Data processing method, model training method, device and equipment based on multiple targets

Abstract

The invention provides a multi-target data processing method and device, a model training method and device and equipment, and relates to the technical field of big data, in particular to the fields of data processing, service recommendation and machine learning in the field of big data. According to the specific implementation scheme, according to basic information of a user and data information of the user for a to-be-planned service, an initial planning path of the user about the to-be-planned service is generated, and the initial planning path comprises a plurality of process targets; monitoring progress data of the user for the plurality of process targets in real time, and adjusting the initial planning path based on the progress data to obtain a planning path; and synchronizing the planned path to a terminal device of the user.

Description

Technical Field

[0001] The present disclosure relates to data processing, service recommendation, and machine learning in the field of big data technology, and in particular to a multi-objective-based data processing method, model training method, device, and equipment. Background Art

[0002] With the continuous development of science and technology, more and more users are gradually abandoning manual planning and adopting scientific planning systems or tools for intelligent planning of their own or their family's education plans, travel plans, etc.

[0003] Taking education planning as an example, current education planning tools usually plan a single short-term goal for new users based on the static data of most users, without considering the user's all-round development based on their own factors. Once the user executes the plan, if the plan is not followed during the actual implementation process (such as due to physical factors or other reasons that affect the progress), the planning tool does not have the function of dynamic adjustment, which leads to the user having to abandon the current plan and re-plan. This not only reduces the operational efficiency of the planning tool, but also affects the user experience. Summary of the invention

[0004] The present invention provides a multi-objective-based data processing method, a model training method, a device and an apparatus.

[0005] According to a first aspect of the present disclosure, a multi-objective data processing method is provided, comprising:

[0006] Generate an initial planning path for the user with respect to the service to be planned according to the basic information of the user and the data information of the user with respect to the service to be planned, wherein the initial planning path includes a plurality of process objectives;

[0007] monitoring the progress data of the user for the multiple process objectives in real time, and adjusting the initial planned path based on the progress data to obtain a planned path;

[0008] The planned path is synchronized to the user's terminal device.

[0009] According to a second aspect of the present disclosure, a model training method is provided, comprising:

[0010] Acquire multi-user sample data, wherein the multi-user sample data includes basic information of multiple user samples, data information about the service to be planned, and an initial planning path;

[0011] Acquire the context data and ultimate goal achievement data of the user sample for the initial planned path, and obtain multi-user context data and ultimate goal achievement data;

[0012] The large language model is trained according to the multi-user sample data, the multi-user situational data and the ultimate goal achievement data to obtain a first machine learning model.

[0013] According to a third aspect of the present disclosure, a model training method is provided, comprising:

[0014] Extracting basic information of user samples and data information about services to be planned from multi-user sample data;

[0015] Obtaining progress data and progress influencing factors of the user sample for the planned path;

[0016] The large language model is trained according to the multi-user sample data, the progress data corresponding to the user samples, and the progress influencing factors to obtain a second machine learning model.

[0017] According to a third aspect of the present disclosure, there is provided a multi-objective-based data processing device, comprising:

[0018] An initial path generation module, which is configured to generate an initial planning path for the user with respect to the service to be planned based on the basic information of the user and the data information of the user with respect to the service to be planned, wherein the initial planning path includes a plurality of process objectives;

[0019] A dynamic adjustment module, which is configured to monitor the progress data of the user for the multiple process objectives in real time, and adjust the initial planned path based on the progress data to obtain a planned path;

[0020] A synchronization module is configured to synchronize the planned path to the user's terminal device.

[0021] According to a fourth aspect of the present disclosure, a model training device is provided, comprising:

[0022] A first data acquisition module, which is configured to acquire multi-user sample data, wherein the multi-user sample data includes basic information of multiple user samples, data information about the service to be planned, and an initial planning path;

[0023] A second data acquisition module is configured to acquire the context data and ultimate goal achievement data of the user sample for the initial planned path, and obtain multi-user context data and ultimate goal achievement data;

[0024] The first training module is configured to train the large language model according to the multi-user sample data, the multi-user situational data and the ultimate goal achievement data to obtain a first machine learning model.

[0025] According to a fourth aspect of the present disclosure, a model training device is provided, comprising:

[0026] A data extraction module, which is configured to extract basic information of user samples and data information about the service to be planned from the multi-user sample data;

[0027] A third data acquisition module, configured to acquire the progress data and progress influencing factors of the user sample for the planned path;

[0028] The second training module is configured to train the large language model according to the multi-user sample data, the progress data corresponding to the user samples, and the progress influencing factors to obtain a second machine learning model.

[0029] According to a fifth aspect of the present disclosure, a server is provided, including:

[0030] at least one processor; and

[0031] a memory communicatively connected to the at least one processor; wherein,

[0032] The memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can execute the multi-objective data processing method provided in the first aspect, or the model training method provided in the second aspect, or the model training method provided in the third aspect.

[0033] According to a sixth aspect of the present disclosure, a non-transitory computer-readable storage medium storing computer instructions is provided, wherein the computer instructions are used to enable the computer to execute the multi-objective-based data processing method provided in the first aspect, or the model training method provided in the second aspect, or the model training method provided in the third aspect.

[0034] According to the seventh aspect of the present disclosure, a computer program product is provided, comprising: a computer program, the computer program being stored in a readable storage medium, at least one processor of a server being able to read the computer program from the readable storage medium, the processor executing the computer program so that when the server executes the computer program, the multi-objective-based data processing method provided in the first aspect above, or the model training method provided in the second aspect above, or the model training method provided in the third aspect above is implemented.

[0035] According to the technology disclosed in the present invention, a more personalized service planning solution that supports dynamic adjustment is provided to users.

[0036] It should be understood that the content described in this section is not intended to identify the key or important features of the embodiments of the present disclosure, nor is it intended to limit the scope of the present disclosure. Other features of the present disclosure will become easily understood through the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] The accompanying drawings are used to better understand the present solution and do not constitute a limitation of the present disclosure.

[0038] Figure 1 It is a flowchart of a multi-objective data processing method provided according to an embodiment of the present disclosure;

[0039] Figure 2 yes Figure 1 One of the flowchart diagrams of step S101;

[0040] Figure 3 yes Figure 1 2. Schematic diagram of the process of step S101;

[0041] Figure 4 yes Figure 1 3. Flow chart of step S101 in FIG.

[0042] Figure 5 It is one of the flowcharts of another multi-objective-based data processing method provided according to an embodiment of the present disclosure;

[0043] Figure 6a This is a second flow chart of another multi-objective data processing method provided by an embodiment of the present disclosure;

[0044] Figure 6b yes Figure 6a Schematic diagram of the process of model optimization;

[0045] Figure 7 is a schematic diagram of a processing flow of multimodal data in an embodiment of the present disclosure;

[0046] Figure 8 is a flowchart of another multi-objective-based data processing method provided in an embodiment of the present disclosure;

[0047] Fig. 9 is a schematic diagram of a process for generating a reverse path in an embodiment of the present disclosure;

[0048] Fig.10 yes Figure 8 Schematic diagram of the process of step S102;

[0049] Fig.11 It is a flowchart of a model training method provided according to an embodiment of the present disclosure;

[0050] Fig.12 is a flowchart of another model training method provided according to an embodiment of the present disclosure;

[0051] Fig.13ais a structural schematic diagram of a multi-objective-based data processing device provided according to an embodiment of the present disclosure;

[0052] Fig.13b is a schematic diagram of interaction between a multi-objective-based data processing device and a user according to an embodiment of the present disclosure;

[0053] Fig.14 is a structural schematic diagram of a model training device provided according to an embodiment of the present disclosure;

[0054] Fig.15 is a structural schematic diagram of another model training device provided according to an embodiment of the present disclosure;

[0055] Fig.16 It is a structural diagram of a server for implementing the multi-objective-based data processing method provided in an embodiment of the present disclosure;

[0056] Fig.17 It is a structural diagram of a server for implementing the model training method provided in an embodiment of the present disclosure;

[0057] Fig.18 is a structural diagram of a server for implementing another model training method provided in an embodiment of the present disclosure;

[0058] Fig.19 The block diagram is a block diagram of an electronic device used to implement the method of the embodiment of the present disclosure. DETAILED DESCRIPTION

[0059] The following is a description of exemplary embodiments of the present disclosure in conjunction with the accompanying drawings, including various details of the embodiments of the present disclosure to facilitate understanding, which should be considered as merely exemplary. Therefore, it should be recognized by those of ordinary skill in the art that various changes and modifications may be made to the embodiments described herein without departing from the scope and spirit of the present disclosure. Similarly, for the sake of clarity and conciseness, descriptions of well-known functions and structures are omitted in the following description.

[0060] The multi-objective data processing solution provided by the embodiments of the present disclosure can be applied to various planning service scenarios, such as education planning services, travel planning services, smart home planning services, etc., which need to be applied to all aspects of goal planning. To facilitate the understanding of the embodiments of the present disclosure, the multi-objective data processing solution is introduced below using education planning as an example:

[0061] The educational planning methods used in related technologies mainly include the following methods: 1) Traditional textbooks and supplementary teaching books, such as parents making study plans for their children based on popular textbooks and supplementary teaching books on the market; 2) Some parents seek professional educational consulting services and obtain study planning suggestions by consulting experts; 3) Some online learning platforms provide learning resources and courses, but usually lack customized plans for individual children and need to rely on parents to choose courses independently. The above solutions mainly rely on human planning, and there are many kinds of textbooks and learning resources on the market. Parents face information overload when choosing and find it difficult to make the best choice; learning plans also often rely on general suggestions, lack of specific analysis of specific individual learning progress, interests and development goals, and cannot fully consider the long-term development needs and personal characteristics of children, resulting in unsatisfactory learning results and growth plans; and some educational consulting services usually require high fees, which is too costly for ordinary families.

[0062] Obviously, with the increasing demand for personalized education and comprehensive development planning in modern society, the above-mentioned manual planning method is difficult to adapt to the dynamic changes and complex growth paths of different individuals during their growth process. Most of the planning systems on the market are based on static, single data sources, which cannot fully respond to real-time changes and are difficult to provide users (especially parents and children) with dynamic adjustments and multi-faceted growth path suggestions. In addition, when determining long-term goals such as academic, competition and career planning, current technology also lacks effective means to cross the time span and achieve the ultimate goal back to the current situation.

[0063] Therefore, the present disclosure provides a more personalized service planning solution that supports dynamic adjustment, namely, a multi-objective data processing method, a model training method, a device and equipment, which are applied to data processing, service recommendation and machine learning in the field of big data. The solution generates an initial planning path for the user regarding the service to be planned based on the user's basic information and the user's data information for the service to be planned. The initial planning path includes multiple process goals, monitors the user's progress data for the multiple process goals in real time, adjusts the initial planning path based on the progress data, obtains the planning path, and synchronizes the planning path to the user's terminal device to achieve a customized multi-objective planning solution for the user. The multi-objective planning solution can be dynamically adjusted according to the user's progress data without the need for re-planning, which effectively improves the operational efficiency of the planning tool and brings a better experience to the user.

