Intelligent decision-making method and system for server customization demand matching

Abstract

The invention relates to an intelligent decision-making method and system for server customization demand matching. The method comprises the steps of obtaining a server customization demand input by a user; constructing a multi-dimensional demand portrait based on the server customization demand, wherein the demand portrait comprises a quantitative index and a configuration constraint condition; according to the demand portrait, executing configuration compatibility check, and identifying whether a compatibility problem exists or not; when the compatibility check is passed, a candidate configuration scheme is generated by adopting a multi-target genetic algorithm based on the demand portrait passing the compatibility check; and sorting and screening the candidate configuration schemes based on user preference data in the demand portraits passing the compatibility check to obtain a recommended scheme list, and transmitting the recommended scheme list to a display interface for a user to select and confirm, the recommended scheme comprising a component list, a price, a performance index and a recommendation reason. The method has the effect of improving the accuracy of server customization recommendation.

Description

Technical Field

[0001] This application relates to the technical field of server configuration and intelligent decision-making, and in particular to an intelligent decision-making method and system for matching server customization needs. Background Technology

[0002] Currently, with the rapid development of cloud computing, big data, and artificial intelligence technologies, servers are playing an increasingly important role in enterprise-level data processing, high-performance computing, and various business deployments. Different users or enterprises have different customized needs for server performance, cost, energy consumption, etc., and server selection platforms or online configuration tools that support personalized configurations have gradually emerged in the market to assist users in selecting hardware component combinations and making purchasing decisions.

[0003] Existing server configuration recommendation methods are mostly based on static templates, manual rules, or experience base matching. They typically only support users to input basic parameters through structured forms and rely on static priorities or manually set scoring models to sort the results. This makes it difficult to cover the logical constraints between complex configuration combinations, lacks dynamic adjustment and intelligent optimization capabilities, and traditional recommendation methods that rely on human experts are inefficient and inconsistent, making it difficult to achieve the optimal balance between cost, performance, and business needs.

[0004] The existing technical solutions mentioned above have the following drawbacks: existing server configuration recommendation methods are mostly based on static templates, manual rules, or experience base matching, which makes it difficult to cover the logical constraints between complex configuration combinations, resulting in low intelligence and low recommendation accuracy in the recommendation process, and therefore there is room for improvement. Summary of the Invention

[0005] To improve the accuracy of server customization recommendations, this application provides an intelligent decision-making method and system for matching server customization needs.

[0006] The above-mentioned objective of this application is achieved through the following technical solution: A smart decision-making method for matching server customization requirements, the method comprising: Obtain server customization requirements input by the user, including performance indicators, cost budget, power consumption limits, application scenario information, and user preference data; Based on the server customization requirements, a multi-dimensional requirement profile is constructed, which includes quantitative indicators and configuration constraints. Based on the aforementioned requirement profile, perform a configuration compatibility check to identify any compatibility issues. When the compatibility check passes, a multi-objective genetic algorithm is used to generate candidate configuration schemes based on the requirement profile that passed the compatibility check. Based on the user preference data in the requirement profile that passed the compatibility check, the candidate configuration schemes are sorted and filtered to obtain a list of recommended schemes. The list of recommended schemes is then transmitted to the display interface for users to select and confirm. The recommended schemes include a component list, price, performance indicators, and reasons for recommendation.

[0007] By adopting the above technical solutions, and by acquiring the server customization requirements input by users, key parameter information affecting server configuration selection can be fully collected, thus providing an accurate data foundation for subsequent intelligent recommendation decisions. By constructing a multi-dimensional requirement profile based on server customization requirements, the raw input can be transformed into a unified quantitative structure that can be processed by the algorithm, thereby ensuring that requirement information from different sources participates in decision-making collaboratively under the same model. By performing configuration compatibility checks, component combinations that do not meet actual installation or operating conditions can be identified in advance before candidate solutions are generated, thereby avoiding the generation of unfeasible or unreliable configuration solutions. By using a multi-objective genetic algorithm to generate candidate configuration solutions, better hardware combinations can be efficiently searched while meeting performance, cost, and power consumption constraints, thereby improving the overall superiority of recommended configurations. By sorting and filtering candidate configuration solutions based on user preference data, personalized output of recommendation results can be achieved, thereby improving user experience and increasing the adoption rate of recommended solutions.

[0008] In one example, this application can be further configured such that the server customization requirements for obtaining user input specifically include: Collect user input information, which includes structured parameters filled in by form, configuration options submitted by drop-down selection, and customized intents expressed by natural language input. The user input information is converted into a structured data format to form the server-customized requirements, which can then be called upon in the subsequent requirement profile generation step.

[0009] By adopting the above technical solutions and collecting user input information, including structured parameters, configuration options, and natural language intent, we can comprehensively cover server configuration requirements under different expression methods, thereby reducing recommendation errors caused by user input bias. By converting user input information into a structured data format, we can avoid semantic ambiguity caused by the direct use of unstructured information, thereby improving the accuracy of subsequent requirement profile construction and compatibility verification.

[0010] In one example, this application can be further configured as follows: the construction of a multi-dimensional requirement profile based on the server customization requirements specifically includes: Extract performance metrics, cost budgets, power consumption limits, application scenario information, and user preference data from the server customization requirements, and establish a requirement parameter vector containing multiple dimensions, including performance, cost, power consumption, and user preference dimensions. Based on a pre-defined NLP model, the natural language content in the user input information is semantically parsed to extract key configuration intents. Combined with a rule engine, each parameter item is standardized to form a unified format requirement profile object for subsequent compatibility checks and candidate configuration scheme generation steps.

[0011] By adopting the above technical solutions, and by extracting demand parameters such as performance, cost, power consumption, application scenarios, and user preferences, and constructing a demand parameter vector, a complete demand model containing different optimization dimensions can be established, thereby improving the consistency between the goals of solution optimization and selection. By performing semantic parsing on natural language content and combining it with a rule engine for standardization, it is possible to ensure that user customization intentions are accurately understood and structurally expressed, thereby improving the adaptability of the recommendation system to complex demands.

