make full use of technology means to efficiently guide students to engage in the practice of intelligent learning. In view of this, there is an urgent need to build an intelligent learning model that can clearly reveal the internal structure and operation mechanism of intelligent learning, and use this as a guide to lead the way of teaching reform and innovative development empowered by intelligent technologies.

[0003] Therefore, it is particularly urgent and important to explore how to efficiently integrate cutting-edge intelligent technologies such as artificial intelligence, big data, "Internet +" and virtual simulation to build a smart learning model that can not only support learners to achieve efficient learning and personalized learning paths, but also conduct accurate evaluations and provide high-quality educational resources, while focusing on cultivating key subject abilities and comprehensively promoting learners' development. This move is not only related to the acceleration of the process of educational modernization, but also the core of driving profound changes and innovations in the field of education and cultivating talents that meet the needs of future society.

[0004] Computational intelligence, as an emerging subfield in the field of artificial intelligence, is gradually showing its unique advantages in dealing with complex problems and optimizing decisions. Swarm intelligence, as a soft simulation and emulation technology of group behavior patterns in nature, fully demonstrates the ability to solve complex problems through the collaboration and cooperation of a large number of simple individuals. This ability provides strong robustness and flexibility for problem-solving solutions, making the solutions more stable and adaptable. With the high efficiency and easy implementation of swarm intelligence algorithms, it has shown significant problem optimization potential in the field of intelligent learning and opened up new paths for the development and application of intelligent learning technology. For example, in the field of personalized resource recommendation ([1] Li Haojun. Research on online learning resource serialization service based on particle swarm optimization algorithm [D]. Zhejiang University of Technology, 2018.), swarm intelligence algorithms effectively improve the accuracy of recommendation systems. In addition, in the intelligent planning of learning paths, the dynamic generation of adaptive evaluation materials, and the optimization of learning platform performance, swarm intelligence algorithms have shown strong optimization capabilities and broad application prospects. These research results not only enrich the theoretical framework of the field of intelligent learning, but also inject new vitality and driving force into the development of intelligent education. Through the application of swarm intelligence algorithms, intelligent learning systems can more accurately meet the personalized needs of learners and provide higher-quality and more efficient learning services, thereby promoting intelligent education to continue to move towards a higher stage of development.

[0005] In recent years, inspired by the unique foraging behavior and exploration strategy of honey badgers in nature, the honey badger algorithm has emerged as a novel swarm intelligence optimization method. The algorithm constructs a self-organizing optimization search framework by finely simulating the dynamic interaction between honey badgers based on food source quality and spatial location. The algorithm logic of this framework is simple and efficient, and it shows good adaptability and robustness in a variety of optimization scenarios, providing a new perspective and powerful tool for complex optimization problems. However, the intrinsic optimization mechanism of its core foraging mechanism and stochastic exploration model needs to be further explored. In particular, how to optimize the algorithm's search strategy to strike a balance between global search and local optimization and achieve capability enhancement has become a key challenge that needs to be solved in current research. Solving this problem is related to the improvement of algorithm performance and its scope of application and efficiency in practical problems.

[0006] Therefore, deeply analyzing the foraging mechanism and stochastic optimization mechanism of the honey badger algorithm and optimizing its search strategy are of great significance for improving the algorithm theory and expanding practical applications. In view of the development limitations and broad application prospects of the honey badger algorithm, the present invention focuses on its in-depth innovation and extensible applications, aiming to break through the existing bottlenecks through technological innovation and enhance its effectiveness in the field of intelligent education, especially in the intelligent recommendation of personalized online learning resources.

[0007] To this end, the present invention provides an online learning resource recommendation method and system based on an improved honey badger algorithm. Summary of the Invention

[0008] The object of the present invention is to overcome the deficiencies in the prior art and provide an online learning resource recommendation method and system based on an improved honey badger algorithm. By deeply analyzing the multi-dimensional characteristics of learners and learning resources, an online learning resource recommendation model for deeply analyzing the differences in multi-dimensional characteristics of learners and learning resources is constructed. The evolutionary strategy of the honey badger algorithm is optimized and innovated, and an elite club guiding strategy based on the similarity of time series trajectories is introduced to strengthen the diversity and effectiveness of the algorithm's search path, so as to provide learners with more personalized learning paths and resource recommendations that meet their needs.

[0009] To achieve the above object, the present invention is implemented by the following technical solutions:

[0010] In the first aspect, the present invention provides a personalized learning resource recommendation method based on an improved honey badger population algorithm, including:

[0011] Obtaining the user profile and learning resource characteristics of the learner;

[0012] Inputting the user profile and learning resource characteristics into a pre-trained learning resource recommendation model, and the learning resource recommendation model constructs a multi-dimensional matching fitness function between the learner and the learning resource according to the user profile and learning resource characteristics of the learner;

[0013] Solving the multi-dimensional matching fitness function by an improved honey badger population algorithm and outputting the recommended learning resources;

[0014] Among them, the improved honey badger population algorithm is based on an elite club guiding strategy of time series trajectory similarity. Through this strategy, a sub-optimal individual is randomly selected from the neighboring area of the optimal individual in the population as the population leader to guide the population to conduct a global search.

[0015] Furthermore, the user profile includes the learner's ability level, preference for information carrier type, preference for content type, and upper limit of learning duration, and the learning resource characteristics include learning resource difficulty, information carrier type, content type, and learning duration required for the learning resource.

[0016] Furthermore, by improving the honey badger population algorithm to solve the multi-dimensional matching fitness function and output the recommended learning resources, it includes:

[0017] Randomly initialize the positions of the honey badger population;

[0018] Execute the iterative steps to obtain the final individual positions until the termination conditions are met. The termination conditions include that the multi-dimensional matching function value corresponding to the decision sequence is lower than the threshold or the algorithm reaches the maximum number of iterations;

[0019] When the termination conditions are met, take the final individual positions as the recommended learning resources and output them;

[0020] The iterative steps include:

[0021] Based on the random mapping model, determine the recommended learning resources, and update the historical best individual recurrence statistics value and discovery probability of the honey badger population;

[0022] Based on the historical best individuals of the honey badger population, implement an elite club guidance strategy based on the similarity of temporal trajectories to update the population individuals;

[0023] According to the updated discovery probability, judge whether to perturb the updated population individuals. If so, implement a mutation strategy based on the multi-strategy weight adaptive mechanism to perturb the individuals, and judge whether to retain the perturbed individual positions as the final individual positions. Otherwise, according to the updated historical best individual recurrence statistics value of the honey badger population, judge whether the number of times the historical best individual appears at the current position reaches the threshold. If so, perturb the optimal individual position according to the variable-scale reverse learning strategy integrating the nonlinear dynamic system theory and judge whether to retain the perturbed position as the final individual position. Otherwise, judge whether the termination conditions are reached.