[0064] It should be noted that the machine learning model in this embodiment is not a model for a specific user and cannot reflect the personal information of a specific user. It should be noted that the basic information, data information or progress data of the user in this embodiment are from a public data set.

[0065] In the technical solution of the present disclosure, the collection, storage, use, processing, transmission, provision and disclosure of user personal information involved are in compliance with the provisions of relevant laws and regulations and do not violate public order and good morals.

[0066] In order to enable readers to more deeply understand the implementation principle of the present disclosure, the following Figure 1-Figure 19 The embodiments of the present disclosure are further refined.

[0067] Figure 1 is a flowchart of a multi-objective data processing method provided by an embodiment of the present disclosure, and the execution subject may be a planning server, such as Figure 1 As shown, the method includes steps S101-S103.

[0068] In this embodiment, the planning server can be a server that generates a planning path corresponding to a specific planning service for the user, or a server that can generate planning paths corresponding to multiple planning services, which generates a corresponding planning path for the user according to the corresponding planning service selected by the user through interaction with the user. In addition, the planning server can be a cloud server or other server (such as an independent server or a distributed server, etc.), which is not particularly limited in this embodiment.

[0069] Combine the following Figure 1 As shown:

[0070] Step S101: Generate an initial planning path of the user for the service to be planned based on basic information of the user and data information of the user for the service to be planned, wherein the initial planning path includes a plurality of process objectives.

[0071] In this embodiment, the basic information of the user may be the inherent information of the user, such as the user's name, gender, etc., and the data information of the user for the service to be planned may be the user information related to the service to be planned. For example, for the education planning service, it is necessary to obtain the user's academic performance (such as academic performance in various subjects, etc.), extracurricular activities, hobbies, etc. information in order to make a more comprehensive planning for the user. Optionally, in order to improve the generation accuracy of the planning path, these data may be data obtained by integrating and analyzing multimodal data, such as text data, image data, and video data.

[0072] Exemplarily, a multi-objective optimization (MOO) algorithm can be used to generate an initial planning path, which is an algorithm for optimizing multiple conflicting objective functions. Specifically, the user's basic information and data information are used to identify and define multiple process goals. For example, the data information can carry the ultimate goal set by the user in advance (in some application scenarios, if the user does not set the ultimate goal, the planning server can also identify the ultimate goal based on the basic information and data information, such as identifying the ultimate goal based on prior data or modeling). According to the ultimate goal, the basic information and data information of the user are decomposed into multiple process goals, and the minimization time, maximization satisfaction, minimization cost, etc. can be defined for multiple process goals, or time constraints, budget constraints, resource availability, etc. can be constrained to generate an initial planning path. In some examples, the initial path can also be generated by decomposing multiple process goals into sub-goals and assigning weights. This process has been described in detail later and will not be repeated here.

[0073] Instead of planning a single short-term goal for a new user based on the static data of most users in related alternative technologies (for example, when planning a path for a primary school student to enter junior high school, planning the path for the primary school student is carried out based on historical user data), without considering the multi-faceted development of the user based on his or her own factors, this embodiment combines the user's basic information and data information to generate a planning path that includes multiple process goals (such as improving academic performance, adding points for extracurricular activities, etc.), thereby achieving the purpose of improving the personalization and comprehensiveness of goal planning.

[0074] In an optional implementation, the planning server may be a server that provides a variety of planning services to users, and obtains user input information through interaction with the user to facilitate subsequent planning path generation. Specifically, the method provided in this embodiment may also include the following steps:

[0075] In response to a request for a service to be planned initiated by the user, obtaining basic information of the user and data information for the service to be planned;

[0076] The service to be planned includes at least one of the following: education planning service, tourism planning service, and smart home planning service.

[0077] In response, it is used to indicate the conditions or states on which the executed operations depend. When the dependent conditions or states are met, one or more operations executed can be in real time or have a set delay. Unless otherwise specified, there is no restriction on the order of execution of the multiple operations executed.

[0078] For example, a user can select one or more of the multiple services to be planned provided by the planning server on his terminal device (such as a mobile phone or smart wearable device, etc.), and initiate a request to the planning server. The planning server responds to the request and obtains basic user information and data information for the services to be planned.

[0079] It should be noted that, in addition to the education planning service, travel planning service and smart home planning service shown above, the services to be planned may also include other planning services, such as financial planning services, etc. This embodiment does not specifically limit the specific planning services.

[0080] Through the above technical solution, user-related data information can be obtained according to the user's planning needs, so as to customize a planning path that meets the user's planning needs, providing convenience for the user.

[0081] Step S102: monitor the progress data of the user for the multiple process objectives in real time, and adjust the initial planned path based on the progress data to obtain a planned path.

[0082] After the server generates the initial planning path, the user can upload his progress data to the server in real time through the terminal device. Based on the interaction between the server and the terminal device, the user's progress data is monitored in real time, and the initial planning path is dynamically adjusted using the progress data, so that a dynamic planning path that adapts to the user can be provided to the user throughout the life cycle of the user's ultimate goal. Exemplarily, the monitoring and synchronization process is a cyclic process, that is, the planning path is adjusted and synchronized according to the user's real-time progress data, and the next cycle is entered to monitor the user's new progress data and adjust it until it is monitored that the user has achieved the ultimate goal of the planning path.

[0083] Exemplarily, the server can evaluate whether the user has achieved various process goals according to the initial planning path (or the planning path adjusted in the previous cycle) based on the monitored progress data. The evaluation criteria can include the completion of each sub-goal, such as the preset time node, the quality standard of task completion, etc. Next, the deviation between the user's actual progress and the initial planning path is identified. The deviation may be caused by the user's progress being too fast or too slow or encountering obstacles. Based on the identified deviation, the initial planning path is dynamically adjusted (for example, adding or deleting process goals or adjusting process goal weights, etc.). In addition, resources can be reallocated, task priorities can be modified, and time arrangements can be adjusted for process goals.

[0084] Compared with the static planning scheme in the related art, once the user executes the planning scheme, if the user fails to execute it according to the planning scheme during the actual execution process (such as due to physical factors or other reasons affecting the progress, etc.), the planning tool does not have the function of dynamic adjustment, which causes the user to abandon the current planning scheme and re-plan, which not only reduces the operational efficiency of the planning tool, but also affects the user experience. This embodiment can realize real-time dynamic adjustment of the planning path by real-time monitoring of the user's progress data for the process goal and adjusting it, so as to improve the dynamic adaptability and real-time responsiveness of the planning path, and provide users with a target planning scheme that is more suitable for their own conditions.

[0085] Step S103: synchronizing the planned path to the user's terminal device.

[0086] Through the above technical solution, a personalized multi-objective planning scheme can be customized for users, which is more suitable for the user's all-round development. The multi-objective planning scheme can be dynamically adjusted according to the user's progress data without the need for re-planning, which effectively improves the operational efficiency of the planning tool and brings a better experience to users.

[0087] Next, the embodiment of the present disclosure further introduces the generation of the initial planning path of the user according to the basic information of the user and the data information of the user for the service to be planned in the above step S101, such as Figure 2 As shown, the following steps S1011-S1013 are included:

[0088] S1011. Based on the basic information of the user and the data information of the user for the service to be planned, the ultimate goal is positively decomposed into multiple process goals.

[0089] Exemplarily, the data information carries a preset ultimate goal of the user, which may be the ultimate goal set during the user input data stage. In some examples, the ultimate goal may also be identified by the server based on the user's basic information and data information and recommended to the user.

[0090] Among them, the ultimate goal is positively decomposed into multiple process goals according to the user's basic information and the user's data information on the service to be planned. This can be achieved by identifying key areas. For example, according to user data, the ultimate goal is decomposed into several key areas or main components. For example, for learning planning, the ultimate goal is to be admitted to the ideal XXX university. By analyzing the total college entrance examination score, score requirements for each subject, and other possible admission conditions (such as interviews, specialties, etc.) required to be admitted to XXX University, and adapting to the user's current academic performance, academic status, interests and hobbies, process goals for improving grades in each subject, process goals for strengthening extracurricular activities, and process goals for strengthening specialties are set (if the user has no interests and hobbies that can add points based on specialties, no setting is required), etc., and the ultimate goal is positively decomposed in combination with the user's actual situation.

[0091] S1012, acquiring multiple sub-goals for achieving the process goal, and assigning weights to the sub-goals, to obtain multiple sub-goals of the process goal and their weights.

[0092] In this embodiment, for each process goal, multiple sub-goals for achieving the process goal may be obtained, and a weight may be assigned to each sub-goal to obtain multiple sub-goals and their weights for each process goal.

[0093] The process goal in the above step S1011 is mainly to adapt to the macro-goal of the user's all-round development, such as improving grades in various subjects, strengthening extracurricular activities, etc. In order to facilitate execution in actual user applications, this embodiment decomposes the process goal into sub-goals executed by the user and assigns corresponding weights, so that the user can achieve the process goal step by step according to the sub-goals, thereby achieving the ultimate goal.

[0094] Exemplarily, an importance assessment can be performed for each sub-goal, and the assessment method can be determined based on a large amount of prior data. For example, in a study plan, the score improvement in Chinese, mathematics, and English accounts for a higher proportion, while the score improvement in other subjects accounts for a relatively lower proportion. Combined with the user's data information, such as excellent Chinese scores (such as above 90 points, which can be determined based on actual application adaptability), the weight of improving mathematics and English scores can be adaptively increased (for example, prior data modeling can be used to assign weights to each sub-goal), thereby assigning a more reasonable weight to each sub-goal, so that users can execute each sub-goal and achieve the corresponding process goals.

[0095] S1013. Generate the initial planning path according to the multiple process goals, the multiple sub-goals of the process goals and their weights.

[0096] Through the above-mentioned multi-objective planning scheme, the ultimate goal is decomposed into multiple process goals, and the sub-goals used to achieve each process goal are obtained and weights are assigned to the sub-goals, so that the corresponding initial planning path can be obtained. The initial planning path can include the multiple process goals, the sub-goals of each process goal and their weights, which helps to balance multiple goals and improve the rationality of the planning path, so that users can execute according to the planned path.

[0097] Considering that the rationality of weight allocation will affect the quality of the planned path, this embodiment provides a weight optimization algorithm. Optionally, the above step S101 generates the initial planned path of the user according to the basic information of the user and the data information of the user for the service to be planned, and may also include the following steps:

[0098] Obtaining a combined weight of the process goal according to a combined calculation result of the weights of the sub-goals of the process goal;

[0099] If the difference between the combined weight of the process objectives and the preset standard weight of the process objectives reaches a preset deviation value, the weights of the sub-objectives are optimized based on a preset optimization algorithm so that the combined weight of the process objectives is within the preset deviation value.