[0012] In one example, this application can be further configured as follows: Performing a configuration compatibility check based on the requirement profile specifically includes: The hardware component knowledge graph in the preset configuration knowledge base is invoked. Based on the component combination in the requirement profile, the constraint rules between each component are identified and extracted. The constraint rules include slot matching, power adaptation, interface standard, and space compatibility. Based on the constraint rules, the rule engine performs logical reasoning on the component combination to determine whether there are compatibility issues in the component combination. The compatibility issues include hardware parameter conflicts, structural incompatibility, or functional exclusion. When the compatibility check passes, the requirement profile is marked as a feasible configuration input for use in the candidate configuration scheme generation step.

[0013] By adopting the above technical solutions, and by calling the hardware component knowledge graph and extracting component constraint rules, it is possible to identify physical and functional contradictions between server components using prior knowledge, thereby improving the professionalism and coverage of configuration checks. By performing logical reasoning based on the rule engine to judge compatibility issues, risky configuration combinations can be excluded before the recommendation stage, thereby avoiding downstream recommendations that cannot be installed or fail to run. By marking the configurations as feasible inputs after the compatibility checks are passed, invalid solutions can be reduced from entering the optimization calculation process, thereby improving the overall recommendation efficiency and feasibility.

[0014] In one example, this application can be further configured as follows: the requirement profile based on the compatibility check passes, and the generation of candidate configuration schemes using a multi-objective genetic algorithm, specifically includes: Using the requirement profile that has passed the compatibility check as input, hardware component combinations are extracted from the preset configuration knowledge base to construct an initial configuration scheme population; Based on the performance indicators, cost budget, and power consumption limits in the requirement profile that passed the compatibility check, a multi-objective fitness function is defined, and an evolutionary operation is performed on the initial configuration scheme population, the evolutionary operation including selection, crossover, and mutation. When the preset evolution termination condition is met, multiple optimized candidate configuration schemes are output for subsequent sorting and filtering steps.

[0015] By adopting the above technical solutions, and constructing an initial population based on a requirement profile that has passed compatibility checks, the evolutionary algorithm can be guaranteed to search efficiently within the feasible solution space, thereby improving the effectiveness of the evolutionary calculation results. By defining a multi-objective fitness function based on performance, budget, and power consumption and performing evolutionary operations, the key objectives of server recommendations can be comprehensively balanced, thereby obtaining a multi-solution configuration set with better overall performance. By outputting optimized candidate configuration schemes based on preset termination conditions, the computational load of the algorithm can be reduced and the recommendation response speed can be improved, thereby improving user interaction efficiency.

[0016] In one example, this application can be further configured as follows: sorting and filtering the candidate configuration schemes based on user preference data in the requirement profile that has passed the compatibility check, and obtaining a recommended scheme list, specifically includes: An initial weighted scoring model is constructed based on the user preference weights in the requirement profile that passed the compatibility check. The preference weights include performance priority, budget sensitivity, and power consumption control tendency. The initial weighted scoring model is optimized based on historical user feedback data to obtain a multi-attribute weighted scoring model. The multi-attribute weighted scoring model is used to calculate the scores of the candidate configuration schemes on each target dimension, and the scores of each dimension are weighted and fused to obtain a comprehensive score. All candidate configuration schemes are sorted based on the comprehensive score to form a recommended scheme list.

[0017] By adopting the above technical solutions and constructing a weighted scoring model based on user preference weights, we can reflect users' subjective emphasis on performance, cost, and energy consumption, thereby improving the personalized matching degree of the recommended solutions. By combining historical user feedback to optimize the scoring model, we can enhance the model's ability to fit the user's real behavioral characteristics, thereby achieving continuous evolution and performance improvement of the recommendation strategy. By weighted scoring and ranking of all candidate solutions, we can select configuration solutions that better meet the priority of needs, thereby significantly improving the user acceptance and adoptability of the recommendation results.

[0018] In one example, this application can be further configured as follows: the intelligent decision-making method for matching server customization needs also includes: Collect user behavior data, including solution click records, preference weight adjustment history, and solution rejection feedback information; The user behavior data is used as training samples, and an incremental learning algorithm is used to dynamically adjust the multi-attribute weighted scoring model.

[0019] By adopting the above technical solutions, and by collecting user click behavior, weight adjustment, and rejection feedback data, it is possible to collect user experience deviations and decision-making patterns in the selection process in real time, thereby accurately depicting the changing preference characteristics of users; by performing incremental learning based on user behavior data, it is possible to continuously optimize the strategy parameters of the multi-attribute scoring model, thereby enabling the system to have adaptive learning capabilities and maintain recommendation accuracy in the long term.

[0020] The second objective of this invention is achieved through the following technical solution: An intelligent decision-making system for matching server customization needs, the intelligent decision-making system for matching server customization needs includes: The requirement acquisition module is used to acquire the server customization requirements input by the user. The server customization requirements include performance indicators, cost budget, power consumption limits, application scenario information, and user preference data. The profile building module is used to build a multi-dimensional requirement profile based on the server customization requirements. The requirement profile includes quantitative indicators and configuration constraints. The compatibility check module is used to perform configuration compatibility checks based on the requirement profile to identify whether there are compatibility issues. The candidate generation module is used to generate candidate configuration schemes based on the requirement profile that has passed the compatibility check, using a multi-objective genetic algorithm. The scheme sorting module is used to sort and filter the candidate configuration schemes based on the user preference data in the requirement profile that has passed the compatibility check, obtain a list of recommended schemes, and transmit the list of recommended schemes to the display interface for the user to select and confirm. The recommended schemes include a component list, price, performance indicators, and reasons for recommendation.

