Electronic component automatic recommendation method and system based on multi-dimensional parameter weight

Abstract

The invention relates to the technical field of electronic component intelligent type selection and industrial internet recommendation, and discloses an electronic component automatic recommendation method and system based on multi-dimensional parameter weight, and the method comprises the following steps: 1, constructing an electronic component multi-dimensional parameter database, the database comprises inherent attribute parameters, performance parameters, application scene parameters and user demand associated parameters of the electronic components. According to the invention, the parameter preprocessing module ensures the validity and comparability of parameter data through data cleaning, improved Z-score abnormal value processing and min-max standardization, and lays a foundation for subsequent calculation; a weight calculation module fuses an AHP subjective weight and an entropy weight method objective weight, the AHP introduces an expert weight adjustment mechanism, the entropy weight method reflects objective dispersion based on parameter information entropy, the AHP subjective weight and the entropy weight method reflect objective dispersion based on parameter information entropy through linear fusion to obtain a combined weight, a core parameter weight is more scientific, and compared with a traditional single weight method, the recommendation accuracy is improved from 60% to 90% or above, and the error selection rate is reduced to 3% or below.

Description

Technical Field

[0001] This invention relates to the field of intelligent selection of electronic components and industrial internet recommendation technology, specifically to an automatic recommendation method and system for electronic components based on multi-dimensional parameter weights. Background Technology

[0002] Traditional recommendation methods for selecting and procuring electronic components suffer from four major pain points, severely limiting selection efficiency, compatibility, and practicality, and failing to meet the needs of modern electronic R&D and manufacturing: The parameters are too narrow and the matching is too one-sided: Traditional recommendations focus on the inherent attributes of components (such as model, package and size) and ignore key performance parameters (such as rated voltage, power loss, frequency response and temperature rating), application scenario parameters (such as electromagnetic interference resistance and system compatibility) and user demand-related parameters (such as budget constraints and delivery cycle). This leads to the recommendation results being out of touch with the actual R&D / procurement needs. For example, recommending ordinary room temperature capacitors for high temperature industrial scenarios has an adaptability of less than 50%.

[0003] The weighting is subjective and inaccurate: weight calculations rely heavily on human experience (such as purchasing personnel preferences) or a single subjective method (such as pure analytic hierarchy process, AHP), without incorporating objective data (such as parameter information entropy, sample distribution), resulting in large weight biases; moreover, there is no expert weight adjustment mechanism, and the evaluation weights of experts from different fields are equal, which dilutes the weights of core parameters (such as chip computing power and power consumption), and the recommendation accuracy is often below 60%, with a wrong selection rate exceeding 15%.

[0004] Lack of dynamic optimization and poor practicality: Recommendations are based solely on parameter matching, without considering users' historical behavior (such as R&D personnel often prioritizing high-reliability models and procurement personnel focusing on inventory) and real-time market supply (such as some recommended models being out of stock or delivery cycles exceeding user requirements). This results in "reliability without accuracy"—even with high similarity, procurement cannot be implemented due to stock shortages or delivery delays, leading to a recommendation implementation rate of less than 50%. Furthermore, there are no alternative models associated with the recommendations, causing selection interruptions when faced with discontinued models, requiring a restart of the recommendation process, which is inefficient.

[0005] Poor data timeliness and delayed updates: Component parameters (such as performance upgrades, price changes, and certification updates) rely on manual input, with an update cycle of over 30 days, and some discontinued models are not marked in a timely manner; there is no parameter version control, and it is impossible to trace historical parameters (such as the need to match the original parameter model for old equipment repair), which leads to a disconnect between recommendations and actually available components, and extends the R&D selection cycle by more than 30%. Summary of the Invention

[0006] The purpose of this invention is to provide an automatic recommendation method and system for electronic components based on multi-dimensional parameter weights, so as to solve the problems mentioned in the background art.