[0100] In this embodiment, for each process goal, the combined weight of the process goal can be obtained based on the combined calculation result of the weight of each sub-goal of the process goal to obtain the combined weight of each process goal. When the difference between the combined weight of one or more process goals and the preset standard weight reaches a preset deviation value, the weight of each sub-goal is optimized based on a preset optimization algorithm, so that the combined weight of each process goal is within the preset deviation value.

[0101] In this embodiment, the combined calculation result of the weights of each sub-goal in the process goal can be the sum of the corresponding value of each sub-goal (which can be the expected value of the sub-goal, and those skilled in the art can make adaptive determinations based on actual applications) and the weight product and the corresponding values ​​of other sub-goals and the weight product, and divided by the corresponding values ​​of multiple process goals to obtain the combined weight of the process goal (in some embodiments, the combined weight can also be calculated directly by weight summation, or other methods are used to calculate the combined weight, which is not particularly limited in this disclosure). For each process goal, the preset standard weight corresponding to the process goal can be compared respectively. When it exceeds the corresponding deviation value, it is adjusted and optimized. A cyclic optimization method can be used until the combined weight of each process goal is within the deviation range, thereby improving the rationality of the weight distribution of each sub-goal.

[0102] It should be noted that the preset standard weight of the process goal in this embodiment can be the weight predetermined when the server decomposes multiple process goals (based on the user's constraints, such as family conditions do not support investing too much money in user interests and hobbies to strengthen features, which can be quantified by statistical methods or mean calculation methods, which will not be repeated here), or it can be the weight set by the user end. This embodiment does not specifically limit the process of determining the preset standard weight. Accordingly, those skilled in the art can adaptively determine the preset deviation value in combination with actual applications.

[0103] The optimization algorithm may include at least one of the following: a Pareto optimization algorithm, a multi-objective genetic algorithm, a dynamic programming algorithm, and a particle swarm optimization algorithm.

[0104] Pareto Optimization Algorithm is a multi-objective optimization technique. It finds a set of solutions called Pareto Front, which is called the optimal solution. Therefore, Pareto's optimization algorithm can be used to find the optimal classifier or model. The Pareto optimization algorithm can be used to solve multi-objectives with weights, thereby optimizing the weights of each sub-objective. For example, assuming that there are n sub-objectives (f_1, f_2, ..., f_n), each sub-objective can be assigned a corresponding weight (w_1, w_2, ..., w_n), and an initial population is generated. The individuals in the initial population represent a possible solution, which contains the value of the decision variable and the corresponding weight. By calculating the objective function value of the individual and evaluating its fitness based on these values, the fitness can be calculated by weighted sum. Next, the population is non-dominated sorted, and the population is divided into different levels. The lower the level, the more superior the individual is, and the better individuals are selected for reproduction (such as using tournament selection, roulette selection, etc.). The selected individuals are subjected to crossover and mutation operations to generate new individuals. At the same time, the newly generated individuals are added to the population and non-dominated sorting is performed to retain the best individuals, thereby outputting the Pareto optimal solution set and its corresponding weights, thereby achieving the optimization of the weights of each sub-goal.

[0105] Correspondingly, Multi-Objective Genetic Algorithm (MOGA) is an optimization algorithm based on evolution, which is suitable for complex multi-objective optimization problems. Through genetic operations (such as selection, crossover and mutation), the solution space can be effectively explored and the weights of sub-objectives can be optimized. The optimization process is guided by setting specific constraints (i.e. preset deviation values); Dynamic Programming (DP) is a method of recursively decomposing problems, which is suitable for problems with overlapping sub-problems and optimal substructure properties, and can also optimize the weights of sub-objectives; Particle Swarm Optimization (PSO) is an optimization algorithm based on swarm intelligence, which simulates the foraging behavior of bird flocks and can optimize the weights of sub-objectives by adjusting the position and speed of particles. For example, constraints (preset deviation values) can be introduced to control the optimization process.

[0106] To facilitate the understanding of the above technical solution, Figure 3 and Figure 4 As shown, Figure 3 The generation and optimization process of the initial planning path is shown. By considering multiple objectives at the same time and balancing the weights of these objectives through the optimization solution process, a comprehensive optimized path or solution is obtained. In the entire multi-objective planning process, steps such as goal decomposition, weight allocation, optimization calculation and result verification are involved. Specifically, Figure 3 As shown, including:

[0107] a. User input.

[0108] b. Identify and define multiple process goals. You can start from the needs and constraints input by the user (which can be determined according to the specific type of service to be planned. For example, in education planning, study time, family conditions, etc. can be used as constraints). Forward decomposition into various process goals that need to be achieved. For example, process goals may include improving academic performance, exploring areas of interest, and school enrollment planning.

[0109] c. Decompose into actionable sub-goals. Multiple process goals can be decomposed into multiple sub-goals, which constitute more specific and quantifiable operations. For example, subject goals can be broken down into academic achievement goals for each subject and each semester.

[0110] d. Assigning weights may include assigning weights to each sub-goal and process goal. The importance of each goal relative to the overall plan may vary. For example, for some families, academic studies may be the most important goal, while social activities and extracurricular practice are relatively secondary. In this embodiment, weights may be assigned to different sub-goals based on empirical values, so that the algorithm can better balance multiple goals.

[0111] e. Optimization solution: optimization algorithms can be used to solve weighted multi-objective problems. Optimization algorithms can include: dynamic programming, multi-objective genetic algorithm (MOGA), particle swarm optimization (PSO), Pareto optimization, etc.

[0112] During the optimization process, a large language model (LLM) can be used to help predict possible strategies during the iteration and solution process, thereby improving the calculation of the optimization path.

[0113] f. Result verification and adjustment: the generated results (i.e. the optimized initial planning path) can be checked one by one to see whether they have achieved the trade-off state of multiple objectives within the constraints of user needs (whether the deviation between the set weights, i.e. the preset standard weights, meets the conditions). If it is found that the optimization results of each objective do not match the set weights (for example, there is a significant deviation in the achievement of a certain objective), the algorithm or the weights of each objective will be adjusted to enter the next step of re-optimization and solution.

[0114] g. Adjust weights or parameters. Since the actual planning environment is usually complex, we can adjust the optimization parameters or target weights again based on the feedback from the verification to gradually approach the optimal solution. This process can be carried out through user input and system self-adjustment.

[0115] g. Output optimized path plan. After multiple rounds of iterations, the optimal path plan that satisfies the target balance will be output. In addition, the server can also provide corresponding strategic adjustments and suggestions based on the planning plan.

[0116] For the above multi-objective optimization, that is, the weight optimization process, we can combine Figure 4 Example to understand:

[0117] A. User input: User input can carry the ultimate goal to be achieved, such as "getting into an ideal university", "improving social skills", etc. This example takes getting into an ideal university as an example.

[0118] B. Break down the ultimate goal into multiple process goals. For example, "improve academic performance" is one of the process goals.

[0119] C. Break down each process goal into smaller, quantifiable or actionable sub-goals, such as "improve math scores", "improve Chinese scores", etc.

[0120] D. Assign weights to each sub-goal. For each sub-goal in the process goal, the corresponding weight can be assigned through system judgment or large model. This process can be determined by analyzing the importance of the goal. For example, for "improving math achievement", which is more important to the process goal, a higher weight (such as 50%) can be assigned for the period.

[0121] E. Combine and optimize the weights of sub-goals within each goal. Specifically, the weight configuration of the sub-goal is calculated by combining the weights of each sub-goal. Path optimization can be performed through optimization algorithms such as the Pareto optimization algorithm to achieve the goal of improving academic performance, and the weights of various subjects are aggregated and adjusted to meet actual needs.

[0122] F. Output the optimized path for each process. The server can plan the corresponding path for each sub-goal, such as specific time allocation, resource allocation, and specific learning path, which can be determined in combination with actual applications or prior data.

[0123] G. Comprehensively optimize the weight distribution and adjustment among multiple process goals. The weights of multiple goals (such as academic performance, extracurricular activities, and social skills) can be summarized globally to determine the path method for resource allocation and optimization, and provide final adjustment suggestions.

[0124] Through the above goal decomposition and weight allocation steps, the path of each sub-goal has been carefully planned, and the generated planning path is more adaptable to personalized comprehensive growth goals.

[0125] Considering that the planning process usually requires a large number of parameter optimization solutions, and machine learning models such as large models (such as the natural language processing model Generative Pre-Trained Transformer, GPT) have strong semantic understanding and data prediction capabilities, they can optimize the model process by learning a large amount of similar scenario data and outputting highly customized solutions, thereby greatly improving the accuracy and feasibility of the planned path. This embodiment uses a machine learning model to assist the planning algorithm in predicting possible future states, thereby optimizing the planned path. Figure 5 This is another multi-objective data processing method provided by the embodiment of the present disclosure. On the basis of the above embodiment, this embodiment introduces a first machine learning model and optimizes the initial path by using the reasoning and prediction capabilities of the model. Specifically, in addition to the above steps S101-S103, after generating the initial planning path in step S101, this embodiment may also include the following steps S501 and S502:

[0126] Step S501: input the initial planning path, the basic information and the data information into a pre-trained first machine learning model, and infer the initial planning path based on the first machine learning model to obtain the prediction result and optimization recommendation information of the initial planning path for predicting the ultimate goal.

[0127] In this embodiment, the first machine learning model can be trained based on multi-user sample data and used to predict and optimize the planned path. Specifically, the basic information of multiple user samples, data information about the service to be planned, and the initial planned path can be obtained, and the user samples' situational data for the initial planned path and the ultimate goal achievement data can be collected to obtain multi-user situational data and ultimate goal achievement data. The large language model can be trained based on the multi-user sample data, the multi-user situational data, and the ultimate goal achievement data to train the first machine learning model. The more detailed training process can be found in the following embodiments, and will not be elaborated here.

[0128] Step S502: Optimize the initial planned path according to the prediction result and the optimization recommendation information, so that step S102 can adjust the optimized initial planned path based on the progress data.

[0129] Exemplarily, after being processed by the first machine learning model, the user's prediction result for the initial path can be obtained. At the same time, the first machine learning model outputs corresponding optimization recommendation information based on the prediction result. The server can use the optimization recommendation information and combine it with the adjustment information input by the user for the prediction result (for example, the prediction result is that the probability of improving math scores is 30% and the probability of improving Chinese scores is 80%. The user can input the goal of deleting the math score improvement goal and optimizing other goals such as improving Chinese scores, etc. This is just a possible example) to optimize the initial planned path.

[0130] By optimizing the initial planning path in combination with the large model, the generated initial planning path can be made more accurate and conducive to achieving the user's goals, thereby reducing the subsequent dynamic tuning process and bringing a better experience to users.

[0131] Next, the prediction result and optimization recommendation information of the initial planning path about the ultimate goal are obtained by reasoning the initial planning path based on the first machine learning model in the above step S501, wherein the first machine learning model may include a situation simulation module, a future scenario deduction module and an optimization suggestion module, and the above step S501 may include the following steps:

[0132] By means of the scenario simulation module, according to the basic information and the data information, the simulation scenario of the user for the planned path under each process target is simulated to obtain the scenario data of the user;

[0133] Through the future scenario deduction module, the future scenarios of the user for the planned path at each process goal are deduced and predicted according to the situation data to obtain a prediction result about the ultimate goal;

[0134] When the prediction result is that the ultimate goal is not achieved, optimization recommendation information is generated through the optimization suggestion module, and the optimization recommendation information includes at least one of the following: process goal replacement information, process goal weight adjustment information and personalized suggestions for the user.