[0021] By adopting the above technical solutions, and by acquiring the server customization requirements input by users, key parameter information affecting server configuration selection can be fully collected, thus providing an accurate data foundation for subsequent intelligent recommendation decisions. By constructing a multi-dimensional requirement profile based on server customization requirements, the raw input can be transformed into a unified quantitative structure that can be processed by the algorithm, thereby ensuring that requirement information from different sources participates in decision-making collaboratively under the same model. By performing configuration compatibility checks, component combinations that do not meet actual installation or operating conditions can be identified in advance before candidate solutions are generated, thereby avoiding the generation of unfeasible or unreliable configuration solutions. By using a multi-objective genetic algorithm to generate candidate configuration solutions, better hardware combinations can be efficiently searched while meeting performance, cost, and power consumption constraints, thereby improving the overall superiority of recommended configurations. By sorting and filtering candidate configuration solutions based on user preference data, personalized output of recommendation results can be achieved, thereby improving user experience and increasing the adoption rate of recommended solutions.

[0022] In summary, this application includes the following beneficial technical effects: 1. By acquiring user-inputted server customization requirements, it is possible to fully collect key parameter information that affects server configuration selection, thereby providing an accurate data foundation for subsequent intelligent recommendation decisions; by constructing a multi-dimensional requirement profile based on server customization requirements, it is possible to transform the raw input into a unified quantitative structure that can be processed by the algorithm, thereby ensuring that requirement information from different sources participates in decision-making collaboratively under the same model; by performing configuration compatibility checks, it is possible to identify component combinations that do not meet actual installation or operating conditions in advance before candidate solutions are generated, thereby avoiding the generation of unimplementable or unreliable configuration solutions; 2. By employing a multi-objective genetic algorithm to generate candidate configuration schemes, it is possible to efficiently search for better hardware combinations while meeting performance, cost, and power consumption constraints, thereby improving the overall superiority of the recommended configuration. By sorting and filtering candidate configuration schemes based on user preference data, personalized output of recommendation results can be achieved, thereby improving user experience and increasing the adoption rate of recommended schemes. Attached Figure Description

[0023] Figure 1 This is a flowchart of an intelligent decision-making method for matching server customization requirements in one embodiment of this application; Figure 2 This is a flowchart illustrating the implementation of step S10 in an intelligent decision-making method for matching customized server requirements in one embodiment of this application. Figure 3 This is a flowchart illustrating the implementation of step S20 in an intelligent decision-making method for matching customized server requirements in one embodiment of this application. Figure 4This is a flowchart illustrating the implementation of step S30 in an intelligent decision-making method for matching customized server requirements in one embodiment of this application. Figure 5 This is a flowchart illustrating the implementation of step S40 in an intelligent decision-making method for matching customized server requirements in one embodiment of this application. Figure 6 This is a flowchart illustrating the implementation of step S50 in an intelligent decision-making method for matching customized server requirements in one embodiment of this application. Figure 7 This is another implementation flowchart of an intelligent decision-making method for matching server customization requirements in one embodiment of this application; Figure 8 This is a principle block diagram of an intelligent decision-making system for matching customized server requirements in one embodiment of this application. Detailed Implementation

[0024] The present application will be further described in detail below with reference to the accompanying drawings.

[0025] In one embodiment, such as Figure 1 As shown, this application discloses an intelligent decision-making method for matching server customization requirements, which specifically includes the following steps: S10: Obtain server customization requirements input by the user, including performance indicators, cost budget, power consumption limits, application scenario information, and user preference data.

[0026] Specifically, the front-end input interface guides users to fill in server performance indicators, cost budgets, power consumption limits, and business usage scenarios. It also allows users to set preference parameters such as performance priority or energy saving tendencies through checkboxes. All input data is collected through a unified form submission interface, and structured parameters and unstructured expressions are marked and stored separately for feature extraction and profile generation by the subsequent requirements modeling module. In the structured input, users can select technical specifications such as the number of CPU cores, memory capacity, and hard drive type and quantity. In the unstructured description, users can fill in intent-type text such as "deployed for AI model training tasks, requiring support for GPU parallelism," thereby ensuring that user input information covers both general and personalized requirements.

[0027] S20: Build a multi-dimensional requirement profile based on server customization needs. The requirement profile includes quantitative indicators and configuration constraints.

[0028] Specifically, by invoking the requirements modeling logic, various customized parameters input by users are mapped to a standard parameter system. In the performance dimension, hard indicators such as CPU frequency, memory capacity, and storage type are extracted and their units are normalized. In the budget dimension, they are mapped to the acceptable unit price limit and total budget range. In the power consumption dimension, the maximum power consumption and heat dissipation tolerance parameters set by the user are extracted. In the scenario dimension, workload types such as "database", "video transcoding" or "AI training" are identified and labeled based on keywords. All processed indicators are combined into a parameter vector. At the same time, potential configuration constraints such as "GPU must be compatible with motherboard" or "storage controller must support RAID5" are identified and uniformly encapsulated into a structured requirement profile data object to support compatibility judgment and solution generation processes.

[0029] S30: Based on the requirement profile, perform a configuration compatibility check to identify any compatibility issues.

[0030] Specifically, by loading a hardware component knowledge graph and matching the component selection schemes in the demand profile with the structural rules stored in the graph, the system identifies whether the motherboard and CPU socket are compatible, whether the memory specifications are supported, whether the power supply's rated power meets the total power consumption requirements of the components, and whether the chassis space allows the selected GPU or cooler to be installed. Combined with the inference engine's logical judgment, if there are parameter conflicts or structural mutual exclusions, they are marked as "incompatible". For example, when a user selects an E-ATX motherboard and a Mini-Tower chassis, the system can identify the problem of incompatible installation space and thus refuse the combination from entering the optimization stage, ensuring that all configuration combinations that enter the next step of optimization calculation are feasible at the physical and structural levels.

[0031] S40: When the compatibility check passes, a multi-objective genetic algorithm is used to generate candidate configuration schemes based on the requirement profile that has passed the compatibility check.