[0024] Furthermore, the multi-dimensional matching fitness function includes six secondary fitness functions, which are respectively expressed as follows:

[0025] (1)

[0026] is the first-level secondary fitness function, representing the difference between the learner's comprehension ability and the difficulty of the recommended resources; represents the th difficulty of the th knowledge point in the th resource; , is the learner's comprehension ability for the

[0027] (2)

[0028] is the secondary fitness function, representing the difference between the information carrier type preference of the learner and the information carrier type of the recommended resource; where is an indicator function used to judge the sequence and at the th position, if , then , otherwise , is the information carrier type weight coefficient; represents the set of information carrier feature identifiers of the th knowledge point in the th resource, , , is the number of information carrier types; being 1 means that the th knowledge point in the th resource is presented in the th information carrier type, being 0 means it is not presented in this type; represents the information carrier type preference of the learner when accepting the th knowledge point, , being 1 means that the learner accepts the th knowledge point presented in the th information carrier type; being 0 means that the learner rejects the knowledge point presented in this media type; is the decision sequence, where , is a decision element. If is 1, it means the th resource is recommended to the learner; if is 0, it means it is not recommended to this learner, represents the time spent learning the th resource;

[0029] (3)

[0031] is the tertiary secondary fitness function, used to evaluate the balance of the difference between the information carrier type preference of the learner and the information carrier type of the recommended resource, is the information carrier type difference variable parameter;

[0032] (4)

[0033] It is the fourth-level secondary fitness function, representing the difference between the learner's preferred material type and the material type of the recommended resources. is the weight coefficient of the material type; represents the th set of material type characteristic identifiers of the th th knowledge point in the is 1, indicating that the th knowledge point in the th resource is presented in the th material type; otherwise th is 0, indicating that it is not presented in this type; represents the learner's preference for the material type when accepting the th knowledge point, is 1, then it is; otherwise is an indicator function used to determine whether the elements at the th position in the sequences and match. If , then ;

[0034] (5)

[0035] is the fifth-level secondary fitness function, used to evaluate the balance of the difference between the learner's preferred material type and the material type of the recommended resources; is the variable parameter of the material type difference;

[0036] (6)

[0037] is the sixth-level secondary fitness function, used to evaluate whether the time required to complete the learning resources exceeds the upper limit of the learner's acceptable learning duration. represents the upper limit of the learner's acceptable learning duration;

[0038] (7)

[0039] is a multi-dimensional matching fitness function, , , , , and are the weight coefficients of the secondary fitness function; The smaller the [[value]] is, the higher the quality of the recommended learning resource is, and the more suitable it is to be recommended to learners. On the contrary, the higher the [[value]] is, the less suitable the learning resource is to be recommended to learners.

[0040] Furthermore, the determination of the learning resources to be recommended based on the random mapping model includes:

[0041] Mapping the discrete element values in the honey badger individual position vector to the binary space. The th learning resource decision sequence is generated by mapping the corresponding honey badger individual position vector , where , , is the population size, is the th individual's learning resource decision sequence at the j-th iteration, is the th individual's position vector at the j-th iteration, The th element in the position vector;

[0042] Combined with the random mapping model, the learning resource decision sequence is expressed as follows:

[0043] (8)

[0044] In equation (8), is a random number with a value range between [0, 1];

[0045] Combined with the decision sequence result in equation (8), calculate the fitness function value of the learning resource corresponding to the decision sequence according to the matching fitness function formula (7);

[0046] Sort the fitness function values of each learning resource from smallest to largest. The fitness function value corresponding to the learning resource to be recommended needs to be the historical minimum, and the individual corresponding to its decision sequence is the historical optimal individual of the honey badger population, denoted as , ;

[0047] The update of the historical optimal individual recurrence statistic value of the honey badger population includes:

[0048] If in the position of the historical optimal individual of the honey badger population, the The absolute value difference between the value of the dimension and the value of the dimension in the historically optimal individual position of the honey badger population in the previous iteration is less than the recurrence statistical value parameter , then the statistical value is processed, otherwise maintain the initial value; the recurrence statistical value of the dimension of the historically optimal individual in the honey badger population is , and the initial value of is 0, the threshold is set to randomly generate dimensional vectors between [0, 1] as the initial positions of the honey badger population;

[0049] The update formula for the discovery probability is as follows:

[0050] (9)

[0051] is the discovery probability at the th iteration, is the maximum number of iterations.

[0052] Furthermore, the implementation of the elite club guidance strategy based on temporal trajectory similarity to update the population individuals includes:

[0053] Calculate the temporal trajectory similarity between the population individuals and the optimal individual , sort the similarity values from large to small, select the positions of the top Tp% of the individuals as candidate solutions, where Tp is the proportion of candidate solutions for temporal trajectory similarity, and randomly select one solution from the candidate solutions as the approximate optimal individual in the current iteration , , is the value of the dimension in the position of the approximate optimal individual in the current iteration;

[0054] The formula for calculating the temporal trajectory similarity between the individuals in the population and the optimal individual is as follows:

[0055] (10)

[0056] is the temporal trajectory similarity between the individuals in the population and the optimal individual;

[0057] Randomly generate , with a value range of [0, 1]. If , the position update formula for the random population individuals is as follows:

[0058] (11)

[0060] like , the position update formula of random population individuals is as follows:

[0061] (12)

[0062] in, , indicating random output -1 or 1, , , and They are all random numbers in the range [0,1];

[0063] Among them, in formulas (10) to (12), ;like , (11) where Replaced by .

[0064] Furthermore, judging whether to disturb the updated population individuals according to the updated discovery probability includes:

[0065] Randomly generate detection comparison parameters , ,judge Is it less than the updated discovery probability? If so, then The individuals at the iteration are perturbed;

[0066] The specific operation of implementing the mutation strategy perturbation individual based on the multi-strategy weight adaptive mechanism is:

[0067] The individual position update formula is as follows:

[0068] (13)

[0069] in, , is the individual position after disturbance; is a random number in the range [-1,1]. is a random number in the range [0,1];

[0070] The determination of whether to retain the disturbed individual comprises:

[0071] According to formula (8) and formula (7) respectively, the fitness function value of the individual after disturbance is calculated. If the fitness function value of the individual after disturbance is less than the value before disturbance, the position after disturbance is retained, otherwise the original position remains unchanged. That is, the basis for judging whether the position of the individual is updated is as follows:

[0072] (14)

[0073] The detailed steps for determining whether to perturb the optimal individual position according to the variable-scale reverse learning strategy of the fused nonlinear dynamic system theory are as follows:

[0074] Determine whether it reaches the threshold If so, perturb the optimal individual position according to the variable-scale reverse learning strategy of the fused nonlinear dynamic system theory;

[0075] Perturbing the optimal individual position according to the variable-scale reverse learning strategy of the fused nonlinear dynamic system theory includes:

[0076] (15)

[0078] Among them, is the control parameter of the nonlinear dynamic system, , is the optimal individual position after perturbation, , is a random number with a value range between [0, 1];

[0079] Determining whether to retain the position after perturbation includes: successively calculating the fitness function values of the historical optimal individuals before and after perturbation according to Equation (8) and Equation (7). If the fitness function value after perturbation is less than the value before perturbation, retain the position after perturbation; otherwise, keep the original position unchanged, that is, determining whether to update the position of the historical optimal individual is based on the following equation:

[0080] (16)

[0081] The specific operation for determining whether there is a multi-dimensional matching function value corresponding to the decision sequence lower than the threshold is as follows: combining Equation (7), calculate the multi-dimensional matching fitness function value corresponding to the individual, and determine whether the minimum function value is lower than the matching threshold .