[0135] In one example, for the deduction process of the situational simulation module, basic information of users, such as age, gender, etc., and data related to the planned services, such as academic performance, hobbies, etc., can be collected in the data preparation stage. The basic information and data information of users can be cleaned and standardized before being input into the model. The model can create a simulation environment, define the behavior rules of users in different situations, and use rule-based models or reinforcement learning models to simulate user behavior. The initial planning path of the user can be run in the simulation environment, and the user's behavior and reaction under each process goal can be simulated. The situational data of the user under each process goal can be output. The situational data can include the user's behavior, feedback, completion, etc.

[0136] In another example, for the future scenario deduction module, the user scenario data can be obtained from the scenario simulation module, and the scenario data can be cleaned and standardized, and the module can use historical scenario data and ultimate goal achievement data to train the prediction model. This module can use a time series prediction model (such as Long Short-Term Memory, LSTM), a regression model, or a graph neural network-based model to perform future scenario deduction by inputting the current scenario data into the corresponding model, predicting the user's performance under each process goal, and thus outputting the prediction result about the ultimate goal to determine whether the user can achieve the ultimate goal.

[0137] In another example, for the optimization suggestion module, the prediction results are obtained from the future scenario deduction module, and the prediction results are analyzed to determine whether the user can achieve the ultimate goal. For example, if some process goals have a large impact but poor results, it is recommended to replace them with other goals, or adjust the importance weights of different process goals according to the specific situation of the user, and generate personalized optimization suggestions based on the user's basic information and situational data. In addition, the optimization suggestion information can also be output to the user, and the initial planning path can be optimized through interaction with the user.

[0138] Through the above steps, by utilizing the collaborative work between the scenario simulation module, future scenario deduction module and optimization suggestion module in the model, the initial planning path can be efficiently adjusted so that the optimized initial planning path can help users better achieve their ultimate goals.

[0139] Understandably, the big model can train virtual scenes that are highly similar to the actual learning environment, and can also accurately lock in the task / goal setting step by step by asking and deducing the model (similar to the dialogue mode). In the pursuit of goals, this causal chain analysis can be used to simulate possible future states-determine the causal relationship of the path, which is critical to the adjustment of the goal. In addition, under the dynamic decision-making requirements of multiple conditions, the big model can perform a large amount of causal reasoning on the data, select multiple key factors that affect learning / goals and combine them into the actual decision path. Through collective discussion of "several small failure possibilities", multiple potential system failure conditions can be avoided in advance. In addition, the big model can also combine the user to provide new input (or infer new options within the algorithm), and the simulation and prediction related modules will dynamically adjust the path according to the current situation. At the same time, since the simulation contains a large number of deduction and iteration processes, it can evolve alternately for multiple situations and continuously improve the planning output. With the help of this feedback process, suboptimal or unfeasible solutions can be gradually eliminated, while increasing the meticulousness and practical significance of the path adjustment for users to successfully achieve a certain ultimate goal.

[0140] The generation and optimization process of the initial planning path is described below in conjunction with an exemplary embodiment. Figure 6a and Figure 6b As shown, Figure 6a The figure shows the initial planning path after the initial planning path is combined with the large model and outputted optimized, including the following steps:

[0141] S61, user inputs data;

[0142] S62. Goal identification and decomposition: define the ultimate goal based on the input and decompose it into multiple process goals.

[0143] S63. Break down multiple goals into specific sub-goals.

[0144] S64, multi-objective programming algorithm, use the multi-objective programming algorithm to generate preliminary learning and growth paths.

[0145] S65. Large model path evaluation: Use the large model to predict, optimize, and supplement the generated path (such as adding personalized features) to generate a more intelligent and personalized solution.

[0146] S66. Result verification and optimization: Multiple interactive calibration system designs are performed to form an optimization solution.

[0147] S67. Output the (initial) planning path, which may be accompanied by dynamic feedback and continuous optimization.

[0148] For the above step S64, Figure 6b The figure shows the basic steps and logical sequence of a possible interactive planning deduction and simulation, and the process from input to simulation and then to output path driven by a large model. Figure 6b As shown:

[0149] S641. At the user interaction layer, input basic data (i.e., basic data and data information, etc.), such as the student's current academic status, individual interests, long-term goals and other key information.

[0150] S642, after receiving the input data, the scenario simulation initialization large model (i.e., scenario simulation module) can simulate the learning scenarios that students may experience through its own knowledge graph and specific domain data, and simulate and generate data for various scenarios, such as predicting grades, external events, and other multivariate variables.

[0151] S643, the learning scenario simulation and generation model (i.e., the future scenario deduction module) simulates multiple possible future scenarios in the built-in scenarios (similar to how a certain learning goal will perform in the future), and preliminarily simulates possible target distribution states, progress, etc. This is an important step in model prediction and logical analysis, which can help predict the possible results of a series of future events in different scenarios.

[0152] S644. Future scenario deduction and selection. In the process, the big model is used to generate possible future scenarios, and multiple deductions are conducted in a virtual environment, and the deduction results are fed back to the user. The deduction includes operations related to the student's future academic or growth development - for example, what impact a key choice has on the learning plan.

[0153] S645. Feedback loop and interactive correction: Based on the feedback from the scenario and its sub-scenario, the user can choose to further evaluate the plan and interactively adjust it, or give the next action suggestion based on the user's new strategy change. This part of the feedback loop can be performed multiple times until the optimization reaches the final satisfactory state.

[0154] S646. Path adjustment and optimization, such as integrating information from multiple feedback loops to determine a reasonable optimal solution and make corresponding adjustments to the initial path plan.

[0155] S647. Output paths and suggestions, provide individualized growth suggestions based on the simulation results generated by interaction and simulation, and output a complete adjustment planning path.

[0156] In this example, the multi-objective planning algorithm provides a basic path construction, while the big model further improves the decision-making ability of the algorithm, and uses its learning, optimization, and simulation capabilities to make the generated path more flexible, more scientific, and ultimately more successful. Therefore, the introduction of the big model provides a new solution for complex, multi-dimensional, and multi-objective path planning, making the path more reasonable, accurate, and personalized.

[0157] In an optional implementation, considering that the diversity of data sources helps to improve the rationality and feasibility of the generated planning path, the data source in this embodiment uses multimodal data instead of a single data source. Specifically, at least one of the above basic information, data information and multi-user sample data is multimodal data, and the multimodal data includes at least two combinations of the following modal data: text, audio, image, video data, biometric data and sensor data.

[0158] Multimodal data refers to data composed of two or more different modalities. Modality can be understood as data of different forms or information sources, such as text, images, audio, video, etc. Multimodal data combination refers to the integration of data of these different modalities to enhance the expressiveness and application effect of the data. Multimodal data can come from a variety of different modalities, including but not limited to: text: information in written language or electronic documents; images: visual information, such as photos, video frames, etc.; audio: sound information, such as voice, ambient sound, etc.; video: dynamic visual and sound information. Among them, biometric data can be health data, real-time physiological data obtained through health equipment, sensor data can be user action trajectories, user movement information obtained through positioning technology, and so on. In some embodiments, in addition to the modal data of the above examples, other modal data may also be included, and the present disclosure does not specifically limit the specific form of multimodal data.

[0159] Based on multimodal data, data from various models such as text, audio, image and video data can be integrated. Combined with the processing method of multimodal data, the model can better understand and optimize the planning path. The processing process of multimodal data is as follows: Figure 7As shown. Taking education planning as an example, by analyzing the education guidance materials, learning records, family environment influence (audio information such as parents' tutoring participation) contained in the text data, the large model can form a personalized knowledge base, thereby inferring possible future paths for the user when the user has a need. For example, pronunciation recognition can accurately determine the child's interest trends, and this trend information can in turn adjust the goal (planning) and recommend the order and type of reasonable planning paths.

[0160] By comprehensively analyzing information in different modes such as text, images, and audio (such as using multimodal machine learning for analysis), the planned path can be optimized and refined. This module uses large model technology to cross-analyze multiple types of data and provide more comprehensive analysis results to improve the path optimization effect.

[0161] Figure 8 It is another multi-objective data processing method provided by the embodiment of the present disclosure. On the basis of the above embodiment, this embodiment takes into account the forward planning process, lacks ultimate goal orientation, and may cause the generated planning path (especially for long-term goals) to be not reasonable and accurate enough. This embodiment provides reverse path orientation for the planning path by means of target reverse calculation to provide a more reasonable initial planning path. Specifically, in addition to the above steps S101-S103, the following steps S801-S803 may also be included:

[0162] Step S801: According to historical path planning information, the ultimate goal of the user is reversely decomposed into multiple intermediate goals.

[0163] For example, historical route planning data of user samples can be collected, which may include user data and planned route data, etc. These planned routes may be routes with ultimate goals and intermediate goals marked, or data mining and machine learning techniques may be used to extract the ultimate goals and intermediate goals of users from the planned routes. Using these historical planned route information, the ultimate goal of the user can be reversely decomposed into multiple corresponding intermediate goals and sub-goals, etc.

[0164] Step S802: Generate a reverse path for the user regarding the service to be planned based on the multiple intermediate goals.

[0165] For example, for each intermediate goal, the resources and conditions (such as time or funds, etc.) required to achieve each intermediate goal can be analyzed, and corresponding weights can be assigned to these intermediate goals. According to the intermediate goals and the ultimate goal, the corresponding reverse path can be generated.

[0166] Step S803, according to the reverse path, adjust the initial planning path, so that step S102 adjusts the adjusted initial planning path based on the progress data.

[0167] In this embodiment, the difference information such as consistency or weight information between the intermediate goals and process goals in the reverse path can be compared, and the initial planning path can be adjusted using the difference information. If the difference is not significant, no adjustment is required.

[0168] In an optional implementation manner, the above step S803 adjusts the initial planned path according to the reversed path, which may include the following steps:

[0169] Determining adjustment directions of the plurality of process objectives according to difference information between the plurality of intermediate objectives and the plurality of process objectives of the reverse path;

[0170] The multiple process objectives are adjusted according to the adjustment direction to obtain the adjusted initial planning path, so that the process objectives of the adjusted initial planning path match the intermediate objectives of the reverse path.

[0171] Exemplarily, the intermediate goal in the backward path is to improve the grades of various subjects, strengthen the bonus items, etc., and the initial planning path is generated based on the user's data information. Since the user has no interest in the bonus items, that is, there is no process goal of strengthening the bonus items in the initial planning path, the difference information is from existence to non-existence, and the adjustment direction is negative, that is, no adjustment is required; as another example, for improving the grades of various subjects, assuming that the intermediate goal planned in the backward path is to improve the math score to 90 points, and the process goal is to improve the data score to 92 points, the difference information is not much different, the adjustment direction is none, and no adjustment is required. If the process goal is to improve the data score to 60, the adjustment direction is positive, that is, the process goal is adjusted toward the intermediate goal.