[0032] Specifically, by calling the genetic algorithm computing engine, a set of server component combinations that meet compatibility constraints are initialized as individuals in the population. Each individual consists of hardware components such as CPU, motherboard, memory, storage, network card, and power supply. Based on the performance indicators, cost budget, and power consumption limits in the demand profile, a multi-objective fitness function is constructed. The fitness function considers performance score, cost penalty function, and energy consumption index simultaneously. Through evolutionary operations such as selection, crossover, and mutation, the population is iteratively optimized. After several rounds of evolution, a set of candidate solutions with the best performance in comprehensive indicators is output. For example, in a certain round, a configuration combination that simultaneously meets GPU acceleration performance, budget within limits, and overall power consumption below 400W is obtained, thereby providing users with a set of alternative solutions with more optimization value.

[0033] S50: Based on user preference data in the requirement profile that has passed the compatibility check, sort and filter the candidate configuration schemes to obtain a list of recommended schemes, and transmit the list of recommended schemes to the display interface for users to select and confirm. The recommended schemes include a list of components, price, performance indicators and reasons for recommendation.

[0034] Specifically, by constructing a weighted scoring model based on user preference data, different weights are assigned to the scores of each dimension in the candidate solutions according to the performance priority, budget sensitivity, and power consumption control tendency set by the user. After calculating the score of each candidate solution on the target dimension, the scores are weighted and fused to obtain a comprehensive score. Then, the solutions are sorted from high to low and the top-ranked configuration solutions are extracted as a recommendation list. The recommended solutions can display the component list, total price, and core performance indicators such as CPU benchmark scores or overall power consumption assessment values ​​for each configuration. At the same time, the recommendation reason text is automatically generated, such as "This solution is recommended because it achieves the optimal balance between performance and budget, and is suitable for lightweight AI inference scenarios". The recommendation results are pushed to the front-end interface through the interface for users to quickly browse and select.

[0035] By adopting the above technical solutions, and by acquiring the server customization requirements input by users, key parameter information affecting server configuration selection can be fully collected, thus providing an accurate data foundation for subsequent intelligent recommendation decisions. By constructing a multi-dimensional requirement profile based on server customization requirements, the raw input can be transformed into a unified quantitative structure that can be processed by the algorithm, thereby ensuring that requirement information from different sources participates in decision-making collaboratively under the same model. By performing configuration compatibility checks, component combinations that do not meet actual installation or operating conditions can be identified in advance before candidate solutions are generated, thereby avoiding the generation of unfeasible or unreliable configuration solutions. By using a multi-objective genetic algorithm to generate candidate configuration solutions, better hardware combinations can be efficiently searched while meeting performance, cost, and power consumption constraints, thereby improving the overall superiority of recommended configurations. By sorting and filtering candidate configuration solutions based on user preference data, personalized output of recommendation results can be achieved, thereby improving user experience and increasing the adoption rate of recommended solutions.

[0036] In one embodiment, such as Figure 2 As shown, in step S10, the server customization requirements input by the user are obtained, which specifically includes: S11: Collect user input information, including structured parameters filled in as a form, configuration options submitted as a drop-down selection, and customized intents expressed in natural language input.

[0037] Specifically, the system loads a server configuration form template by calling the front-end interactive interface, and receives numerical fields such as performance requirements, budget range, and expected power consumption filled in by the user in the specified input area. It then guides the user to select options such as processor type, storage interface type, and whether GPU acceleration is supported from the standard configuration options through a drop-down selection control. At the same time, it enables text input boxes to collect natural language descriptions of the user's usage scenarios, performance focus, or scalability requirements. During the information collection process, parameter integrity verification logic is introduced to avoid missing key information. Unfilled fields are highlighted and guided to complete. For example, if the user does not fill in the budget field, a prompt window pops up asking for the upper limit of the budget range, thereby ensuring that the input information is usable and complete at both the structural and semantic levels.

[0038] S12: Convert user input information into a structured data format to form server-customized requirements, which can then be called upon in subsequent requirement profile generation steps.

[0039] Specifically, by constructing a parameter parsing and semantic extraction model, the user-input form parameters, drop-down selection values, and natural language text are formatted and standardized. All structured parameters are uniformly converted into numerical or enumerated fields and mapped to the standard field names set in the server parameter definition table. At the same time, the keyword matching and intent classification module is called to extract scene tags such as "AI model training" and "edge computing deployment" from the natural language description part and map them to the scene fields. After processing, all data is encapsulated into JSON format structured objects and assigned a unique request number to identify the customized requirement information, so that the subsequent requirement profile construction logic can carry out feature vector extraction and multi-dimensional parameter modeling based on the unified format data.

[0040] In one embodiment, such as Figure 3 As shown, in step S20, a multi-dimensional requirement profile is constructed based on server customization needs, specifically including: S21: Extract performance metrics, cost budget, power consumption limits, application scenario information, and user preference data from the server customization requirements, and establish a requirement parameter vector containing multiple dimensions, including performance, cost, power consumption, and user preference dimensions.

[0041] Specifically, based on the user-customized requirements data that has been converted into a structured format, performance-related metrics such as processor clock speed, number of cores, memory size, storage type, and interface bandwidth are extracted sequentially. These parameters are then incorporated into feature vector construction items under the performance dimension. Simultaneously, the user-entered cost budget upper limit is numerically normalized and mapped to budget weight parameters. Power supply capacity, target power consumption upper limit, or energy-saving level requirements are organized into power consumption vector elements. For structured labels of business application scenarios, such as "video transcoding" or "database hosting," the default workload type is associated to supplement environmental constraints. User-marked preferences, such as "emphasis on GPU performance" or "no concern for noise control," are mapped to priority weight parameters in the preference dimension. Finally, all parameter dimensions are combined into a multi-dimensional vector structure according to a preset template for subsequent genetic optimization and sorting modules to call.

[0042] S22: Based on a pre-defined NLP model, semantic parsing is performed on the natural language content in the user input information to extract key configuration intents. Combined with a rule engine, each parameter item is standardized to form a unified format requirement profile object for subsequent compatibility checks and candidate configuration scheme generation steps.