[0007] To achieve the above objectives, the present invention provides the following technical solution: an automatic recommendation method for electronic components based on multi-dimensional parameter weights, comprising the following steps: Step 1: Construct a multi-dimensional parameter database for electronic components. This database includes the inherent attribute parameters, performance parameters, application scenario parameters, and user requirement-related parameters of electronic components. Step 2: Preprocess the multi-dimensional parameters, including data cleaning, outlier handling, and standardization transformation. The data cleaning step removes parameter records from the database where the percentage of missing values ​​exceeds a preset threshold, which is 20% of the total number of parameter items in the record. The outlier handling step uses an improved Z-score method; for each parameter item, its Z-score value is calculated. ,in For parameter values, This is the mean of the parameter. The standard deviation of this parameter term, when When the absolute value of a parameter is greater than 3, it is considered an outlier, and the median of that parameter is used to replace the outlier. The standardization transformation step uses the min-max standardization method to transform all parameter values ​​to the [0,1] interval. The transformation formula is as follows: ,in These are the standardized parameter values. This is the minimum value of this parameter. This is the maximum value of the parameter. Step 3: Determine the weight values ​​for each dimension parameter. This determination employs a combined weighting method that integrates subjective and objective weights. The subjective weights are determined using the Analytic Hierarchy Process (AHP), specifically by constructing a hierarchical model that divides the multi-dimensional parameters into a target layer, a criterion layer, and an indicator layer. The target layer represents the recommended compatibility of electronic components; the criterion layer comprises four categories: inherent attributes, performance, application scenarios, and user needs; and the indicator layer consists of each specific parameter item. A judgment matrix is ​​constructed by inviting at least five experts in the field of electronic components to conduct pairwise comparisons of the parameter items at the same level. ,in Indicates the first The parameter item is relative to the first The degree of importance of each parameter item The value of is an integer in the 1-9 scale and satisfies , =1, calculate the largest eigenvalue of the judgment matrix. The subjective weight vector is obtained by normalizing the feature vectors and their corresponding eigenvectors, and its expression is: ; in This represents the weight vector of sub-level factors. For the first The importance weight of each sub-level factor relative to its parent-level goal. The number of factors in each sub-level is determined, and a consistency test is performed. The consistency index is... When the consistency ratio When <0.1, the judgment matrix has satisfactory consistency, where The average random consistency index; Based on matrix order Select the corresponding values; the objective weights are determined using the entropy weight method, specifically including calculating the... Information entropy of each parameter item ,in The number of component samples. For the first The sample at the th The proportion of the standardized value of each parameter to the sum of all standardized values ​​of that parameter, i.e. ,when When =0, define Then calculate the objective weight. To obtain the objective weight vector Combined weight vector It is obtained by linearly fusing subjective weights and objective weights, i.e. ; Step 4: Obtain the target requirement parameters input by the user. These target requirement parameters include the performance index range specified by the user, application scenario description, budget constraints, and delivery time requirements. Then, convert the target requirement parameters into a standardized requirement vector corresponding to the parameter items in the database. Each of them For users to the first A standardized representation of the required value or range of each parameter item; Step 5: Calculate the similarity between each electronic component sample in the database and the target demand vector. The similarity calculation uses the weighted cosine similarity algorithm, and the formula is as follows: ,in For the first The standardized values ​​of each parameter in the component sample. For the first The combined weights of the parameter items; Step 6: Sort all electronic component samples according to the similarity calculation results, and select the top N components with the highest similarity scores as the initial recommendation list, where N is the number of recommendations that can be set by the user, with a default value of 10. Step seven optimizes and adjusts the initial recommendation list. This optimization process includes introducing a user historical behavior feedback factor (β). This factor is calculated based on user clicks, favorites, purchases, and reviews of historical recommendation results. ,in This is the normalized value of the click frequency. This is a normalized value for the number of collections. To improve conversion rates, To evaluate the score normalization values, all normalization values ​​were transformed to the [0,1] interval, and the similarity score of each component in the initial recommendation list was calculated. With the corresponding feedback factor Multiply by each product to obtain the final fit score. The system then re-sorts the components based on their final fit scores, generating and outputting the final recommended list of electronic components.