[0172] It should be noted here that the above is only a logical example to facilitate understanding of the relationship between the adjustment direction and the difference information. In actual application, the intermediate target may be more complicated.

[0173] In some embodiments, the intermediate goal may correspond to a secondary goal path, an intermediate goal path, a short-term goal path, and a current task, such as Fig. 9The reverse process shown in the figure. The specific reverse process is as follows: the ultimate goal definition, the server can obtain the ultimate goal (such as entering a top university, achieving a certain professional achievement, winning an international award, etc.) from the user input or the built-in goal library, decompose the ultimate goal into phased goals, and decompose the long-term ultimate goal into multiple phased intermediate goals (such as achieving a certain college entrance examination score, ranking among the top in a key competition, etc.). The goals of each stage can constitute the necessary conditions for achieving the ultimate goal. In the reverse process, the requirements and resources of each stage goal can be analyzed. For each stage goal, for example, the resources and conditions required to achieve the goal, such as skill level, knowledge field, time management, etc., can be analyzed. And key steps and milestones can be identified. By identifying and setting key steps (such as specific exams, academic research or extracurricular activities), and marking some key milestones, these steps or milestones are the necessary basis for achieving a certain intermediate goal. Combined with the above process, a reverse path is generated. For example, based on the analysis of key steps and resources, the path is gradually reversed from the latest time of achieving the ultimate goal. Each step is closely dependent on the completion of the next step (i.e., the previous step). To improve the accuracy of the backward path generation, the rationality and consistency of the path can be verified. For example, the consistency of the generated path in timing logic and feasibility can be checked. By evaluating the requirements of each stage, it is ensured that the planned target path does not conflict with the actual situation and target requirements. If the path verification finds that the setting of a certain stage is unreasonable or the timing conflicts, the target allocation is adjusted, and the specific path is corrected through feedback to ensure that each link is optimized reasonably and the result is optimal globally. When the path is verified and optimized, the final backward path can be output and completed. This path can be further used for the optimization and adjustment of the initial path.

[0174] Through the above technical solution, the established long-term ultimate goal is used to reversely infer the goals of each stage and the corresponding paths, and the reversed path is used to optimize the initial planning path generated forward, which plays a guiding role in the initial planning path and improves the availability of the planned path.

[0175] After the initial planning path is generated, considering that the static planning path is difficult to adapt to the actual planning needs of the user (such as changes in the user's interests and hobbies during the journey or other objective factors), in order to improve the efficiency and accuracy of dynamic adjustment, this embodiment combines the large model technology to dynamically adjust the initial planning path, thereby improving the optimization effect of the path. Next, the above step S102 adjusts the initial planning path based on the progress data to obtain the planning path, which is further introduced:

[0176] Inputting the progress data into a second machine learning model, processing the progress data based on the second machine learning model, and obtaining a progress trend and a progress generating factor corresponding to the progress data;

[0177] Based on the progress trend and the progress generating factor, adjusting the corresponding process target in the initial planning path to obtain the planning path;

[0178] The second machine learning model is a model trained based on multi-user sample data and used to predict progress trends and progress generation factors.

[0179] Exemplarily, the training process of the second machine learning model may be to extract basic information of each user sample in multi-user sample data and data information about the service to be planned, obtain the progress data and progress influencing factors of the user sample for each user sample, train the large language model according to the multi-user sample data and the progress data and progress influencing factors corresponding to each user sample, and obtain the second machine learning model. The specific training process can be found in the embodiments below and will not be elaborated here.

[0180] It should be noted that the progress data of each process goal in this embodiment may include the progress data of each sub-goal in each process goal, so that a more refined prediction can be achieved when predicting the progress trend and tracing back the factors causing the progress, thereby making a more refined path adjustment.

[0181] Furthermore, in order to improve the model processing efficiency and thus efficiently adjust the planned path, this embodiment may further include the following steps before inputting the progress data into the second machine learning model:

[0182] Preprocessing the progress data, and extracting feature data of the preprocessed progress data to obtain progress features;

[0183] The progress feature is converted into format feature data corresponding to the second machine learning model to obtain converted progress data, so as to input the converted progress data into the second machine learning model.

[0184] In this embodiment, the preprocessing of progress data may include data cleaning, standardization, and structuring, etc., to ensure that the input data is consistent with the feature construction requirements in the large model. After preprocessing, feature engineering is used to extract and convert key features that can reflect progress data, such as key features that can reflect user learning and growth characteristics in education planning. Since the input of the model may have certain format requirements, the original feature data can be converted and encoded, and the converted features are used as the input of the large model, so that the data input format meets the requirements of the large model.

[0185] To facilitate the understanding of the real-time adjustment process, the real-time adjustment of the planned path is further introduced below with a specific example. Fig.10As shown, the following process is included:

[0186] S1001. Real-time data collection. The real-time monitoring system captures and collects various types of user-related data, such as academic performance, psychological state, current progress, competition performance, etc.

[0187] S1002: Data preprocessing: performing preprocessing operations on the data, such as cleaning, standardization, and structuring, to ensure that the input data is consistent with the feature construction requirements in the large model.

[0188] S1003, feature construction, after preprocessing, through the feature engineering step, key features that can reflect the user's learning and growth characteristics are extracted and converted. These features will serve as input for large model analysis.

[0189] S1004: Data feature conversion: converting the original features or engineering features into data formats and encoding them so that the data input format meets the requirements of the large model.

[0190] S1005: Pass it to the input layer of the large model. The constructed features are sent to the input layer of the large model to prepare for real-time analysis.

[0191] S1006. The big model performs real-time analysis. The big model analyzes and processes the input data and features, which may include running its embedded complex deep learning, natural language processing and multimodal models to draw further judgments and conclusions.

[0192] S1007, Pattern Recognition and Classification, large models automatically identify patterns, categories, and structural information from data, which helps to identify some important dynamic features in individual learning trajectories, such as large fluctuations in short-term performance.

[0193] S1008, trend prediction. The big model can use time series analysis, space-time characteristics and other methods to predict future trends. This step determines the students’ possible next path or behavioral characteristics.

[0194] S1009. Causal inference, for example, the causal model in the model can be used to infer the causal relationship of the underlying data, such as which factors lead to learning improvement, decline or success.

[0195] S1010. Adjust and optimize the planned path. Based on the patterns, trends and causal reasoning conclusions generated by the large model, the module optimizes the learning and growth path in real time to generate the best path adjustment plan suitable for the current situation.

[0196] S1011. Feedback and action suggestions can also be generated, generating detailed feedback for users and execution systems, including optimization paths, action suggestions, future plans, and path descriptions on how to continue tuning or relearning.

[0197] Next, the new optimization path can be released. That is, after internal verification, the results of the optimization adjustment will be published to the system execution level, ready to push the new path, and the published and updated planning path will be pushed down to the execution module for further implementation (such as adjustment of learning plans, task allocation, resource recommendations, etc.).

[0198] Through the above technical solution, the planned path can be dynamically updated according to the user's real-time data to optimize the performance of the planned path.

[0199] Considering that the data source of the relevant data is single, it may be difficult to accurately profile the user, thereby affecting the accuracy of the generated planning path. In the above steps of this embodiment, obtaining the basic information of the user and the data information for the service to be planned may include the following steps:

[0200] Acquire initial user data input by a user, wherein the initial user data is multimodal data;

[0201] The initial user data is analyzed and integrated to obtain basic information of the user and data information for the service to be planned.

[0202] It should be noted that the specific content of the multimodal data has been introduced in the above embodiment, and the relevant description and beneficial effects are not repeated here. The analysis and integration of the initial user data can be processed by a multimodal machine learning model.

[0203] In some embodiments, to further optimize the user experience, this embodiment may also recommend personalized service resources to the user. Specifically, the method may further include the following steps:

[0204] According to the planned path, personalized service resources corresponding to each process target stage are recommended to the user.

[0205] Exemplarily, the personalized service resources may include course recommendations (such as online video courses, exercise question banks, tutoring books, etc.), simulated exam systems, and time management tools, etc. This not only improves the user's learning experience, but also helps the user achieve the ultimate goal more effectively.

[0206] Compared with the planning tools in the related art, which rely on the static and single-dimensional data of most users to plan a single short-term goal for new users, without considering the multi-faceted development of users based on their own factors, ignoring the comprehensiveness and personalized needs of users, and it is difficult to adjust the planning path in real time according to the real-time data update of users. For the planning of long-term complex goals, there is a lack of calculation ability of the reverse path, and it is impossible to effectively trace back the path requirements from the final goal. This embodiment generates an initial planning path for the user about the service to be planned according to the basic information of the user and the data information of the user for the service to be planned, and the initial planning path includes multiple process goals, monitors the progress data of the user for the multiple process goals in real time, adjusts the initial planning path based on the progress data, obtains the planning path, and synchronizes the planning path to the user's terminal device to achieve customized personalized multi-objective planning solutions for users, and combines large model technology to make the planning path more accurate and intelligent. In addition, the path reverse technology and multimodal data technology are used to solve the path analysis of long-term complex goals, improve the efficiency of path planning, and provide users with comprehensive and full-perspective path suggestions.

[0207] Fig.11 is a flow chart of a model training method provided by an embodiment of the present disclosure, such as Fig.11 As shown, the method may include the following step S111.

[0208] Step S111: Acquire multi-user sample data, where the multi-user sample data includes basic information of multiple user samples, data information about the service to be planned, and an initial planned path.

[0209] The process of acquiring multi-user sample data has been described in detail in the above embodiments and will not be repeated here.

[0210] Step S112: Obtain the context data and ultimate goal achievement data of the user sample for the initial planned path, and obtain multi-user context data and ultimate goal achievement data.

[0211] Exemplarily, the context data of user samples when using the initial planned path can be collected. These data may include user feedback, behavior data, environmental data, etc. in different contexts, such as user feedback on using services at different times and locations, user behavior patterns in different contexts, etc., and data on whether the user sample has achieved the ultimate goal after using the initial planned path, such as whether the user has achieved the expected service effect, and the collected context data and ultimate goal achievement data of the user sample can be summarized and sorted to form multi-user context data and ultimate goal achievement data.

[0212] Step S113: training a large language model according to the multi-user sample data, the multi-user situational data and the ultimate goal achievement data to obtain a first machine learning model.

[0213] Before the model training phase, multi-user sample data, multi-user context data, and ultimate goal achievement data can also be pre-processed to form a data set suitable for training a large language model, for example, data cleaning, data standardization, feature extraction, etc.