[0043] Specifically, the system takes natural language paragraphs, keyword tags, and ambiguous expressions contained in user input as input, calls the loaded semantic parsing model to perform intent recognition, entity extraction, and grammatical dependency judgment on the text, and outputs structured semantic tags as intermediate results. For example, from the user's statement "for AI model training, dual GPUs and more than 128G of memory are required," the scene tag "AI training," the component constraint "dual GPU support," and the performance minimum "128GRAM" are extracted. Then, the rule engine looks up tables to standardize parameter items with different expressions but the same meaning into field names and legal value ranges recognized by the system. For example, ambiguous words such as "high performance" and "energy saving" are converted into parameters with high performance priority and medium power consumption limit. Finally, all parameter combinations, mappings, and constraint information after semantic extraction and rule correction are aggregated into a unified data structure and output as a requirement profile object to ensure that the subsequent compatibility verification and optimization generation stages can call semantically consistent and structurally standard input data for inference and calculation.

[0044] In one embodiment, such as Figure 4 As shown, in step S30, a configuration compatibility check is performed based on the requirement profile, specifically including: S31: Call the hardware component knowledge graph in the preset configuration knowledge base, identify and extract the constraint rules between each component based on the component combination in the requirement profile, and the constraint rules include slot matching, power adaptation, interface standard and space compatibility.

[0045] Specifically, a pre-defined component knowledge graph data structure is loaded, and the component combinations explicitly specified or deduced by default in the requirement profile are used as input entity nodes. Based on the connection edge types defined in the graph, the relationship attributes between different components are extracted, including but not limited to the slot matching relationship between the motherboard and the processor, the adaptation relationship between the power module and the overall power supply capacity, the compatibility relationship between interface types such as PCIe channel version and onboard expansion cards, and the spatial adaptation relationship between standard chassis size and the length of the selected GPU or the height of the air cooler. During the parsing process, the corresponding subgraph is called according to the component type to improve processing efficiency. For example, when identifying the connection relationship between NVMe solid-state drives and motherboards, it is only necessary to determine the number of M.2 slots and the protocol support, thereby efficiently establishing a set of constraint rules for component combinations for subsequent logical reasoning.

[0046] S32: Based on constraint rules, the rule engine performs logical reasoning on the component combination to determine whether there are compatibility issues in the component combination. These compatibility issues include hardware parameter conflicts, structural incompatibility, or functional exclusion.

[0047] Specifically, the constraint rules extracted from the component knowledge graph are used as preconditions for inference and input into the logic rule engine. Various logical judgment conditions are matched in turn, including whether the processor TDP power consumption exceeds the motherboard's support limit, whether the graphics card thickness exceeds the chassis's allowable installation space, whether the power supply has sufficient 12V current output capability to support the operation of all load modules, and whether the number of memory channels meets the multi-channel bandwidth requirements. During inference, combination verification and counterexample matching are used to mark component combinations that do not meet the conditions as having "compatibility issues". For example, if a configuration combination includes a dual-slot RTX card and a MATX structure motherboard, the logical inference result will mark it as structural incompatibility. If a low-wattage power supply is used to match a high-power CPU and multiple hard drives, it will be marked as a power conflict. This ensures that all configuration combinations entering the subsequent optimization process meet the engineering-level hardware and software compatibility requirements.

[0048] S33: When the compatibility check passes, mark the requirement profile as a feasible configuration input for use in the candidate configuration scheme generation step.

[0049] Specifically, if no compatibility violations are detected during the logical reasoning process, the status of the corresponding requirement profile object is updated from "pending verification" to "available input," and a flag field is added to record the compatibility check result as "passed." At the same time, the object is added to the feasible configuration input pool for subsequent multi-objective genetic algorithm calls. During the update process, the pass timestamp and call count are recorded to support subsequent evaluation and tracking mechanisms. For example, if a certain type of GPU configuration scheme passes the compatibility check in multiple user requirements, it can be used to construct a priority reference set of recommended candidate schemes, thereby improving the convergence speed and adaptation rate of the candidate scheme generation stage.

[0050] In one embodiment, such as Figure 5 As shown, in step S40, based on the requirement profile that has passed the compatibility check, a multi-objective genetic algorithm is used to generate candidate configuration schemes, specifically including: S41: Using the requirement profile that has passed the compatibility check as input, extract hardware component combinations from the preset configuration knowledge base to construct an initial configuration scheme population.

[0051] Specifically, the requirement profile, confirmed to be conflict-free through compatibility checks, is used as input. Available hardware components are retrieved from the configuration knowledge base according to the performance level, budget range, and power consumption constraints included in the requirement profile. An initial population of configuration schemes that meets basic engineering feasibility is constructed through two methods: random combination and rule-driven combination. Random combination increases solution space coverage, while rule-driven combination improves the quality of the initial population based on historically highly rated configuration templates. For example, when a high-performance GPU is specified in the user's requirements, a high-power motherboard matching the target GPU's capabilities and a chassis with sufficient airflow channels are prioritized to form individual configurations. This yields a first batch of diverse and feasible candidate schemes, providing a search foundation for subsequent evolution.

[0052] S42: Define a multi-objective fitness function based on the performance metrics, cost budget, and power consumption limits in the requirement profile that has passed the compatibility check, and perform evolutionary operations on the initial configuration population, including selection, crossover, and mutation.

[0053] Specifically, a multi-objective fitness function is constructed based on the performance indicators, cost budget, and power consumption constraints in the requirement profile that has passed the compatibility check. This allows each configuration individual to be quantitatively scored in three directions: performance, cost, and power consumption. The fitness function is used as the evaluation basis for the genetic algorithm. A fitness-based selection operation is performed on the initial configuration scheme population to retain high-quality schemes. Then, crossover operations are performed on core components to achieve superior gene recombination, and probabilistic mutations are performed on some minor parameters to avoid the population getting trapped in local optima. For example, for schemes that meet the budget but have high power consumption, a CPU model with lower TDP or some unnecessary extended storage modules are selected during mutation, thereby continuously improving the overall optimizability of the population and enabling candidate schemes to iteratively evolve towards a balance between high performance, low power consumption, and economic cost.

[0054] S43: When the preset evolution termination condition is met, output multiple optimized candidate configuration schemes for subsequent sorting and filtering steps.