[0082] On the second aspect, the present invention provides an online personalized learning resource recommendation system based on an improved honey badger population algorithm, including:

[0083] An acquisition module: used to acquire the user profile and learning resource characteristics of the learner;

[0084] Output module: It is used to input the user profile and learning resource features into a pre-trained learning resource recommendation model. The learning resource recommendation model constructs a multi-dimensional matching fitness function between learners and learning resources based on the user profile and learning resource features of learners; solves the multi-dimensional matching fitness function through an improved honey badger population algorithm and outputs the recommended learning resources. Among them, the improved honey badger population algorithm is based on an elite club guiding strategy of temporal trajectory similarity. Through this strategy, sub-optimal individuals are randomly selected from the neighboring area of the optimal individual in the population as the population leader to guide the population to search globally.

[0085] In a third aspect, the present invention provides a computer device, including a processor and a storage medium;

[0086] The storage medium is used to store instructions;

[0087] The processor is used to operate according to the instructions to execute the steps of the method according to the first aspect.

[0088] In a fourth aspect, the present invention provides a computer-readable storage medium, on which a computer program is stored. It is characterized in that when the program is executed by a processor, the steps of the method according to the first aspect are realized.

[0089] Compared with the prior art, the beneficial effects achieved by the present invention:

[0090] The present invention relates to the cross field of educational technology and swarm intelligence algorithms. Specifically, it is a technical solution aimed at improving the matching degree between the learning path and the personalized needs of learners, optimizing the online learning experience, and solving the problem of accurate recommendation of learning resources. By deeply analyzing the multi-dimensional features of learners and learning resources, the present invention innovatively uses an improved honey badger algorithm to promote the development of personalized learning in the context of "Internet +" education, accelerate the progress of technology-driven online education, and realize a more accurate and efficient online learning mode.

[0091] The present invention first constructs an online learning resource recommendation model, which deeply explores the multi-dimensional feature differences between learners and learning resources. The core of this model is that it integrates the learner preferences estimated based on the element weight matching technology as the key features. In view of the problem that the traditional honey badger algorithm is prone to fall into local optimal solutions during the search process, thus causing premature convergence, the present invention optimizes and innovates the evolutionary strategy of the algorithm to improve its global search ability and recommendation accuracy.

[0092] Specifically, based on the diversified leadership decision-making mechanism, the present invention introduces an elite club guiding strategy based on the similarity of time-series trajectories. This strategy randomly selects sub-optimal individuals from the neighboring regions of the population's optimal individuals as the population leaders to guide the population to conduct searches globally. This strategy fully explores and utilizes the diversity characteristics of the dominant solutions, effectively enhancing the path diversity during the search process of the algorithm and significantly improving the search efficiency. To further enhance the optimization performance of the algorithm, the present invention proposes an elimination mechanism that incorporates the discovery probability, aiming to effectively eliminate inferior solutions and thereby enhance the diversity of the population. Specifically, a mutation strategy integrating a multi-strategy weight adaptive mechanism is incorporated in the present invention. In the initial stage of the algorithm iteration, this strategy endows the population with strong global search capabilities, enabling the algorithm to quickly locate potential high-quality solution sets within a vast search space. As the iteration progresses, by dynamically adjusting the discovery probability, the population is enabled to possess superior local search capabilities in the later stage of iteration, thereby effectively improving the accuracy and convergence speed of the solutions. In addition, to significantly enhance the anti-stagnation characteristics of the algorithm, the present invention introduces a variable-scale reverse learning strategy integrating the theory of nonlinear dynamic systems. This strategy aims to enhance the algorithm's ability to break through the constraints of local extrema and improve the robustness of global optimization. During the generation of reverse solutions, this strategy comprehensively considers the dynamic global information of the population. By combining the chaotic perturbation term with the adaptive reverse solution generation mechanism, the distribution range of the population within the search space is effectively expanded, ensuring that the algorithm can traverse the solution space more comprehensively and achieve a full exploration of potential high-quality solutions. The simulation experiment results show that the improved honey badger algorithm proposed by the present invention demonstrates good application prospects in the field of personalized online learning resource recommendation. Its recommendation matching degree is superior to the existing learning resource recommendation methods based on swarm intelligence optimization algorithms, providing more efficient and robust technical support for the precise recommendation of online education resources. BRIEF DESCRIPTION OF THE DRAWINGS

[0093] Figure 1 is the algorithm flowchart of the present invention;

[0094] Figure 2 is the comparison of the convergence curves of different algorithms in each environment in the embodiment of the present invention;

[0095] Figure 3 is the schematic flowchart of an online learning resource recommendation method based on an improved honey badger population algorithm provided by the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0096] The technical solutions of the present invention will be described in detail below through the accompanying drawings and specific embodiments. It should be understood that the specific features in the embodiments of the present application and the embodiments are detailed descriptions of the technical solutions of the present application, rather than limitations on the technical solutions of the present application. Without conflict, the technical features in the embodiments of the present application and the embodiments can be combined with each other.

[0097] In this text, the term "and / or" is merely a description of the relationship between associated objects, indicating that there can be three relationships. For example, A and / or B can represent three situations: A exists alone, A and B exist simultaneously, and B exists alone. Additionally, in this text, the character " / " generally indicates that the associated objects before and after are in an "or" relationship.