[0214] Exemplarily, the above-mentioned processed training data can be used to train a large language model, for example, using a deep learning framework (such as TensorFlow, PyTorch) to train the model, adjust model parameters, optimize model performance, and use a validation data set to validate the trained model to evaluate the model's prediction and optimization capabilities. Based on the validation results, the model is further optimized and adjusted to improve the accuracy and robustness of the model. In addition, the trained first machine learning model is deployed in an actual application environment for predicting and optimizing the planning path. For example, the model can be integrated into a server to process the user's planning path request in real time and provide optimization suggestions.

[0215] The model trained in the above manner can be used to optimize the tuning process of the initial planned path, making the initial planned path more feasible.

[0216] Fig.12 It is a flowchart of another model training method provided by an embodiment of the present disclosure, which may include the following steps S121-S123.

[0217] Step S121: extract basic information of user samples and data information about the service to be planned from the multi-user sample data.

[0218] The process of acquiring multi-user sample data has been described in detail in the above embodiments and will not be repeated here.

[0219] Step S122: Obtain progress data and progress influencing factors of the user sample for the planned path.

[0220] In this embodiment, progress data of each user sample can be obtained, which reflects the progress of the user in the planned path corresponding to the service to be planned. For example, the progress data in the learning plan can be the number of courses completed by the user, test scores, study time, etc.

[0221] By identifying and collecting factors that affect user progress, which can be internal factors (such as users' learning habits and learning methods) and external factors (such as the quality of learning resources and the impact of the environment), and integrating user sample data, progress data and progress influencing factors, a complete user sample data set is formed.

[0222] Step S123: training the large language model according to the multi-user sample data, the progress data corresponding to the user samples, and the progress influencing factors to obtain a second machine learning model.

[0223] Before model training, as mentioned above, the user sample data set can be preprocessed first, and the relevant instructions will not be repeated here. Next, enter the model training process. You can select a suitable large language model (such as GPT, etc.) as the basic model and input the user sample data set into the model for training. During the training process, the model learns the relationship between the user's behavior pattern and the factors affecting the progress. Through methods such as cross-validation and grid search, it adjusts the model's hyperparameters, optimizes the model performance, and evaluates the trained model. The performance indicators of the model (such as accuracy, precision, recall, etc.) are calculated, and the model can be verified to ensure that the model can perform well on unseen data, thereby training a more robust model to predict the user's progress trend and progress factors, thereby adjusting the planning path in real time.

[0224] The model trained in the above manner can be applied to the process of dynamically adjusting the planned path, can adapt to user changes and respond intelligently, and improve the intelligence and real-time performance of the planned path.

[0225] Fig.13a is a schematic diagram of a structure of a multi-objective data processing device provided by an embodiment of the present disclosure, such as Fig.13a As shown, the device includes an initial path generation module 131, a dynamic adjustment module 132 and a synchronization module 133, wherein:

[0226] An initial path generation module 131 is configured to generate an initial planning path for the user with respect to the service to be planned based on the basic information of the user and the data information of the user with respect to the service to be planned, wherein the initial planning path includes a plurality of process objectives;

[0227] A dynamic adjustment module 132, which is configured to monitor the progress data of the user for the multiple process goals in real time, and adjust the initial planned path based on the progress data to obtain a planned path;

[0228] The synchronization module 133 is configured to synchronize the planned path to the terminal device of the user.

[0229] In an optional embodiment, the device further comprises:

[0230] A response module, configured to respond to a request for a service to be planned initiated by the user, and obtain basic information of the user and data information for the service to be planned;

[0231] The service to be planned includes at least one of the following: education planning service, tourism planning service, and smart home planning service.

[0232] In an optional implementation manner, the data information carries a preset ultimate goal of the user;

[0233] The initial path generation module 131 includes:

[0234] A forward decomposition unit, configured to forward decompose the ultimate goal into a plurality of process goals according to the basic information of the user and the data information of the user for the service to be planned;

[0235] A weight allocation unit, which is configured to obtain a plurality of sub-goals for achieving the process goal and allocate weights to the sub-goals to obtain a plurality of sub-goals of the process goal and their weights;

[0236] A generating unit is configured to generate the initial planning path according to the multiple process goals, the multiple sub-goals of the process goals and their weights.

[0237] In an optional implementation, the initial path generation module 131 further includes:

[0238] A combined weight unit, which is configured to obtain a combined weight of the process goal according to a combined calculation result of the weights of the sub-goals of the process goal;

[0239] The weight optimization unit is configured to optimize the weights of the sub-goals based on a preset optimization algorithm if the difference between the combined weight of the process objectives and the preset standard weight of the process objectives reaches a preset deviation value, so that the combined weight of the process objectives is within the preset deviation value.

[0240] In an optional implementation, the optimization algorithm includes at least one of the following: a Pareto optimization algorithm, a multi-objective genetic algorithm, a dynamic programming algorithm, and a particle swarm optimization algorithm.

[0241] In an optional embodiment, the device further comprises:

[0242] An inference module, configured to input the initial planning path, the basic information and the data information into a pre-trained first machine learning model, and infer the initial planning path based on the first machine learning model to obtain a prediction result and optimization recommendation information of the initial planning path with respect to a predicted ultimate goal;

[0243] The initial path optimization module is actually configured to optimize the initial planned path according to the prediction result and the optimization recommendation information;

[0244] Among them, the first machine learning model is a model trained based on multi-user sample data and used to predict and optimize the planned path.

[0245] In an optional implementation, the first machine learning model includes a situation simulation module, a future scenario deduction module, and an optimization suggestion module;

[0246] The reasoning module comprises:

[0247] A simulation unit, configured to simulate, through the scenario simulation module, the simulation scenario of the user under each process target for the planned path according to the basic information and the data information, to obtain the scenario data of the user;

[0248] A prediction unit, configured to perform deduction and prediction of the future scenarios of the user for each process target of the planned path according to the situation data through the future scenario deduction module, and obtain a prediction result about the ultimate target;

[0249] An optimization recommendation unit is configured to generate optimization recommendation information through the optimization suggestion module when the prediction result fails to achieve the ultimate goal, and the optimization recommendation information includes at least one of the following: process goal replacement information, process goal weight adjustment information and personalized suggestions for the user.

[0250] In an optional implementation, at least one of the basic information, the data information and the multi-user sample data is multimodal data, and the multimodal data includes modal data of at least two combinations of the following: text, audio, image, video data, biometric data and sensor data.

[0251] In an optional implementation, the dynamic adjustment module includes:

[0252] a model processing unit, configured to input the progress data into a second machine learning model, process the progress data based on the second machine learning model, and obtain a progress trend and a progress generating factor corresponding to the progress data;

[0253] An adjustment unit, configured to adjust the corresponding process target in the initial planning path based on the progress trend and the progress generating factor to obtain the planning path;

[0254] The second machine learning model is a model trained based on multi-user sample data and used to predict progress trends and progress generation factors.

[0255] In an optional implementation, the dynamic adjustment module 132 further includes:

[0256] a feature extraction unit, configured to preprocess the progress data and extract feature data of the preprocessed progress data to obtain progress features;

[0257] A format conversion unit is configured to convert the progress feature into format feature data corresponding to the second machine learning model to obtain converted progress data, so as to input the converted progress data into the second machine learning model.

[0258] In an optional embodiment, the device further comprises:

[0259] A reverse decomposition module, which is configured to reversely decompose the user's ultimate goal into multiple intermediate goals based on historical path planning information;

[0260] A reverse path generation module, configured to generate a reverse path for the user with respect to the service to be planned according to the multiple intermediate goals;

[0261] An initial path adjustment module is configured to adjust the initial planned path according to the reversed path.

[0262] In an optional implementation, the initial path adjustment module includes:

[0263] an adjustment information determining unit, configured to determine adjustment directions of the plurality of process objectives according to difference information between the plurality of intermediate objectives of the reverse path and the plurality of process objectives;

[0264] A target adjustment unit is configured to adjust the multiple process targets according to the adjustment direction to obtain the adjusted initial planning path, so that the process target of the adjusted initial planning path matches the intermediate target of the reverse path.

[0265] In an optional implementation, the response module includes:

[0266] a data acquisition unit, configured to acquire initial user data input by a user, wherein the initial user data is multimodal data;

[0267] A data processing unit is configured to analyze and integrate the initial user data to obtain basic information of the user and data information for the service to be planned.

[0268] In an optional embodiment, the device further comprises:

[0269] The resource recommendation module is configured to recommend personalized service resources corresponding to each process target stage to the user according to the planned path.

[0270] In an exemplary embodiment, to facilitate understanding of the present device, the relationship between the modules in the device provided by the present embodiment is as follows: Fig.13b As shown:

[0271] The user first inputs initial data, and the data enters the device through the user data acquisition module (ie, the response module).

[0272] The multimodal data analysis module (ie, the data acquisition unit) processes the data, generates multivariate information output and sends it to the planning generation module (ie, the initial path generation module 131).

[0273] The planning generation module starts to formulate a preliminary path and sends it to the target reverse module (i.e., the reverse path generation module).

[0274] The target back-pushing module constructs the back-pushing path by the back-pushing method and sends it back to the planning generation module for initial planning path adjustment.

[0275] Next, the path is adjusted in real time by the dynamic adjustment module, and the adjusted path is provided with real-time feedback through the planning generation module. The device provides the user with the final planned path and feedback suggestions through the feedback module (ie, the synchronization module 133).

[0276] During the whole process, the data flow starts from the user input, and the data is preprocessed by the user data collection module. The preprocessed data is passed to the multimodal data analysis module for further data integration and analysis. The analysis results are used as the input of the planning generation module. Through the initial planning, a reasonable path is constructed. The preliminary path is reversed through the target reverse deduction module, and finally a comprehensive adjusted initial planning path is established. The dynamic adjustment module adjusts the existing path according to real-time data to enhance flexibility and timeliness. The optimized path generated at the end is returned to the feedback module, and a comprehensive planning path is generated and pushed to the user end to complete the interactive closed loop.

[0277] It should be noted that the unexplained parts can be understood by referring to the relevant descriptions and effects corresponding to the steps in the embodiments of the method disclosed herein, and no further elaboration will be made here.

[0278] Fig.14is a structural diagram of a model training device provided by an embodiment of the present disclosure, such as Fig.14 As shown, the device includes a first data acquisition module 141, a second data acquisition module 142 and a first training module 143, wherein:

[0279] A first data acquisition module 141, which is configured to acquire multi-user sample data, wherein the multi-user sample data includes basic information of multiple user samples, data information about the service to be planned, and an initial planning path;

[0280] The second data acquisition module 142 acquires the context data and ultimate goal achievement data of the user sample for the initial planned path, and obtains multi-user context data and ultimate goal achievement data;

[0281] The first training module 143 is configured to train the large language model according to the multi-user sample data, the multi-user situational data and the ultimate goal achievement data to obtain a first machine learning model.

[0282] The relevant instructions can be understood by referring to the relevant descriptions and effects corresponding to the steps in the embodiment of the method disclosed in the present invention, and no further details will be given here.