[0055] Specifically, during the iteration of the genetic algorithm, the current optimal fitness change is judged in real time. When the preset evolution termination condition is reached, the evolution process stops. The termination condition may include reaching the maximum number of iterations, the improvement of the population fitness being lower than a threshold, or finding a solution that satisfies strong constraints. The top-ranked configuration individuals in the population at this time are output as optimized candidate configuration schemes, and the main performance contribution parameters and cost composition information of each scheme are recorded for subsequent sorting, screening, and interpretability recommendation. For example, the top few output schemes may focus on computational performance, energy efficiency ratio, or budget optimization, respectively, thereby ensuring that the recommendation stage can provide diverse options for users with different usage preferences.

[0056] In one embodiment, such as Figure 6 As shown, in step S50, the candidate configuration schemes are sorted and filtered based on the user preference data in the requirement profile that has passed the compatibility check, and a list of recommended schemes is obtained. This specifically includes: S51: Construct an initial weighted scoring model based on the user preference weights in the requirement profile that has passed the compatibility check. The preference weights include performance priority, budget sensitivity, and power consumption control tendency.

[0057] Specifically, based on the user preference data contained in the requirement profile that passed the compatibility check, three key weight factors are extracted: performance priority, budget sensitivity, and power consumption control tendency. These factors are then mapped to floating-point weight coefficients. When initializing the scoring model, the basic scoring function for each target dimension is set and corresponding weight values ​​are assigned. During the construction process, default inputs are filled with default median values ​​to prevent the scoring model from becoming unbalanced. For example, when the user does not explicitly set the power consumption tendency, a neutral coefficient of 0.33 is assigned by default. At the same time, adaptive normalization rules are defined for each dimension to ensure that the performance score, budget score, and energy consumption score are additive in the scoring space. Finally, a basic scoring calculation framework is formed for subsequent optimization and integration.

[0058] S52: Optimize the initial weighted scoring model based on historical user feedback data to obtain a multi-attribute weighted scoring model.

[0059] Specifically, by collecting users' click behavior, confirmation behavior, and rejection feedback in historical recommendation interactions, and establishing a correlation mapping relationship between behavioral results and recommendation scheme attributes, the weights of each dimension in the initial weighted scoring model are gradient adjusted. An incremental optimization method with a sample memory window is used to correct the influence coefficient of each dimension score on the final result round by round. For example, when users consistently tend to choose the scheme with slightly higher power consumption but excellent performance in multiple recommendations, the weight ratio of the performance dimension is automatically increased and the penalty coefficient of power consumption control is appropriately reduced, so that the scoring model gradually tends to conform to the actual choice tendency of a specific user group, completing the transformation process from a static scoring structure to a dynamic learnable structure.

[0060] S53: Calculate the scores of candidate configuration schemes on each target dimension using a multi-attribute weighted scoring model, and then weight and fuse the scores of each dimension to obtain a comprehensive score.

[0061] Specifically, each candidate configuration scheme output by the genetic algorithm is input into the scoring model. The performance score is calculated based on the processor performance, memory bandwidth, and I / O throughput metrics in the performance dimension. The economic score is calculated based on the degree of matching between the total cost and the budget in the budget dimension. The energy efficiency score is evaluated based on the total power consumption of the selected components and the power consumption limit set by the user in the power consumption dimension. The scores of the three dimensions are multiplied by the weight coefficients in the current multi-attribute scoring model and the weighted sum is calculated. Finally, a score value representing the scheme under the user's overall expectations is generated. For example, if a scheme has a slightly higher budget but significantly improved performance, it may still rank higher in the weighted result, thus supporting reasonable recommendations after personalized trade-offs.

[0062] S54: Sort all candidate configuration schemes according to the comprehensive score to form a recommended scheme list.

[0063] Specifically, all candidate configuration schemes that have completed the scoring are sorted from high to low according to their comprehensive score. A number of top-ranked schemes are extracted according to a preset quantity threshold to form a recommendation list. During the sorting process, some high-potential alternative schemes can be reserved for cross-validation demonstration or for later user adjustments. Each item in the recommendation scheme list retains its comprehensive score and key component information and generates a summary of the recommendation reasons. For example, when a scheme ranks high but has high power consumption, it can be marked "due to its significantly higher performance score, it has an overall superior score". This supports visualization and enhanced interpretability when it is subsequently transmitted to the display interface, thereby increasing users' trust in the recommended schemes and their willingness to choose them.

[0064] In one embodiment, such as Figure 7 As shown, this intelligent decision-making method for matching server customization needs also includes: S60: Collect user behavior data, including plan click records, preference weight adjustment history, and plan rejection feedback information.

[0065] Specifically, while users browse the list of recommended options, the system continuously records the number of clicks, dwell time, and confirmation actions of each option on the interface. These actions are stored as response markers for candidate options. Simultaneously, the system captures users' modifications to preference parameters such as performance priority, budget sensitivity, and power consumption tendency in the interaction settings and generates time-series records. When a user explicitly rejects an option, the system collects the rejection reason options or the content they fill in and performs text parsing processing. For example, when the user's rejection reason is "budget exceeds the limit," the system can mark the current configuration cost exceeding the limit in the dataset. Finally, all the above behavioral data are aggregated into a continuous behavioral trajectory dataset according to user ID and time window, providing a data foundation with labeled features for subsequent model optimization.

[0066] S70: Use user behavior data as training samples and employ an incremental learning algorithm to dynamically adjust the multi-attribute weighted scoring model.

[0067] Specifically, a training sample set with labeled feedback is constructed based on continuously collected user behavior data. User click preferences and rejection feedback are used as supervisory signals input to the gradient optimization module of the multi-attribute weighted scoring model. An incremental learning algorithm that supports online updates is used to dynamically correct the weight distribution in the current scoring model. A fine-tuning mechanism is triggered after each recommendation iteration to ensure that the model responds promptly to changes in user preferences. During training, a sample memory pool is used to retain short-term historical behavior data to avoid overfitting. At the same time, a feedback confidence weight mechanism is introduced to assign higher adjustment weights to training samples with clear intention feedback. For example, when a user selects the high-performance, high-power solution three times in a row and rejects all energy-saving solutions, the global weight coefficient of the performance dimension can be automatically increased, thereby enabling the scoring model to gradually approach an accurate fit to user preferences and achieve an improvement in behavior-driven recommendation accuracy.