[0008] Preferably, the construction of the multi-dimensional parameter database also includes a dynamic parameter update mechanism. This mechanism automatically crawls the latest parameter information from the official websites of electronic component manufacturers, industry databases, and e-commerce platforms according to a preset time period of 7 days. For newly added component models, their multi-dimensional parameters are automatically extracted, standardized, and then entered into the database. For existing component models, when a change in parameter information is detected, such as performance parameter adjustments or price changes, the corresponding record in the database is automatically updated, and the parameter change time and content are marked. Simultaneously, a version control mechanism is established for each parameter item, saving at least the most recent three parameter change histories so that different time-point parameter versions can be selected for matching calculations based on user needs during the recommendation process. Furthermore, the database also stores information on alternative models for each component, including direct replacement models, functionally compatible replacement models, and performance-upgraded replacement models, and records the parameter difference between the replacement model and the original model. The parameter difference is obtained by calculating the weighted Euclidean distance between the two models on core parameter items, including rated voltage, rated current, power, and size. The weighted Euclidean distance formula is... ,in The weights of the core parameter items, The first of the original models One core parameter value, The replacement model number The core parameter values.

[0009] Preferably, the construction of the judgment matrix in the analytic hierarchy process also incorporates an expert weight adjustment mechanism, assigning expert weights to each expert based on their years of industry experience, past parameter evaluation accuracy, and professional field matching degree. Among them, experts with more than 10 years of experience in the industry The value is 1.0 for periods of 5-10 years, 0.8 for periods of less than 5 years, and 0.6 for periods of less than 5 years. The accuracy of past parameter assessments is calculated based on the consistency between historical expert assessments and feedback from actual applications, with an accuracy exceeding 90%. An extra 0.2 points are awarded for scores between 70% and 90%, and an extra 0.1 points are awarded for scores below 70%. The professional field matching degree is determined by the overlap between the expert's research field and the field of the currently recommended component. A perfect match is awarded... An extra 0.3 is awarded for a partial match, an extra 0.15 for a partial match, and no extra points are awarded for a non-match. Expert weighting applies. The total value range is [0.6, 1.5], and for all experts... Normalization is performed to make ,in , For the first The original weights of the experts, For the first The normalized weights of the experts are used to determine the final judgment matrix, which is obtained by weighted averaging of the judgment matrices of all experts. ,in For the first A judgment matrix constructed by experts is used to improve the objectivity and reliability of subjective weights.

[0010] Preferably, in the similarity calculation step, when the user sets hard constraints on certain parameters, such as requiring a minimum or maximum value for a performance parameter, a two-step calculation method is used: The first step is to screen all component samples and remove those that do not meet the hard constraints. Meeting the hard constraints is determined by the lower limit of the parameters specified by the user. The corresponding parameter value y in the component sample is ≥ For user-specified parameter upper limit The corresponding parameter value y in the component sample is ≤ For user-specified mandatory parameter attributes, such as specific certification standards, the component sample must include that attribute; The second step involves calculating the weighted cosine similarity of the selected samples, while simultaneously introducing a dynamic adjustment factor for parameter importance into the similarity calculation. When a user explicitly emphasizes the importance of a certain parameter, such as marking it as "important" or "very important" when inputting their requirements, the combined weight of that parameter will be adjusted. Multiply by the corresponding The word "important" corresponds to =1.5, corresponding to "very important" =2.0, unmarked parameter item =1.0, the adjusted weight is The adjusted weights are then normalized to ensure that... This is to ensure the accuracy of similarity calculations.