[0283] Fig.15 is a structural diagram of a model training device provided by an embodiment of the present disclosure, such as Fig.15 As shown, the device includes a data extraction module 151, a third data acquisition module 152 and a second training module 153, wherein:

[0284] A data extraction module 151, which is configured to extract basic information of user samples and data information about the service to be planned from the multi-user sample data;

[0285] A third data acquisition module 152, which is configured to obtain the progress data and progress influencing factors of the user sample for the planned path;

[0286] The second training module 153 is configured to train the large language model according to the multi-user sample data, the progress data corresponding to the user samples, and the progress influencing factors to obtain a second machine learning model.

[0287] The relevant instructions can be understood by referring to the relevant descriptions and effects corresponding to the steps in the embodiment of the method disclosed in the present invention, and no further details will be given here.

[0288] Fig.16 A server provided in an embodiment of the present disclosure, such as Fig.16As shown, the server 160 includes: at least one processor 161; and a memory 161 that is communicatively connected to the at least one processor 162; wherein the memory stores instructions that can be executed by the at least one processor 162, and the instructions are executed by the at least one processor 162 so that the at least one processor 162 can execute the multi-objective-based data processing method corresponding to the above-mentioned embodiment.

[0289] The server 160 further includes a receiver 163 and a transmitter 164. The receiver 163 is used to receive instructions and data sent by other devices, and the transmitter 164 is used to send instructions and data to external devices.

[0290] The relevant instructions can be understood by referring to the relevant descriptions and effects corresponding to the steps in the embodiment of the method disclosed in the present invention, and no further details will be given here.

[0291] Fig.17 A server provided in an embodiment of the present disclosure, such as Fig.17 As shown, the server 170 includes: at least one processor 171; and a memory 171 communicatively connected to the at least one processor 172; wherein the memory stores instructions executable by the at least one processor 172, and the instructions are executed by the at least one processor 172 so that the at least one processor 172 can execute the model training method corresponding to the above-mentioned embodiment.

[0292] The server 170 further includes a receiver 173 and a transmitter 174. The receiver 173 is used to receive instructions and data sent by other devices, and the transmitter 174 is used to send instructions and data to external devices.

[0293] The relevant instructions can be understood by referring to the relevant descriptions and effects corresponding to the steps in the embodiment of the method disclosed in the present invention, and no further details will be given here.

[0294] Fig.18 A server provided in an embodiment of the present disclosure, such as Fig.18 As shown, the server 180 includes: at least one processor 181; and a memory 181 communicatively connected to the at least one processor 182; wherein the memory stores instructions executable by the at least one processor 182, and the instructions are executed by the at least one processor 182 so that the at least one processor 182 can execute another model training method corresponding to the above-mentioned embodiment.

[0295] The server 180 further includes a receiver 183 and a transmitter 184. The receiver 183 is used to receive instructions and data sent by other devices, and the transmitter 184 is used to send instructions and data to external devices.

[0296] The relevant instructions can be understood by referring to the relevant descriptions and effects corresponding to the steps in the embodiment of the method disclosed in the present invention, and no further details will be given here.

[0297] An embodiment of the present disclosure also provides a non-transitory computer-readable storage medium storing computer instructions, wherein the computer instructions are used to enable the computer to execute the multi-objective data processing method, or the model training method, or another model training method of an embodiment of the present disclosure.

[0298] The relevant instructions can be understood by referring to the relevant descriptions and effects corresponding to the steps in the embodiment of the method disclosed in the present invention, and no further details will be given here.

[0299] The embodiments of the present disclosure also provide a computer program product, which includes: a computer program, which is stored in a readable storage medium. At least one processor of the server can read the computer program from the readable storage medium, and at least one processor executes the computer program so that the server executes the solution provided by any of the above embodiments.

[0300] Fig.19 A schematic block diagram of an example electronic device 190 that can be used to implement an embodiment of the present disclosure is shown. The electronic device is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device can also represent various forms of mobile devices, such as personal digital processing, cellular phones, smart phones, wearable devices, and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely examples and are not intended to limit the implementation of the present disclosure described and / or claimed herein.

[0301] like Fig.19 As shown, the device 190 includes a computing unit 191, which can perform various appropriate actions and processes according to a computer program stored in a read-only memory (ROM) 192 or a computer program loaded from a storage unit 198 into a random access memory (RAM) 193. In the RAM 193, various programs and data required for the operation of the device 190 can also be stored. The computing unit 191, the ROM 192, and the RAM 193 are connected to each other via a bus 194. An input / output (I / O) interface 805 is also connected to the bus 194.

[0302] A number of components in the device 190 are connected to the I / O interface 805, including: an input unit 196, such as a keyboard, a mouse, etc.; an output unit 197, such as various types of displays, speakers, etc.; a storage unit 198, such as a disk, an optical disk, etc.; and a communication unit 199, such as a network card, a modem, a wireless communication transceiver, etc. The communication unit 199 allows the device 190 to exchange information / data with other devices through a computer network such as the Internet and / or various telecommunication networks.

[0303] The computing unit 191 may be a variety of general and / or special processing components with processing and computing capabilities. Some examples of the computing unit 191 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various dedicated artificial intelligence (AI) computing chips, various computing units running machine learning model algorithms, digital signal processors (DSPs), and any appropriate processors, controllers, microcontrollers, etc. The computing unit 191 performs the various methods and processes described above, such as a multi-objective data processing method. For example, in some embodiments, any of the methods in the above embodiments may be implemented as a computer software program, which is tangibly contained in a machine-readable medium, such as a storage unit 198. In some embodiments, part or all of the computer program may be loaded and / or installed on the device 190 via ROM 192 and / or communication unit 199. When the computer program is loaded into RAM 193 and executed by the computing unit 191, one or more steps of the method provided by any of the embodiments described above may be performed. Alternatively, in other embodiments, the computing unit 191 may be configured to perform the method of the above disclosed embodiment in any other appropriate manner (e.g., by means of firmware).

[0304] Various implementations of the systems and techniques described above herein can be implemented in digital electronic circuit systems, integrated circuit systems, field programmable gate arrays (FPGAs), application specific integrated circuits (ASICs), application specific standard products (ASSPs), systems on chips (SOCs), complex programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various implementations can include: being implemented in one or more computer programs that can be executed and / or interpreted on a programmable system including at least one programmable processor, which can be a special purpose or general purpose programmable processor that can receive data and instructions from a storage system, at least one input device, and at least one output device, and transmit data and instructions to the storage system, the at least one input device, and the at least one output device.

[0305] The program code for implementing the method of the present disclosure may be written in any combination of one or more programming languages. These program codes may be provided to a processor or controller of a general-purpose computer, a special-purpose computer, or other programmable data processing device, so that the program code, when executed by the processor or controller, enables the functions / operations specified in the flow chart and / or block diagram to be implemented. The program code may be executed entirely on the machine, partially on the machine, partially on the machine and partially on a remote machine as a stand-alone software package, or entirely on a remote machine or server.

[0306] In the context of the present disclosure, a machine-readable medium may be a tangible medium that may contain or store a program for use by or in conjunction with an instruction execution system, device, or equipment. A machine-readable medium may be a machine-readable signal medium or a machine-readable storage medium. A machine-readable medium may include, but is not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, device, or equipment, or any suitable combination of the foregoing. A more specific example of a machine-readable storage medium may include an electrical connection based on one or more lines, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.

[0307] To provide interaction with a user, the systems and techniques described herein can be implemented on a computer having: a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to the user; and a keyboard and pointing device (e.g., a mouse or trackball) through which the user can provide input to the computer. Other types of devices can also be used to provide interaction with the user; for example, the feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including acoustic input, voice input, or tactile input).

[0308] The systems and techniques described herein may be implemented in a computing system that includes back-end components (e.g., as a data server), or a computing system that includes middleware components (e.g., an application server), or a computing system that includes front-end components (e.g., a user computer with a graphical user interface or a web browser through which a user can interact with implementations of the systems and techniques described herein), or a computing system that includes any combination of such back-end components, middleware components, or front-end components. The components of the system may be interconnected by any form or medium of digital data communication (e.g., a communication network). Examples of communication networks include: a local area network (LAN), a wide area network (WAN), and the Internet.

[0309] A computer system may include a client and a server. The client and the server are generally remote from each other and usually interact through a communication network. The relationship between the client and the server is generated by computer programs running on the corresponding computers and having a client-server relationship with each other. The server may be a cloud server, also known as a cloud computing server or cloud host, which is a host product in the cloud computing service system to solve the defects of difficult management and weak business scalability in traditional physical hosts and VPS services ("Virtual Private Server", or "VPS" for short). The server may also be a server of a distributed system, or a server combined with a blockchain.

[0310] It should be understood that the various forms of processes shown above can be used to reorder, add or delete steps. For example, the steps recorded in this disclosure can be executed in parallel, sequentially or in different orders, as long as the desired results of the technical solutions disclosed in this disclosure can be achieved, and this document does not limit this.

[0311] The above specific implementations do not constitute a limitation on the protection scope of the present disclosure. It should be understood by those skilled in the art that various modifications, combinations, sub-combinations and substitutions can be made according to design requirements and other factors. Any modification, equivalent substitution and improvement made within the spirit and principle of the present disclosure shall be included in the protection scope of the present disclosure.

Claims

1. A multi-objective data processing method, comprising: Generate an initial planning path for the user with respect to the service to be planned according to the basic information of the user and the data information of the user with respect to the service to be planned, wherein the initial planning path includes a plurality of process objectives; monitoring the progress data of the user for the multiple process objectives in real time, and adjusting the initial planned path based on the progress data to obtain a planned path; The planned path is synchronized to the user's terminal device.

2. The method according to claim 1, further comprising: In response to a request for a service to be planned initiated by the user, obtaining basic information of the user and data information for the service to be planned; The service to be planned includes at least one of the following: education planning service, tourism planning service, and smart home planning service.

3. The method according to claim 1 or 2, wherein: The data information carries the preset ultimate goal of the user; The generating the initial planning path of the user according to the basic information of the user and the data information of the user for the service to be planned includes: According to the basic information of the user and the data information of the user for the service to be planned, the ultimate goal is positively decomposed into multiple process goals; Acquire multiple sub-goals for achieving the process goal, and assign weights to the sub-goals to obtain multiple sub-goals of the process goal and their weights; The initial planning path is generated according to the multiple process goals, the multiple sub-goals of the process goals and their weights.

4. The method according to claim 3, wherein: The generating the initial planned path of the user according to the basic information of the user and the data information of the user for the service to be planned also includes: Obtaining a combined weight of the process goal according to a combined calculation result of the weights of the sub-goals of the process goal; If the difference between the combined weight of the process objectives and the preset standard weight of the process objectives reaches a preset deviation value, the weights of the sub-objectives are optimized based on a preset optimization algorithm so that the combined weight of the process objectives is within the preset deviation value.

5. The method according to claim 4, wherein: The optimization algorithm includes at least one of the following: a Pareto optimization algorithm, a multi-objective genetic algorithm, a dynamic programming algorithm, and a particle swarm optimization algorithm.