[0098] Figure 3 It is a flowchart of an online learning resource recommendation method based on an improved honey badger population algorithm in the first embodiment of the present invention. The online learning resource recommendation method provided in this embodiment can be applied to a terminal and can be executed by an online learning resource recommendation system based on the improved honey badger population algorithm. This system can be implemented in software and / or hardware and can be integrated into the terminal. For example: any smart phone, tablet computer or computer device with communication functions. See Figure 3 , the method of this embodiment includes:

[0099] Obtain the user portrait and learning resource characteristics of the learner;

[0100] Obtain the user portrait and learning resource characteristics of the learner;

[0101] Input the user portrait and learning resource characteristics into a pre-trained learning resource recommendation model. The learning resource recommendation model constructs a multi-dimensional matching fitness function between the learner and the learning resources according to the user portrait and learning resource characteristics of the learner;

[0102] Solve the multi-dimensional matching fitness function through the improved honey badger population algorithm and output the recommended learning resources;

[0103] Among them, the improved honey badger population algorithm is based on an elite club guiding strategy of temporal trajectory similarity. Through this strategy, a sub-optimal individual is randomly selected from the neighboring area of the optimal individual in the population as the population leader to guide the population to search globally;

[0104] The user portrait includes the learner's ability level, information carrier type preference, content type preference, and learning duration limit. The learning resource characteristics include learning resource difficulty, information carrier type, content type, and learning duration required for the learning resource;

[0105] Solving the multi-dimensional matching fitness function through the improved honey badger population algorithm and outputting the recommended learning resources includes:

[0106] Randomly initialize the positions of the honey badger population;

[0107] Execute iterative steps to obtain the final individual position until the termination condition is met. The termination condition includes that there is a multi-dimensional matching function value corresponding to the decision sequence lower than the threshold or the algorithm reaches the maximum number of iterations;

[0108] When the termination condition is met, take the final individual position as the recommended learning resource and output it;

[0109] The iterative steps include:

[0110] Determine the recommended learning resource based on the random mapping model, and update the historical optimal individual recurrence statistic value and discovery probability of the honey badger population;

[0111] Based on the historical optimal individual of the honey badger population, implement an elite club guidance strategy based on the similarity of time series trajectories to update the population individuals;

[0112] According to the updated discovery probability, judge whether to perturb the updated population individuals. If so, implement a mutation strategy based on the multi-strategy weight adaptive mechanism to perturb the individuals, and judge whether to retain the perturbed individual position as the final individual position. Otherwise, according to the updated historical optimal individual recurrence statistic value of the honey badger population, judge whether the number of times the historical optimal individual appears at the current position reaches the threshold. If so, perturb the optimal individual position according to the variable-scale reverse learning strategy integrating the nonlinear dynamic system theory and judge whether to retain the perturbed position as the final individual position. Otherwise, judge whether the termination condition is reached;

[0113] As Figure 1 shown, the method specifically includes: Step A: Construct a multi-dimensional matching fitness function according to the characteristics of learners and learning resources. Set the matching threshold, population size, maximum number of algorithm iterations, historical optimal individual recurrence statistic value of the honey badger population, threshold, and randomly initialize the positions of the honey badger population.

[0114] Step B: Determine the recommended learning resource based on the random mapping model; update the historical optimal individual recurrence statistic value of the honey badger population;

[0115] Step C: Update the discovery probability;

[0116] Step D: Implement an elite club guidance strategy based on the similarity of time series trajectories to update the population individuals;

[0117] Step E: Judge whether to perturb the individuals according to the discovery probability. If so, execute Step F, otherwise execute Step G;

[0118] Step F: Implement a mutation strategy based on the multi-strategy weight adaptive mechanism to perturb the individuals, and judge whether to retain the perturbed individuals;

[0119] Step G: Determine whether the number of times the historical optimal individual appears at the current position reaches the threshold. If so, perturb the position of the optimal individual according to the variable-scale reverse learning strategy integrating the non-linear dynamic system theory and determine whether to retain the perturbed position. Otherwise, execute Step H;

[0120] Step H: Determine whether there is a multi-dimensional matching function value corresponding to the decision sequence that is lower than the threshold or whether the algorithm has reached the maximum number of iterations. If either condition is met, execute Step I. Otherwise, execute Step B;

[0121] Step I: Determine the recommended learning resources.

[0122] Step Aa: is the learner's comprehension ability for the th knowledge point. , is the number of knowledge points, represents the th knowledge point in the th resource, , is the number of resources, represents the set of information carrier feature identifiers of the th knowledge point in the th resource, , , is the number of information carrier types, being 1 indicates that the th knowledge point in the th resource is presented in the th information carrier type. Otherwise, 0 indicates that it is not presented in this type. represents the learner's preference for the information carrier type when receiving the th knowledge point, , being 1 indicates that the learner accepts the th knowledge point presented in the th information carrier type. Otherwise, 0 indicates that the learner rejects the knowledge point being presented in this information carrier type. represents the set of material type feature identifiers of the th knowledge point in the th resource, , , is the number of knowledge material types, being 1 indicates that the th knowledge point in the th resource is presented in the th material type. Otherwise, 0 indicates that it is not presented in this type. Indicates the preference for the type of material when the learner accepts the th knowledge point, , with 1 indicating that the learner accepts the th knowledge point presented in the th type of material, and conversely, 0 indicates that the learner rejects the knowledge point presented in that type of material. Indicates the time taken to learn the th resource, and represents the upper limit of the learning duration acceptable to the learner. , is the decision sequence, where , if is 1, it means the th resource is recommended to the learner, and if

[0123] (1)

[0124] represents the difference between the learner's comprehension ability and the difficulty of the recommended resource.

[0125] (2)

[0126] represents the difference between the learner's preference for the type of information carrier and the type of information carrier of the recommended resource. Among them, is an indicator function used to determine whether the elements in the sequences and at the th position match. If , then , otherwise , is the weight coefficient of the information carrier type.

[0127] (3)

[0129] evaluates the balance of the difference between the learner's preference for the type of information carrier and the type of information carrier of the recommended resource, is the variable parameter of the information carrier type difference.

[0130] (4)

[0131] Indicates the difference between the learner's preference for material types and the material types of the recommended resources. is the weight coefficient of the material type.

[0132] (5)

[0133] is to evaluate the balance of the difference between the learner's preference for material types and the material types of the recommended resources. is the variable parameter of the material type difference.

[0134] (6)

[0135] is to evaluate whether the time required to complete the learning resources exceeds the upper limit of the learner's acceptable learning duration.

[0136] (7)

[0137] Equation (6) is the multi-dimensional matching fitness function, where 、 、 、 、 and are the weight coefficients of the secondary fitness functions. Let the matching threshold be . The smaller the value, the higher the quality of the recommended learning resources and the more suitable they are for recommendation to learners. Conversely,

[0138] Step Ab: is the population size, and the maximum number of iterations is set. is the current number of iterations. The -dimensional recurrence statistic value of the historical best individual in the honey badger population is , and 、the threshold is set to . Randomly generate -dimensional vectors between [0,1] as the initial positions of the honey badger population.

[0139] Step Ba: Map the discrete element values in the honey badger individual position vector to the binary space. The th learning resource decision sequence is generated by mapping the corresponding honey badger individual position vector , where , . Combining the random mapping model, the learning resource decision sequence is expressed as follows:

[0140] (8)

[0141] In equation (8), is a random number with a value range between [0, 1].

[0142] Combined with the decision sequence result in equation (8), calculate the fitness function value of the learning resource corresponding to the decision sequence according to the matching fitness function equation (7). Sort the fitness function values of each learning resource from small to large. The fitness function value corresponding to the recommended learning resource needs to be the smallest in history. The individual corresponding to its decision sequence is the historical optimal individual of the honey badger population, denoted as , .