[0011] Preferably, the optimization and adjustment of the recommended list also includes introducing a real-time market supply information factor. This factor is calculated based on the current inventory quantity of each component, supplier delivery cycle, and price fluctuations. Specifically, ,in To ensure sufficient inventory, when the inventory quantity is more than 5 times the amount the user may purchase, =1.0, 2-5 times =0.7, 1-2 times =0.4, less than 1 time =0.1, For delivery cycle satisfaction, when the delivery cycle is less than 80% of the user's required time, =1.0, t=0.8 when 80%-100%, t=1.0 when 100%-120% =0.5, when it is greater than 120%, t=0.2. For price stability, when the price fluctuation range is less than 5% in the past 30 days =1.0, 5%-10% =0.7, greater than 10% =0.4, all factor values ​​were converted to the [0,1] interval, and the final fit score was adjusted to The recommended list also displays market supply information for each component. This value is provided for users to refer to comprehensively.

[0012] Another technical problem to be solved by the present invention is to provide a system for the automatic recommendation method of electronic components based on multi-dimensional parameter weights as described above, comprising: The multi-dimensional parameter database module is used to store the inherent attribute parameters, performance parameters, application scenario parameters, and user requirement-related parameters of electronic components. This database module also has a parameter dynamic update function, which can crawl and update the latest parameter information according to a preset cycle, and establish a parameter version control mechanism and alternative model association. The parameter preprocessing module is connected to the multi-dimensional parameter database module and is used to perform data cleaning, outlier handling and standardization transformation on the multi-dimensional parameters in the database. Data cleaning is used to remove parameter records with missing values ​​exceeding 20%, and outlier handling adopts the improved Z-score method. The weight calculation module, connected to the parameter preprocessing module, is used to determine the combined weights of parameters in each dimension. This module includes a subjective weight calculation submodule and an objective weight calculation submodule. The subjective weight calculation submodule uses the Analytic Hierarchy Process (AHP) to obtain the subjective weight vector by constructing a hierarchical structure model, expert judgment matrix, consistency check and eigenvector calculation. The objective weight calculation submodule uses the entropy weight method to obtain the objective weight vector by calculating information entropy and objective weights. The combined weight vector is obtained by linear fusion. The requirement acquisition module is used to receive target requirement parameters input by users, including performance index range, application scenario description, budget constraints and delivery time requirements, and convert them into standardized requirement vectors. The similarity calculation module is connected to the parameter preprocessing module, the weight calculation module, and the demand acquisition module, respectively, and is used to calculate the similarity between the electronic component samples and the target demand vector. The recommendation list generation module is connected to the similarity calculation module. It is used to generate an initial recommendation list based on the similarity score, select the top N components, and optimize the initial list by introducing user historical behavior feedback factors and real-time market supply information factors. The final recommendation list is generated by sorting the components according to the score. The output module, connected to the recommendation list generation module, is used to display the final recommendation list to the user in a visual manner, including parameter details, similarity scores, feedback factors, and market supply information for each component.

[0013] Preferably, the subjective weight calculation submodule in the weight calculation module further includes an expert weight adjustment unit.

[0014] Preferably, the similarity calculation module further includes a hard constraint screening unit and a parameter importance adjustment unit.

[0015] This invention provides a method and system for automatically recommending electronic components based on multi-dimensional parameter weights. It has the following beneficial effects: 1. The multi-dimensional parameter database module of this invention not only stores inherent attributes, but also incorporates performance parameters, application scenario parameters, and user demand-related parameters. It ensures the timeliness of parameters through dynamic crawling and updating every 7 days. At the same time, it establishes parameter version control and association with alternative models to avoid recommendation failure due to missing parameters or model discontinuation.

[0016] 2. The parameter preprocessing module of this invention ensures the validity and comparability of parameter data through data cleaning, improved Z-score outlier handling, and min-max standardization, laying the foundation for subsequent calculations. The weight calculation module integrates AHP subjective weights and entropy weight method objective weights. AHP introduces an expert weight adjustment mechanism, while entropy weight method reflects objective dispersion based on parameter information entropy. The two are linearly fused to obtain combined weights, making the core parameter weights more scientific. Compared with traditional single weight methods, the recommendation accuracy is increased from 60% to over 90%, and the misselection rate is reduced to below 3%.