6. The method according to any one of claims 1 to 5, after generating the initial planned path, the method further comprises: Inputting the initial planning path, the basic information and the data information into a pre-trained first machine learning model, and inferring the initial planning path based on the first machine learning model to obtain a prediction result and optimization recommendation information of the initial planning path with respect to a predicted ultimate goal; Optimizing the initial planning path according to the prediction result and the optimization recommendation information; Among them, the first machine learning model is a model trained based on multi-user sample data and used to predict and optimize the planned path.

7. The method according to claim 6, wherein: The first machine learning model includes a situation simulation module, a future scenario deduction module, and an optimization suggestion module; The reasoning of the initial planning path based on the first machine learning model to obtain a prediction result and optimization recommendation information of the initial planning path with respect to the ultimate goal includes: By means of the scenario simulation module, according to the basic information and the data information, the simulation scenario of the user for the planned path under each process target is simulated to obtain the scenario data of the user; Through the future scenario deduction module, the future scenarios of the user for the planned path at each process goal are deduced and predicted according to the situation data to obtain a prediction result about the ultimate goal; When the prediction result is that the ultimate goal is not achieved, optimization recommendation information is generated through the optimization suggestion module, and the optimization recommendation information includes at least one of the following: process goal replacement information, process goal weight adjustment information and personalized suggestions for the user.

8. The method according to claim 6, wherein: At least one of the basic information, the data information and the multi-user sample data is multimodal data, and the multimodal data includes modal data of at least two combinations of the following: text, audio, image, video data, biometric data and sensor data.

9. The method according to any one of claims 1 to 8, wherein: The adjusting the initial planned path based on the progress data to obtain a planned path includes: Inputting the progress data into a second machine learning model, processing the progress data based on the second machine learning model, and obtaining a progress trend and a progress generating factor corresponding to the progress data; Based on the progress trend and the progress generating factor, adjusting the corresponding process target in the initial planning path to obtain the planning path; The second machine learning model is a model trained based on multi-user sample data and used to predict progress trends and progress generation factors.

10. The method according to claim 9, before inputting the progress data into the second machine learning model, the method further comprises: Preprocessing the progress data, and extracting feature data of the preprocessed progress data to obtain progress features; The progress feature is converted into format feature data corresponding to the second machine learning model to obtain converted progress data, so as to input the converted progress data into the second machine learning model.

11. The method according to any one of claims 1 to 8, further comprising: According to the historical path planning information, the ultimate goal of the user is reversely decomposed into multiple intermediate goals; Generating a reverse path for the user regarding the service to be planned according to the multiple intermediate goals; The initial planned path is adjusted according to the reversed path.

12. The method according to claim 11, wherein: The adjusting the initial planned path according to the reversed path includes: Determining adjustment directions of the plurality of process objectives according to difference information between the plurality of intermediate objectives and the plurality of process objectives of the reverse path; The multiple process objectives are adjusted according to the adjustment direction to obtain the adjusted initial planning path, so that the process objectives of the adjusted initial planning path match the intermediate objectives of the reverse path.

13. The method according to claim 2, wherein: The acquiring of the basic information of the user and the data information for the service to be planned includes: Acquire initial user data input by a user, wherein the initial user data is multimodal data; The initial user data is analyzed and integrated to obtain basic information of the user and data information for the service to be planned.

14. The method according to any one of claims 1 to 8, further comprising: According to the planned path, personalized service resources corresponding to each process target stage are recommended to the user.

15. A model training method, comprising: Acquire multi-user sample data, wherein the multi-user sample data includes basic information of multiple user samples, data information about the service to be planned, and an initial planning path; Acquire the context data and ultimate goal achievement data of the user sample for the initial planned path, and obtain multi-user context data and ultimate goal achievement data; The large language model is trained according to the multi-user sample data, the multi-user situational data and the ultimate goal achievement data to obtain a first machine learning model.

16. A model training method, comprising: Extracting basic information of user samples and data information about services to be planned from multi-user sample data; Obtaining progress data and progress influencing factors of the user sample for the planned path; The large language model is trained according to the multi-user sample data, the progress data corresponding to the user samples, and the progress influencing factors to obtain a second machine learning model.

17. A multi-objective data processing device, comprising: An initial path generation module, which is configured to generate an initial planning path for the user with respect to the service to be planned based on the basic information of the user and the data information of the user with respect to the service to be planned, wherein the initial planning path includes a plurality of process objectives; A dynamic adjustment module, which is configured to monitor the progress data of the user for the multiple process objectives in real time, and adjust the initial planned path based on the progress data to obtain a planned path; A synchronization module is configured to synchronize the planned path to the user's terminal device.

18. The apparatus according to claim 17, further comprising: A response module, configured to respond to a request for a service to be planned initiated by the user, and obtain basic information of the user and data information for the service to be planned; The service to be planned includes at least one of the following: education planning service, tourism planning service, and smart home planning service.

19. The device according to claim 17 or 18, wherein: The data information carries the preset ultimate goal of the user; The initial path generation module comprises: A forward decomposition unit, configured to forward decompose the ultimate goal into a plurality of process goals according to the basic information of the user and the data information of the user for the service to be planned; A weight allocation unit, which is configured to obtain a plurality of sub-goals for achieving the process goal and allocate weights to the sub-goals to obtain a plurality of sub-goals of the process goal and their weights; A generating unit is configured to generate the initial planning path according to the multiple process goals, the multiple sub-goals of the process goals and their weights.

20. The device according to claim 19, wherein The initial path generation module further includes: A combined weight unit, which is configured to obtain a combined weight of the process goal according to a combined calculation result of the weights of the sub-goals of the process goal; The weight optimization unit is configured to optimize the weights of the sub-goals based on a preset optimization algorithm if the difference between the combined weight of the process objectives and the preset standard weight of the process objectives reaches a preset deviation value, so that the combined weight of the process objectives is within the preset deviation value.

21. The device according to claim 20, wherein: The optimization algorithm includes at least one of the following: a Pareto optimization algorithm, a multi-objective genetic algorithm, a dynamic programming algorithm, and a particle swarm optimization algorithm.

22. The apparatus according to claim 19, further comprising: An inference module, configured to input the initial planning path, the basic information and the data information into a pre-trained first machine learning model, and infer the initial planning path based on the first machine learning model to obtain a prediction result and optimization recommendation information of the initial planning path with respect to a predicted ultimate goal; The initial path optimization module is actually configured to optimize the initial planned path according to the prediction result and the optimization recommendation information; Among them, the first machine learning model is a model trained based on multi-user sample data and used to predict and optimize the planned path.

23. The device according to claim 22, wherein: The first machine learning model includes a situation simulation module, a future scenario deduction module, and an optimization suggestion module; The reasoning module comprises: A simulation unit, configured to simulate, through the scenario simulation module, the simulation scenario of the user under each process target for the planned path according to the basic information and the data information, to obtain the scenario data of the user; A prediction unit, configured to perform deduction and prediction of the future scenarios of the user for each process target of the planned path according to the situation data through the future scenario deduction module, and obtain a prediction result about the ultimate target; An optimization recommendation unit is configured to generate optimization recommendation information through the optimization suggestion module when the prediction result fails to achieve the ultimate goal, and the optimization recommendation information includes at least one of the following: process goal replacement information, process goal weight adjustment information and personalized suggestions for the user.

24. The device according to claim 22, wherein: At least one of the basic information, the data information and the multi-user sample data is multimodal data, and the multimodal data includes text, audio, image, video data, biometric data and sensor data.

25. The device according to any one of claims 17 to 24, wherein: The dynamic adjustment module includes: a model processing unit, configured to input the progress data into a second machine learning model, process the progress data based on the second machine learning model, and obtain a progress trend and a progress generating factor corresponding to the progress data; An adjustment unit, configured to adjust the corresponding process target in the initial planning path based on the progress trend and the progress generating factor to obtain the planning path; The second machine learning model is a model trained based on multi-user sample data and used to predict progress trends and progress generation factors.

26. The apparatus according to claim 25, wherein the dynamic adjustment module further comprises: a feature extraction unit, configured to preprocess the progress data and extract feature data of the preprocessed progress data to obtain progress features; A format conversion unit is configured to convert the progress feature into format feature data corresponding to the second machine learning model to obtain converted progress data, so as to input the converted progress data into the second machine learning model.

27. The device according to any one of claims 17 to 24, further comprising: A reverse decomposition module, which is configured to reversely decompose the user's ultimate goal into multiple intermediate goals based on historical path planning information; A reverse path generation module, configured to generate a reverse path for the user with respect to the service to be planned according to the multiple intermediate goals; An initial path adjustment module is configured to adjust the initial planned path according to the reversed path.

28. The device according to claim 27, wherein The initial path adjustment module includes: an adjustment information determining unit, configured to determine adjustment directions of the plurality of process objectives according to difference information between the plurality of intermediate objectives of the reverse path and the plurality of process objectives; A target adjustment unit is configured to adjust the multiple process targets according to the adjustment direction to obtain the adjusted initial planning path, so that the process target of the adjusted initial planning path matches the intermediate target of the reverse path.

29. The device according to claim 18, wherein The response module comprises: a data acquisition unit, configured to acquire initial user data input by a user, wherein the initial user data is multimodal data; A data processing unit is configured to analyze and integrate the initial user data to obtain basic information of the user and data information for the service to be planned.

30. The device according to any one of claims 17 to 24, further comprising: The resource recommendation module is configured to recommend personalized service resources corresponding to each process target stage to the user according to the planned path.

31. A model training device, comprising: A first data acquisition module, which is configured to acquire multi-user sample data, wherein the multi-user sample data includes basic information of multiple user samples, data information about the service to be planned, and an initial planning path; A second data acquisition module is configured to acquire the context data and ultimate goal achievement data of the user sample for the initial planned path, and obtain multi-user context data and ultimate goal achievement data; The first training module is configured to train the large language model according to the multi-user sample data, the multi-user situational data and the ultimate goal achievement data to obtain a first machine learning model.

32. A model training device, comprising: A data extraction module, which is configured to extract basic information of user samples and data information about the service to be planned from the multi-user sample data; A third data acquisition module, configured to acquire the progress data and progress influencing factors of the user sample for the planned path; The second training module is configured to train the large language model according to the multi-user sample data, the progress data corresponding to the user samples, and the progress influencing factors to obtain a second machine learning model.

33. A server, comprising: at least one processor; as well as a memory communicatively connected to the at least one processor; wherein, The memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can execute the multi-objective data processing method described in any one of claims 1 to 14, or the model training method described in claim 15, or the model training method described in claim 16.

34. A non-transitory computer-readable storage medium storing computer instructions, wherein: The computer instructions are used to enable the computer to execute the multi-objective data processing method according to any one of claims 1-14, or the model training method according to claim 15, or the model training method according to claim 16.

35. A computer program product, comprising a computer program, which, when executed by a processor, implements the multi-objective data processing method described in any one of claims 1 to 14, or the model training method described in claim 15, or the model training method described in claim 16.

Images