[0143] Step Bb: If the value of the th dimension in the position of the historical optimal individual of the honey badger population is close to the value of the th dimension in the position of the historical optimal individual of the honey badger population in the previous iteration process, that is, the absolute value difference between the two values is less than , then perform processing on the statistical value , otherwise .

[0144] The specific operation of updating the discovery probability is as follows:

[0145] (9)

[0146] Step D: Randomly select an individual with a relatively high comprehensive similarity to as .

[0147] The formula for calculating the temporal trajectory similarity between an individual in the population and the optimal individual is as follows:

[0148] (10)

[0149] Calculate the temporal trajectory similarity between the population individual and the optimal individual according to equation (11), sort the similarity values from large to small, select the positions of the top Tp% of the individuals as candidate solutions, and randomly select one solution from them as , ;

[0150] Randomly generate , with a value range of [0, 1]. If , the position update formula for the random population individual is as follows:

[0151] (11)

[0153] If , the position update formula for the random population individual is as follows:

[0154] (12)

[0155] Among them, , indicating a random output of -1 or 1, , , and are all random numbers with values in the range [0, 1];

[0156] Among them, in formulas (10) to (12), ; if , in formula (11) is replaced by .

[0157] Step E: Randomly generate detection and comparison parameters , , and judge whether it is less than the discovery probability ; if so, perturb the individual in the population.

[0158] Step Fa: The position update formula for an individual is as follows:

[0159] (13)

[0160] Among them, , is the position of the individual after perturbation; is a random number with a value range between [-1, 1], is a random number with a value range between [0, 1].

[0161] Step Fb: Sequentially calculate the fitness function value of the individual after perturbation according to formula (8) and formula (7). If the fitness function value of the individual after perturbation is less than the value before perturbation, then retain the position after perturbation; otherwise, keep the original position unchanged, that is, judge whether the position of the individual is updated according to the following formula:

[0162] (14)

[0163] Step Ga: Judge whether it reaches the threshold ; if so, perturb the position of the optimal individual according to the variable-scale reverse learning strategy integrating the nonlinear dynamic system theory.

[0164] Step Gb: The specific operation of perturbing the position of the optimal individual according to the variable-scale reverse learning strategy integrating the nonlinear dynamic system theory is as follows:

[0165] (15)

[0167] in, is the control parameter of the nonlinear dynamic system, , is the optimal individual position after disturbance, , is a random number in the range [0,1].

[0168] Step Gc: According to formula (8) and formula (7), calculate the fitness function value of the historical best individual before and after the disturbance. If the fitness function value after the disturbance is less than the value before the disturbance, the position after the disturbance is retained, otherwise the original position remains unchanged. That is, the position of the historical best individual is updated according to the following formula:

[0169] (16)

[0170] Step Ha: Combined with formula (7), calculate the multidimensional matching fitness function value corresponding to the individual, and determine whether the minimum function value is lower than the matching threshold .

[0171] Step Hb: Determination Is it equal to the maximum number of iterations? .

[0172] Step I: According to formula (7), the multi-dimensional matching fitness function value corresponding to the decision sequence of the recommended learning resource is the minimum.

[0173] S10: Construct multi-dimensional matching fitness function and set population size

[0174] For learners The ability to understand knowledge points. , is the number of knowledge points, Indicates Resources The difficulty of each knowledge point, , is the number of resources, Indicates Resources The information carrier feature identifier set of knowledge points, , , is the number of information carrier types, 1 means the Resources Knowledge points are If it is 0, it means that the information carrier type is not presented. Indicates that learners accept The information carrier type preference when learning a knowledge point , 1 indicates that the learner accepts the th knowledge point presented in the th information carrier type. Otherwise, 0 indicates that the learner rejects the knowledge point presented in this information carrier type. Indicates the set of material type characteristic identifiers of the th knowledge point in the th resource, , , where is the number of knowledge material types, 1 indicates that the th knowledge point in the th resource is presented in the th material type. Otherwise, 0 indicates that it is not presented in this type. Indicates the material type preference of the learner when accepting the th knowledge point, 1 indicates that the learner accepts the th knowledge point presented in the th material type. Otherwise, 0 indicates that the learner rejects the knowledge point presented in this material type. Indicates the time taken to learn the th resource, where is the decision sequence, where , is the decision element. If is 1, it indicates that the th resource is recommended to the learner. If is 0, it indicates that it is not recommended to this learner. To match personalized learning resources with learner characteristics and improve the quality of personalized learning resource generation services, 7 secondary fitness functions are established and are publicly announced as follows:

[0175] (1)

[0176] Indicates the difference between the learner's comprehension ability and the difficulty of the recommended resources.

[0177] (2)

[0178] Indicates the difference between the learner's information carrier type preference and the information carrier type of the recommended resources. Among them, is the indicator function used to judge the sequence and Whether the element at the -th position matches. If , then , otherwise , is the weight coefficient of the information carrier type.

[0179] (3)

[0181] is to evaluate the balance of the difference between the information carrier type preference of the learner and the information carrier type of the recommended resource. is the variable parameter of the information carrier type difference.

[0182] (4)

[0183] represents the difference between the material type preference of the learner and the material type of the recommended resource. is the weight coefficient of the material type.

[0184] (5)

[0185] is to evaluate the balance of the difference between the material type preference of the learner and the material type of the recommended resource. is the variable parameter of the material type difference.

[0186] (6)

[0187] is to evaluate whether the time required to complete the learning resource exceeds the upper limit of the acceptable learning duration of the learner.

[0188] (7)

[0189] Equation (6) is the multi-dimensional matching fitness function, where , , , , and are the weight coefficients of the secondary fitness functions. Let the matching threshold be . The smaller the value, the higher the quality of the recommended learning resource, and the more suitable it is to be recommended to the learner. On the contrary, the higher the

[0190] S11: Set the matching threshold, the maximum number of iterations of the algorithm, and randomly initialize the positions of the honey badger population

[0191] Set the maximum number of iterations , is the current number of iterations, and the -dimensional recurrence statistical value of the best individual in the honey badger population history is , and The threshold is set to , randomly generate [0,1] -dimensional vectors as the initial positions of the honey badger population.

[0192] S12: Determine the recommended learning resources based on the random mapping model

[0193] Map the discrete element values in the honey badger individual position vector to the binary space. The th learning resource decision sequence is generated by mapping the corresponding honey badger individual position vector , where , , combined with the random mapping model, the learning resource decision sequence is expressed as follows:

[0194] (8)

[0195] In formula (8), is a random number with a value range between [0,1].

[0196] Combined with the decision sequence result in formula (8), calculate the fitness function value of the learning resource corresponding to the decision sequence according to the matching fitness function formula (7). Sort the fitness function values of each learning resource from small to large. The fitness function value corresponding to the recommended learning resource needs to be the smallest in history, and its decision sequence corresponding individual is the best individual in the honey badger population history, expressed as , .