[0017] 3. The similarity calculation module of this invention supports hard constraint screening and dynamic adjustment of parameter importance to further ensure matching accuracy. For example, when recommending components with a temperature resistance of -55℃ to 125℃ for aerospace scenarios, it can accurately eliminate room temperature models.

[0018] 4. This invention utilizes a dynamic optimization mechanism for the recommendation list, ensuring that the recommendations are not only accurate but also better suited to actual procurement and selection needs, significantly improving the implementation rate. User behavior feedback aligns with preferences: The recommendation list generation module incorporates historical user behavior feedback factors. For example, if R&D personnel frequently collect high-reliability components, the system automatically increases the final fit score for such models, making recommendations more aligned with user habits and improving repeated selection efficiency by 50%. Furthermore, real-time market supply ensures implementation: Introducing a real-time market supply factor, the final score is adjusted to Score = Sim(1+β)(1+θ), and the θ value is displayed in the recommendation list, allowing users to intuitively assess procurement feasibility—avoiding recommendations of "highly similar but out-of-stock" models, increasing the implementation rate from 50% to over 85%. In addition, the database stores alternative models. When a recommended model is out of stock or discontinued, the system automatically recommends alternative solutions with low parameter differences, reducing the selection interruption rate from 30% to below 5%, significantly shortening the R&D and procurement cycle. Attached Figure Description

[0019] Figure 1 This is an overall flowchart of the automatic recommendation method for electronic components based on multi-dimensional parameter weights in this invention; Figure 2 This is a flowchart of the multi-dimensional parameter preprocessing process of the present invention; Figure 3 This is a flowchart of the multi-dimensional parameter combination weight calculation process of the present invention; Figure 4 This is a flowchart illustrating the interaction of the electronic component automatic recommendation system module based on multi-dimensional parameter weights according to the present invention. Detailed Implementation

[0020] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0021] Examples of the embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and intended to explain the invention, and should not be construed as limiting the invention.