[0068] Furthermore, by analyzing users' adoption behavior, negative feedback, and solution modification paths for specific hardware combinations in historical recommendation results, the system identifies the adaptability bias and logical coverage blind spots of the rules defined in the current knowledge graph. It then constructs a rule fitness evaluation index. When users frequently adopt component combinations that are logically judged as "suboptimal" or "incompatible," the system automatically records their configuration characteristics and scenario context. By comparing local statistical frequency with conflict label results, it determines whether the rule is overly restrictive or lacks conditions. If the confidence interval threshold is met, the graph rule adjustment module is invoked to perform rule relaxation, logical premise updates, or relational edge reconstruction operations. For example, when a certain type of consumer-grade GPU is repeatedly adopted by users in multiple video processing tasks without causing actual compatibility issues, an exception condition is added to the original "professional GPU only" rule: "allowed if power consumption is below 200W and the motherboard supports PCIe 4.0." This enables dynamic evolution and adaptive adjustment of knowledge graph rules, improving the fit to actual usage scenarios and enhancing the recommendation system's ability to accept boundary configuration combinations.

[0069] It should be understood that the sequence number of each step in the above embodiments does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of this application.

[0070] In one embodiment, an intelligent decision-making system for matching server customization requirements is provided, which corresponds one-to-one with the intelligent decision-making method for matching server customization requirements described in the above embodiments. For example... Figure 8 As shown, this intelligent decision-making system for matching customized server requirements includes a requirement acquisition module, a profile construction module, a compatibility check module, a candidate generation module, and a solution ranking module. Detailed descriptions of each functional module are as follows: The requirement acquisition module is used to obtain the server customization requirements input by the user. The server customization requirements include performance indicators, cost budget, power consumption limits, application scenario information, and user preference data. The profile building module is used to build multi-dimensional requirement profiles based on server customization requirements. The requirement profiles include quantitative indicators and configuration constraints. The compatibility check module is used to perform configuration compatibility checks based on the requirement profile to identify whether there are any compatibility issues. The candidate generation module is used to generate candidate configuration schemes based on the requirement profile that has passed the compatibility check, using a multi-objective genetic algorithm. The solution sorting module is used to sort and filter candidate configuration solutions based on user preference data in the requirement profile that has passed the compatibility check, obtain a list of recommended solutions, and transmit the list of recommended solutions to the display interface for users to select and confirm. The recommended solutions include a component list, price, performance indicators, and reasons for recommendation.

[0071] Optionally, the requirements gathering module includes: The information collection submodule is used to collect user input information, including structured parameters filled in by form, configuration options submitted by drop-down selection, and customized intents expressed by natural language input. The structure transformation submodule is used to convert user input information into a structured data format to form server-customized requirements, which can then be called by the subsequent requirement profile generation steps.

[0072] Optionally, the profile building module includes: The parameter extraction submodule is used to extract performance indicators, cost budgets, power consumption limits, application scenario information and user preference data from server customization requirements, and to establish a requirement parameter vector containing multiple dimensions, including performance, cost, power consumption and user preference dimensions. The semantic modeling submodule is used to perform semantic parsing on the natural language content in user input information based on a preset NLP model, extract key configuration intents, and combine the rule engine to standardize each parameter item to form a unified format requirement profile object for subsequent compatibility checks and candidate configuration scheme generation steps.

[0073] Optional, the compatibility check module includes: The constraint extraction submodule is used to call the hardware component knowledge graph in the preset configuration knowledge base, identify and extract the constraint rules between each component based on the component combination in the requirement profile, and the constraint rules include slot matching, power adaptation, interface standard, and space compatibility. The logical reasoning submodule is used to perform logical reasoning on the component combination based on constraint rules and the rule engine to determine whether there are compatibility issues in the component combination. These compatibility issues include hardware parameter conflicts, structural incompatibility, or functional exclusion. The Feasibility Marking submodule is used to mark the requirement profile as a feasible configuration input when the compatibility check passes, so that it can be called by the candidate configuration scheme generation step.

[0074] Optionally, the candidate generation module includes: The population building submodule is used to extract hardware component combinations from a preset configuration knowledge base to build an initial configuration scheme population, taking the requirement profile that has passed the compatibility check as input. The fitness calculation submodule is used to define a multi-objective fitness function based on the performance indicators, cost budget and power consumption limits in the requirement profile that has passed the compatibility check, and to perform evolutionary operations on the initial configuration scheme population. The evolutionary operations include selection, crossover and mutation. The scheme output submodule is used to output multiple optimized candidate configuration schemes when the preset evolution termination conditions are met, for subsequent sorting and filtering steps to call.

[0075] Optionally, the scheme sorting module includes: The model building submodule is used to build an initial weighted scoring model based on the user preference weights in the requirement profile that has passed the compatibility check. The preference weights include performance priority, budget sensitivity and power consumption control tendency. The model optimization submodule is used to optimize the initial weighted scoring model based on historical user feedback data to obtain a multi-attribute weighted scoring model. The scoring calculation submodule is used to calculate the scores of candidate configuration schemes on each target dimension using a multi-attribute weighted scoring model, and to weight and fuse the scores of each dimension to obtain a comprehensive score. The scheme ranking submodule is used to rank all candidate configuration schemes based on the comprehensive score and form a list of recommended schemes.

[0076] Optionally, this intelligent decision-making system for matching server customization needs also includes: The behavior collection module is used to collect user behavior data, including solution click records, preference weight adjustment history, and solution rejection feedback information. The model update module is used to dynamically adjust the multi-attribute weighted scoring model by using user behavior data as training samples and employing an incremental learning algorithm.