[0197] S13: Update the recurrence statistical value of the best individual in the honey badger population history

[0198] If the value of the -dimensional in the position of the best individual in the honey badger population history is close to the value of the -dimensional in the position of the best individual in the honey badger population history during the previous iteration, that is, the absolute value difference between the two values is less than , then perform processing on the statistical value , otherwise .

[0199] S14: Update the discovery probability

[0200] The specific operation to update the discovery probability is:

[0201] (9)

[0202] S15: Implement the elite club guiding strategy based on the similarity of temporal trajectories to update the population individuals

[0203] Randomly select and Individuals with relatively high comprehensive similarity as .

[0204] The calculation formula for the similarity of temporal trajectories between an individual in the population and the optimal individual is as follows:

[0205] (10)

[0206] Calculate the similarity of temporal trajectories between the population individuals and the optimal individual according to formula (10), sort the similarity values from large to small, select the positions of the top Tp% of the individuals as candidate solutions, and randomly select one solution from them as , ;

[0207] Randomly generate , with a value range of [0, 1]. If , the position update formula for the random population individuals is as follows:

[0208] (11)

[0210] If , the position update formula for the random population individuals is as follows:

[0211] (12)

[0212] Among them, indicates randomly outputting -1 or 1, , , and are all random numbers with a value range between [0, 1];

[0213] Among them, in formulas (10) to (12), ; if , in formula (11), is replaced by .

[0214] S16: Judge whether to perturb the individual according to the discovery probability

[0215] Randomly generate the detection and comparison parameter , , judge whether it is less than the discovery probability . If so, execute S17.

[0216] S17: Implement a mutation strategy to perturb individuals based on a multi-strategy weight adaptive mechanism

[0217] (13)

[0218] Wherein, , is the position of the perturbed individual; is a random number with a value range between [-1, 1], is a random number with a value range between [0, 1].

[0219] S18: Determine whether the fitness value of the perturbed individual becomes smaller

[0220] Successively according to Equation (8) and Equation (7), calculate the fitness function value of the individual after perturbation. If the fitness function value of the individual after perturbation is less than the value before perturbation, execute S19; otherwise, execute S20.

[0221] S19: Retain the position of the individual after perturbation

[0222] The position of the individual is updated to the position after perturbation, which is equivalent to updating the position of the population individuals according to Equation (10).

[0223] (14)

[0224] S20: The position of the individual remains unchanged

[0225] Equivalent to Equation (15), the individual retains the position before perturbation unchanged.

[0226] (15)

[0227] S21: Determine whether the number of times the historical optimal individual appears at the current position reaches the threshold

[0228] Judge whether it reaches the threshold , if so, execute S22; otherwise, execute S26.

[0229] S22: Perturb the position of the historical optimal individual according to the variable-scale reverse learning strategy integrating the dynamic step size (16)

[0231] Wherein, is the control parameter of the nonlinear dynamic system, , is the position of the optimal individual after perturbation, , is a random number with a value range between [0, 1].

[0232] S23: Determine whether the fitness value of the perturbed historical best individual becomes smaller

[0233] Calculate the fitness function values of the historical best individual before and after perturbation successively according to Equation (8) and Equation (7). If the fitness function value after perturbation is less than the value before perturbation, execute S24; otherwise, execute S25.

[0234] S24: Retain the position of the historical best individual after perturbation

[0235] Updating the individual position to the position after perturbation is equivalent to updating the position of the population individuals according to Equation (17).

[0236] (17)

[0237] S25: The position of the historical best individual remains unchanged

[0238] Equivalent to Equation (18), the individual remains in the position before perturbation.

[0239] (18)

[0240] S26: Determine whether the multi-dimensional matching function value is lower than the threshold or whether the algorithm has reached the maximum number of iterations

[0241] Combined with Equation (7), calculate the multi-dimensional matching fitness function value corresponding to the individual, and determine whether the minimum function value of the population is lower than the matching threshold , determine whether it is equal to the maximum number of iterations .

[0242] S27: Determine whether the multi-dimensional matching function value is lower than the threshold or whether the algorithm has reached the maximum number of iterations

[0243] The multi-dimensional matching fitness function value corresponding to the decision sequence of the recommended learning resource is the minimum.

[0244] Experiment: Select the example in the reference (Li Haojun. Research on the Serialization Service of Online Learning Resources Based on Particle Swarm Optimization Algorithm [D]. Zhejiang University of Technology, 2018.) for comparison with the results of the present invention.

[0245] In order to verify the performance of the improved honey badger population algorithm (IHBA) in the personalized learning resource recommendation problem in the present invention, 4 groups of experiments were implemented. The unified parameter part is set as shown in Table 1, and the specific learner characteristics and learning resource characteristics are shown in Table 2.

[0246] Table 1 Parameter Settings of the Improved Honey Badger Population Algorithm in the Present Invention

[0247]

[0248] Table 2 Learner Characteristics and Learning Resource Characteristics

[0249]

[0250] Under the conditions of different learner characteristics and learning resource characteristics, the fitness function values corresponding to the recommended learning resources obtained by each technology are as Figure 2 shown. It can be seen that when the IHBA algorithm solves the problem of personalized learning resource recommendation, the multi-dimensional matching degree corresponding to the recommended learning resources is better than that of the competitive algorithms.

[0251] Example Two:

[0252] An online learning resource recommendation system based on an improved honey badger algorithm, comprising:

[0253] An acquisition module: used to acquire the user profile of the learner and the characteristics of learning resources;

[0254] An output module: used to input the user profile and the characteristics of learning resources into a pre-trained learning resource recommendation model. The learning resource recommendation model constructs a multi-dimensional matching fitness function between the learner and the learning resources according to the user profile of the learner and the characteristics of learning resources; solves the multi-dimensional matching fitness function through an improved honey badger population algorithm and outputs the recommended learning resources. Among them, the improved honey badger population algorithm is based on an elite club guiding strategy of time series trajectory similarity. Through this strategy, sub-optimal individuals are randomly selected from the neighboring area of the population optimal individual as the population leader to guide the population to search globally.

[0255] The online learning resource recommendation system based on the improved honey badger algorithm provided by the embodiment of the present invention can execute the online learning resource recommendation method provided by any embodiment of the present invention, and has the corresponding functional modules and beneficial effects for executing the method.

[0256] Example Three:

[0257] The embodiment of the present invention also provides a computer device, comprising a processor and a storage medium;

[0258] The storage medium is used to store instructions;

[0259] The processor is used to operate according to the instructions to execute the steps of the following method:

[0260] Acquire the user profile of the learner and the characteristics of learning resources;

[0261] Input the user portrait and learning resource features into a pre-trained learning resource recommendation model, which constructs a multi-dimensional matching fitness function between learners and learning resources based on the user portrait and learning resource features of the learners;

[0262] Solve the multi-dimensional matching fitness function through an improved honey badger population algorithm and output the recommended learning resources; wherein, the improved honey badger population algorithm is based on an elite club guiding strategy of temporal trajectory similarity, and through this strategy, randomly select sub-optimal individuals from the adjacent area of the optimal individual in the population as the population leader to guide the population to search globally.