[0022] Example 1 A preferred embodiment of the automatic electronic component recommendation method and system based on multi-dimensional parameter weights provided by this invention is as follows: Figure 1-4 The following is an example of an automatic recommendation method for electronic components based on multi-dimensional parameter weights, comprising the following steps: Step 1: Construct a multi-dimensional parameter database for electronic components. This database includes inherent attribute parameters, performance parameters, application scenario parameters, and user requirement-related parameters for electronic components. Inherent attribute parameters include component model specifications, material, packaging form, manufacturer, production date, and certification standards. Performance parameters include electrical performance parameters, mechanical performance parameters, environmental adaptability performance parameters, and reliability parameters. Electrical performance parameters further include rated voltage, rated current, resistance value, capacitance value, inductance value, power loss, frequency response range, and noise figure. Mechanical performance parameters include size, weight, vibration resistance level, and impact resistance. Environmental adaptability performance parameters include operating temperature range, humidity adaptability range, electromagnetic interference resistance, and corrosion resistance level. Reliability parameters include mean time between failures (MTBF), expected lifespan, and failure rate. Application scenario parameters include applicable industry sectors, typical application scenarios, supporting equipment models, and system compatibility requirements. User requirement-related parameters include historical purchase records, user evaluation scores, price sensitivity coefficient, and delivery cycle requirements. Step 2: Preprocess the multi-dimensional parameters, including data cleaning, outlier handling, and standardization transformation. The data cleaning step removes parameter records from the database where the percentage of missing values ​​exceeds a preset threshold (20% of the total number of parameter items in the record). The outlier handling step uses a modified Z-score method; for each parameter item, its Z-score value is calculated. ,in For parameter values, This is the mean of the parameter. The standard deviation of this parameter term, when When the absolute value of a parameter is greater than 3, it is considered an outlier, and the median of that parameter is used to replace the outlier. The standardization transformation step uses the min-max standardization method to transform all parameter values ​​to the [0,1] interval. The transformation formula is as follows: ,in These are the standardized parameter values. This is the minimum value of this parameter. This is the maximum value of the parameter. Step 3: Determine the weight values ​​for each dimension parameter. This determination employs a combined weighting method that integrates subjective and objective weights. The subjective weights are determined using the Analytic Hierarchy Process (AHP), specifically by constructing a hierarchical model that divides the multi-dimensional parameters into a target layer, a criterion layer, and an indicator layer. The target layer represents the recommended compatibility of electronic components; the criterion layer comprises four categories: inherent attributes, performance, application scenarios, and user needs; and the indicator layer consists of each specific parameter item. A judgment matrix is ​​constructed by inviting at least five experts in the field of electronic components to conduct pairwise comparisons of the parameter items at the same level. ,in Indicates the first The parameter item is relative to the first The degree of importance of each parameter item The value of is an integer in the 1-9 scale and satisfies , =1, calculate the largest eigenvalue of the judgment matrix. The subjective weight vector is obtained by normalizing the feature vectors and their corresponding eigenvectors, and its expression is: ; in This represents the weight vector of sub-level factors. For the first The importance weight of each sub-level factor relative to its parent-level goal. The number of factors in each sub-level is determined, and a consistency test is performed. The consistency index is... When the consistency ratio When <0.1, the judgment matrix has satisfactory consistency, where The average random consistency index; Based on matrix order Select the corresponding values; the objective weights are determined using the entropy weight method, specifically including calculating the... Information entropy of each parameter item ,in The number of component samples. For the first The sample at the th The proportion of the standardized value of each parameter to the sum of all standardized values ​​of that parameter, i.e. ,when When =0, define Then calculate the objective weight. To obtain the objective weight vector Combined weight vector It is obtained by linearly fusing subjective weights and objective weights, i.e. ,in These are the weighting coefficients. The value range is [0,1]. It is determined by fitting and optimizing historical recommendation data using the least squares method, so that the recommendation result under the combined weight has the highest matching degree with the actual user selection result. Step 4: Obtain the target requirement parameters input by the user. These target requirement parameters include the performance index range specified by the user, application scenario description, budget constraints, and delivery time requirements. Then, convert the target requirement parameters into a standardized requirement vector corresponding to the parameter items in the database. Each of them For users to the first A standardized representation of the required value or range of each parameter item; Step 5: Calculate the similarity between each electronic component sample in the database and the target demand vector. The similarity calculation uses the weighted cosine similarity algorithm, and the formula is as follows: ,in For the first The standardized values ​​of each parameter in the component sample. For the first The combined weights of the parameter items; Step 6: Sort all electronic component samples according to the similarity calculation results, and select the top N components with the highest similarity scores as the initial recommendation list, where N is the number of recommendations that can be set by the user, with a default value of 10. Step seven involves optimizing and adjusting the initial recommendation list. This optimization process includes introducing a user history behavior feedback factor (β). This factor is calculated based on user clicks, favorites, purchases, and reviews of historical recommendations. ,in This is the normalized value of the click frequency. This is a normalized value for the number of collections. To improve conversion rates, To evaluate the score normalization values, all normalization values ​​were transformed to the [0,1] interval, and the similarity score of each component in the initial recommendation list was calculated. With the corresponding feedback factor Multiply by each product to obtain the final fit score. The system then re-sorts the components based on their final fit scores, generating and outputting the final recommended list of electronic components.

[0023] The construction of the multi-dimensional parameter database also includes a dynamic parameter update mechanism. This mechanism automatically crawls the latest parameter information from the official websites of electronic component manufacturers, industry databases, and e-commerce platforms according to a preset time period of 7 days. For newly added component models, its multi-dimensional parameters are automatically extracted, standardized, and then entered into the database. For existing component models, when a change in parameter information is detected, such as performance parameter adjustments or price changes, the corresponding record in the database is automatically updated, and the parameter change time and content are marked. Simultaneously, a version control mechanism is established for each parameter item, saving at least the most recent three parameter change histories. This allows for matching calculations of parameter versions at different time points based on user needs during the recommendation process. Furthermore, the database also stores information on alternative models for each component, including direct replacement models, functionally compatible replacement models, and performance-upgraded replacement models. It records the parameter differences between the replacement and original models. The parameter difference is obtained by calculating the weighted Euclidean distance between the two models on core parameter items, including rated voltage, rated current, power, and size. The weighted Euclidean distance formula is... ,in The weights of the core parameter items, The first of the original models One core parameter value, The replacement model number The core parameter values.