[0077] For specific limitations regarding the intelligent decision-making system for matching server customization needs, please refer to the limitations of the intelligent decision-making method for matching server customization needs described above, which will not be repeated here. Each module in the aforementioned intelligent decision-making system for matching server customization needs can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in or independent of the processor in a computer device in hardware form, or stored in the memory of a computer device in software form, so that the processor can call and execute the operations corresponding to each module.

[0078] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the above-described division of functional units and modules is used as an example. In practical applications, the above functions can be assigned to different functional units and modules as needed, that is, the internal structure of the system can be divided into different functional units or modules to complete all or part of the functions described above.

[0079] The above-described embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application, and should all be included within the protection scope of this application.

Claims

1. A method for intelligent decision making for server customization requirement matching, characterized in that, The server customization demand matching intelligent decision method comprises the following steps: Obtain the server customization demand input by a user, wherein the server customization demand comprises a performance index, a cost budget, a power consumption limit, application scenario information, and user preference data; Construct a multi-dimensional demand portrait based on the server customization demand, wherein the demand portrait comprises a quantitative index and a configuration constraint condition; Perform configuration compatibility checking according to the demand portrait, and identify whether there is a compatibility problem; When the compatibility checking is passed, generate a candidate configuration scheme based on the demand portrait passed by the compatibility checking by using a multi-objective genetic algorithm; Sort and filter the candidate configuration scheme based on the user preference data in the demand portrait passed by the compatibility checking, obtain a recommended scheme list, and transmit the recommended scheme list to a display interface for user selection and confirmation, wherein the recommended scheme comprises a component list, a price, a performance index, and a recommended reason.

2. The method of claim 1, wherein, The server customization demand input by a user comprises the following steps: Collect user input information, wherein the input information comprises structured parameters filled in a form, configuration options submitted in a drop-down selection manner, and customization intentions expressed in a natural language input manner; Convert the user input information into a structured data format to form the server customization demand for subsequent demand portrait generation steps. 3.The intelligent decision method for server customization demand matching of claim 2, wherein, The multi-dimensional demand portrait based on the server customization demand comprises the following steps: Extract the performance index, the cost budget, the power consumption limit, the application scenario information, and the user preference data in the server customization demand to establish a demand parameter vector comprising multiple dimensions, wherein the dimensions comprise a performance dimension, a cost dimension, a power consumption dimension, and a user preference dimension; Based on a preset NLP model, perform semantic analysis on the natural language content in the user input information to extract a key configuration intention, and combine a rule engine to standardize each parameter item to form a demand portrait object in a unified format for subsequent compatibility checking and candidate configuration scheme generation steps.

4. The method of claim 1, wherein the server customization requirement matching intelligent decision method is characterized in that, The configuration compatibility checking based on the demand portrait comprises the following steps: Call a hardware component knowledge graph in a preset configuration knowledge base, identify and extract constraint rules between components based on the component combination in the demand portrait, wherein the constraint rules comprise slot matching, power supply adaptation, interface standard, and space compatibility; Based on the constraint rules, perform logical reasoning on the component combination by using a rule engine to determine whether the component combination has a compatibility problem, wherein the compatibility problem comprises a hardware parameter conflict, structural incompatibility, or functional exclusion; When the compatibility checking is passed, mark the demand portrait as a feasible configuration input for the candidate configuration scheme generation step.

5. The method of claim 1, wherein the server customization requirement matching intelligent decision method is characterized in that, The candidate configuration scheme generated based on the demand portrait passed by the compatibility checking by using a multi-objective genetic algorithm comprises the following steps: Take the demand portrait passed by the compatibility checking as input, extract a hardware component combination from a preset configuration knowledge base to construct an initial configuration scheme population; According to the performance indicators, cost budget and power consumption limit in the requirement profile passed by the compatibility check, a multi-objective fitness function is defined, and an evolution operation is performed on the initial configuration scheme population, the evolution operation including selection, crossover and mutation; When the preset evolution termination condition is met, a plurality of optimized candidate configuration schemes are output for subsequent sorting and screening steps.

6. The method of claim 1, wherein the server customization requirement matching intelligent decision method is characterized in that, The user preference data in the requirement profile passed by the compatibility check is used to sort and screen the candidate configuration schemes to obtain a recommended scheme list, specifically including: An initial weighted scoring model is constructed according to the user preference weight in the requirement profile passed by the compatibility check, the preference weight including performance priority, budget sensitivity and power consumption control tendency; Based on the user historical feedback data, the initial weighted scoring model is optimized to obtain a multi-attribute weighted scoring model; The multi-attribute weighted scoring model is used to calculate the scores of the candidate configuration schemes in each target dimension respectively, and the dimension scores are weighted and fused to obtain a comprehensive score; All candidate configuration schemes are sorted according to the comprehensive score to form a recommended scheme list.

7. The method of claim 6, wherein the server customization requirement matching intelligent decision method is characterized in that, The intelligent decision-making method for server customization requirement matching further includes: User behavior data is collected, including scheme click record, preference weight adjustment history and scheme veto feedback information; The user behavior data is used as a training sample, and an incremental learning algorithm is used to dynamically adjust the multi-attribute weighted scoring model.

8. An intelligent decision system for server customization requirement matching, characterized in that, The intelligent decision-making system for server customization requirement matching includes: A requirement collection module is configured to obtain a server customization requirement input by a user, the server customization requirement including performance indicators, cost budget, power consumption limit, application scenario information and user preference data; An image construction module is configured to construct a multi-dimensional requirement profile based on the server customization requirement, the requirement profile including quantitative indicators and configuration constraint conditions; A compatibility check module is configured to perform a configuration compatibility check based on the requirement profile to identify whether there is a compatibility problem; A candidate generation module is configured to generate candidate configuration schemes based on the requirement profile passed by the compatibility check using a multi-objective genetic algorithm when the compatibility check is passed; A scheme sorting module is configured to sort and screen the candidate configuration schemes based on the user preference data in the requirement profile passed by the compatibility check to obtain a recommended scheme list, and transmit the recommended scheme list to a display interface for user selection and confirmation, the recommended scheme including a component list, price, performance indicators and recommended reasons.

Images