[0263] For the detailed content of each step in this embodiment, reference can be made to Embodiment 1, which will not be elaborated here. In view of the fact that this embodiment and Embodiment 1 adopt the same technical concept, it also has the technical effects such as those described in Embodiment 1.

[0264] Embodiment 4:

[0265] The embodiment of the present invention also provides a computer-readable storage medium, on which a computer program is stored, and when the program is executed by a processor, the steps of the following method are implemented:

[0266] Obtain the user portrait and learning resource features of the learner;

[0267] Input the user portrait and learning resource features into a pre-trained learning resource recommendation model, which constructs a multi-dimensional matching fitness function between learners and learning resources based on the user portrait and learning resource features of the learners;

[0268] Solve the multi-dimensional matching fitness function through an improved honey badger population algorithm and output the recommended learning resources; wherein, the improved honey badger population algorithm is based on an elite club guiding strategy of temporal trajectory similarity, and through this strategy, randomly select sub-optimal individuals from the adjacent area of the optimal individual in the population as the population leader to guide the population to search globally.

[0269] For the detailed content of each step in this embodiment, reference can be made to Embodiment 1, which will not be elaborated here. In view of the fact that this embodiment and Embodiment 1 adopt the same technical concept, it also has the technical effects such as those described in Embodiment 1.

[0270] Those skilled in the art should understand that the embodiments of the present application can be provided as a method, a system, or a computer program product. 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.

[0271] The present application is described with reference to the flowcharts and / or block diagrams of methods, apparatuses (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 can be implemented by computer program instructions, and the combination of the flows and / or blocks in the flowchart and / or block diagram can also be implemented. 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, such that the instructions executed by the processor of the computer or other programmable data processing devices generate means for implementing the functions specified in Figure 1 one or more of the flows Figure 1 or blocks.

[0272] 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, such that the instructions stored in the computer-readable memory generate a manufactured article including instruction means that implement the functions specified in Figure 1 one or more of the flows Figure 1 or blocks.

[0273] These computer program instructions can also be loaded onto a computer or other programmable data processing device, such that a series of operation steps are executed on the computer or other programmable device to generate a computer-implemented process, and thus the instructions executed on the computer or other programmable device provide steps for implementing the functions specified in Figure 1 one or more of the flows Figure 1 or blocks.

[0274] The above is only the preferred embodiment of the present invention. It should be noted that for those of ordinary skill in the art, without departing from the technical principle of the present invention, several improvements and modifications can be made, and these improvements and modifications should also be regarded as the protection scope of the present invention.

Claims

1. An online learning resource recommendation method based on an improved honey badger population algorithm, characterized in that, Including: Obtain the user profile and learning resource characteristics of the learner; Input the user profile and learning resource characteristics into a pre-trained learning resource recommendation model, and the learning resource recommendation model constructs a multi-dimensional matching fitness function between the learner and the learning resource according to the user profile and learning resource characteristics of the learner; Solve the multi-dimensional matching fitness function by improving the honey badger population algorithm and output the recommended learning resources; Among them, the improved honey badger population algorithm is based on an elite club guiding strategy based on temporal trajectory similarity. Through this strategy, sub-optimal individuals are randomly selected from the neighboring area of the population's optimal individual as the population leader to guide the population to search globally.

2. The online learning resource recommendation method based on the improved honey badger population algorithm according to claim 1, characterized in that, The user profile includes the learner's ability level, information carrier type preference, content type preference, and learning duration limit, and the learning resource characteristics include learning resource difficulty, information carrier type, content type, and learning resource required duration.

3. The online learning resource recommendation method based on the improved honey badger population algorithm according to claim 2, characterized in that, Solving the multi-dimensional matching fitness function by improving the honey badger population algorithm and outputting the recommended learning resources includes: Randomly initialize the positions of the honey badger population; Execute iterative steps to obtain the final individual positions until the termination conditions are met. The termination conditions include that the multi-dimensional matching function value corresponding to the decision sequence is lower than the threshold or the algorithm reaches the maximum number of iterations; When the termination conditions are met, take the final individual positions as the recommended learning resources and output them; The iterative steps include: Determine the recommended learning resources based on the random mapping model, and update the historical optimal individual recurrence statistics value and discovery probability of the honey badger population; Based on the historical optimal individual of the honey badger population, implement an elite club guiding strategy based on temporal trajectory similarity to update the population individuals; Judge whether to perturb the updated population individuals according to the updated discovery probability. If so, implement a mutation strategy based on a multi-strategy weight adaptive mechanism to perturb the individuals, and judge whether to retain the perturbed individual positions as the final individual positions. Otherwise, judge whether the number of times the historical optimal individual appears at the current position reaches the threshold according to the updated historical optimal individual recurrence statistics value of the honey badger population. If so, perturb the optimal individual position according to the variable scale reverse learning strategy integrating the theory of nonlinear dynamic systems and judge whether to retain the perturbed position as the final individual position. Otherwise, judge whether the termination conditions are reached.

4. The online learning resource recommendation method based on the improved honey badger population algorithm according to claim 3, wherein, The multi-dimensional matching fitness function includes six secondary fitness functions, which are respectively expressed as follows: (1) is the primary-secondary fitness function, representing the difference between the learner's comprehension ability and the difficulty of the recommended resources; represents the th knowledge point in the th resource; is the learner's comprehension ability of the th knowledge point; , is the number of knowledge points, , is the number of resources; (2) is the secondary sub - fitness function, representing the difference between the information carrier type preference of the learner and the information carrier type of the recommended resource; among them, is the indicator function, used to judge the sequence and at the th position, whether the elements match. If , then , otherwise . is the information carrier type weight coefficient; represents the set of information carrier feature identifiers of the th knowledge point in the rd resource, , , is the number of information carrier types; being 1 means that the rd knowledge point in the th resource is presented in the th information carrier type. being 0 means that it is not presented in this type; represents the information carrier type preference of the learner when accepting the th knowledge point, , being 1 means that the learner accepts the th knowledge point presented in the th information carrier type; being 0 means that the learner rejects the knowledge point being presented in this media type; is the decision sequence, where , is the decision element. If is 1, it means that the rd resource is recommended to the learner; if is 0, it means that it is not recommended to this learner, represents the time spent on learning the th resource; (3)； It is a three-level secondary fitness function, which is used to evaluate the balance of the difference between the information carrier type preference of the learner and the information carrier type of the recommended resources. It is a variable parameter for the difference in information carrier type; (4) is the fourth-level secondary fitness function, representing the difference between the learner's preference for the data type and the data type of the recommended resource. is the data type weight coefficient; represents the -th resource, and is the set of data type characteristic identifiers of the -th knowledge point; is the number of knowledge data types; If it is 1, it means that the -th knowledge point in the -th resource is presented in the -th data type; otherwise if it is 0, it means that it is not presented in this type; represents the learner's preference for the data type when accepting the -th knowledge point, represents whether the learner accepts the -th knowledge point presented in the -th data type. If it is 1, it is; otherwise if it is 0, it means that the learner rejects the knowledge point presented in this content type; is an indicator function used to judge whether the elements at the -th and -th positions in the sequence match. If is true, then , otherwise ;​​ (5) It is a fifth-level secondary fitness function, which is used to evaluate the balance of the difference between the preferred material type of the learner and the material type of the recommended resources; It is a variable parameter for the difference in material type; (6) It is a sixth-level secondary fitness function, which is used to evaluate whether the time required to complete the learning resources exceeds the upper limit of the acceptable learning duration of the learner. represents the upper limit of the acceptable learning duration of the learner; (7) is a multi-dimensional matching fitness function, , , , , and are the weight coefficients of the secondary fitness function; The smaller the value, the higher the quality of the recommended learning resources and the more suitable they are for recommendation to learners. Conversely,the higher the value, the less suitable the learning resources are for recommendation to learners.

5. The online learning resource recommendation method based on the improved honey badger population algorithm according to claim 4, characterized in that, Determining the recommended learning resources based on the random mapping model includes: Map the discrete element values in the honey badger individual position vector to the binary space. The th learning resource decision sequence is generated by mapping the corresponding honey badger individual position vector , where , , is the population size, is the th individual's learning resource decision sequence at the th iteration, is the th individual's position vector at the th iteration, is the th element in the position vector; Combined with the random mapping model, the learning resource decision sequence is expressed as follows: (8) In formula (8), is a random number with a value range between [0, 1]; Combined with the decision sequence result in formula (8), calculate the fitness function value of the decision sequence corresponding learning resource according to the matching fitness function formula (7); Sort the fitness function values of each learning resource from smallest to largest. The fitness function value corresponding to the recommended learning resource needs to be the smallest in history, and the individual corresponding to its decision sequence is the historically optimal individual of the honey badger population, denoted as , ; Updating the historical optimal individual recurrence statistics value of the honey badger population includes: If the absolute value difference between the value of the -th dimension in the historical optimal individual position of the honey badger population and the value of the -th dimension in the historical optimal individual position of the honey badger population during the previous iteration is less than the recurrence statistical value parameter , then perform processing on the statistical value , otherwise maintain the initial value; the recurrence statistical value of the -th dimension of the historical optimal individual of the honey badger population is , and has an initial value of 0, the threshold is set to , randomly generate -dimensional vectors between [0, 1] as the initial positions of the honey badger population; The update formula for updating the discovery probability is as follows: (9) is the discovery probability at the th iteration, and is the maximum number of iterations.

6. The online learning resource recommendation method based on the improved honey badger population algorithm according to claim 5, wherein Implementing an elite club guiding strategy based on temporal trajectory similarity to update the population individuals includes: Calculate the temporal trajectory similarity between the individuals of the population and the optimal individual Sort the similarity values from large to small, and select the positions of the top Tp% of the individuals as candidate solutions, where Tp is the proportion of candidate solutions for the temporal trajectory similarity. Randomly select one solution from the candidate solutions as the approximate optimal individual in the current iteration , where , is the value of the -th dimension in the approximate optimal individual position in the current iteration; The similarity calculation formula for the temporal trajectory between an individual and the optimal individual in the population is as follows: (10) is the temporal trajectory similarity between an individual in the population and the optimal individual; Random generation , the value range is [0, 1]. If , the position update formula of the random population individuals is as follows: (11)； If , the position update formula for individuals in the random population is as follows: (12) Among them, , indicating a random output of -1 or 1, , , and are all random numbers with values in the range of [0, 1]; Among them, in formulas (10) to (12), ; if , in formula (11) is replaced by .

7. The online learning resource recommendation method based on the improved honey badger population algorithm according to claim 6, characterized in that, Judging whether to perturb the updated population individuals according to the updated discovery probability includes: Randomly generate detection and comparison parameters , , judge whether it is less than the updated discovery probability : If so, perturb the individual at the th iteration in the population; Implementing a mutation strategy based on a multi-strategy weight adaptive mechanism to perturb the individuals includes: The position update formula of the individual is as follows: (13) Among them, , is the individual position after perturbation; is a random number with a value range between [-1, 1], is a random number with a value range between [0, 1]; The judgment of whether to retain the perturbed individual includes: Successively according to Equation (8) and Equation (7), calculate the fitness function value of the individual after perturbation. If the fitness function value of the perturbed individual is less than the value before perturbation, then retain the position after perturbation; otherwise, keep the original position unchanged. The judgment of whether the position of the individual is updated is based on the following formula: (14) The steps of judging whether to perturb the position of the optimal individual according to the variable-scale reverse learning strategy integrating the nonlinear dynamic system theory are as follows: Judge whether the threshold is reached , if so, perturb the optimal individual position according to the variable-scale reverse learning strategy that integrates the nonlinear dynamic system theory; Perturbing the position of the optimal individual according to the variable-scale reverse learning strategy integrating the nonlinear dynamic system theory includes: (15)； Among them, is the control parameter of the nonlinear dynamic system, , is the optimal individual position after perturbation, , is a random number with a value range between [0, 1]; The judgment of whether to retain the position after perturbation includes: successively according to Equation (8) and Equation (7), calculate the fitness function values of the historical optimal individuals before and after perturbation. If the fitness function value after perturbation is less than the value before perturbation, then retain the position after perturbation; otherwise, keep the original position unchanged. The judgment of whether the position of the historical optimal individual is updated is based on the following formula: (16)。 8. An online personalized learning resource recommendation system based on an improved honey badger population algorithm, characterized in that, Includes: An acquisition module: used to acquire the user portrait and learning resource characteristics of the learner; An output module: used to input the user portrait and learning resource characteristics into a pre-trained learning resource recommendation model. The learning resource recommendation model constructs a multi-dimensional matching fitness function between the learner and the learning resources according to the user portrait and learning resource characteristics of the learner; solve the multi-dimensional matching fitness function through an improved honey badger population algorithm and output the recommended learning resources. Among them, the improved honey badger population algorithm is based on an elite club guidance strategy of temporal trajectory similarity, and randomly selects a sub-optimal individual from the adjacent area of the population optimal individual as the population leader through this strategy to guide the population to search globally.

9. A computer device, characterized in that, Includes a processor and a storage medium; The storage medium is used to store instructions; The processor is used to operate according to the instructions to execute the steps of the method according to any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the program is executed by the processor, it implements the steps of the method according to any one of claims 1 to 7.

Images