[0024] The analytic hierarchy process (AHP) incorporates an expert weight adjustment mechanism in the construction of the judgment matrix. This mechanism assigns expert weights to each expert based on their years of industry experience, past parameter evaluation accuracy, and professional domain relevance. Among them, experts with more than 10 years of experience in the industry The value is 1.0 for periods of 5-10 years, 0.8 for periods of less than 5 years, and 0.6 for periods of less than 5 years. The accuracy of past parameter assessments is calculated based on the consistency between historical expert assessments and feedback from actual applications, with an accuracy exceeding 90%. An extra 0.2 points are awarded for scores between 70% and 90%, and an extra 0.1 points are awarded for scores below 70%. The professional field matching degree is determined by the overlap between the expert's research field and the field of the currently recommended component. A perfect match is awarded... An extra 0.3 is awarded for a partial match, an extra 0.15 for a partial match, and no extra points are awarded for a non-match. Expert weighting applies. The total value range is [0.6, 1.5], and for all experts... Normalization is performed to make ,in , For the first The original weights of the experts, For the first The normalized weights of the experts are used to determine the final judgment matrix, which is obtained by weighted averaging of the judgment matrices of all experts. ,in For the first A judgment matrix constructed by experts is used to improve the objectivity and reliability of subjective weights.

[0025] In the similarity calculation process, when the user sets hard constraints on certain parameters, such as requiring a minimum or maximum value for a performance parameter, a two-step calculation method is used: The first step is to screen all component samples and remove those that do not meet the hard constraints. Meeting the hard constraints is determined by the lower limit of the parameters specified by the user. The corresponding parameter value y in the component sample is ≥ For user-specified parameter upper limit The corresponding parameter value y in the component sample is ≤ For user-specified mandatory parameter attributes, such as specific certification standards, the component sample must include that attribute; The second step involves calculating the weighted cosine similarity of the selected samples, while simultaneously introducing a dynamic adjustment factor for parameter importance into the similarity calculation. When a user explicitly emphasizes the importance of a certain parameter, such as marking it as "important" or "very important" when inputting their requirements, the combined weight of that parameter will be adjusted. Multiply by the corresponding The word "important" corresponds to =1.5, corresponding to "very important" =2.0, unmarked parameter item =1.0, the adjusted weight is The adjusted weights are then normalized to ensure that... This is to ensure the accuracy of similarity calculations.

[0026] The optimization and adjustments to the recommended list also include the introduction of a real-time market supply information factor. This factor is calculated based on the current inventory quantity of each component, supplier delivery cycle, and price fluctuations. ,in To ensure sufficient inventory, when the inventory quantity is more than 5 times the amount the user may purchase, =1.0, 2-5 times =0.7, 1-2 times =0.4, less than 1 time =0.1, For delivery cycle satisfaction, when the delivery cycle is less than 80% of the user's required time, =1.0, t=0.8 when 80%-100%, t=1.0 when 100%-120% =0.5, when it is greater than 120%, t=0.2. For price stability, when the price fluctuation range is less than 5% in the past 30 days =1.0, 5%-10% =0.7, greater than 10% =0.4, all factor values ​​were converted to the [0,1] interval, and the final fit score was adjusted to The recommended list also displays market supply information for each component. This value is provided for users to refer to comprehensively.

[0027] Example 2 Please see Figures 1-4 Furthermore, based on Example 1, a system for automatically recommending electronic components using the above-mentioned multi-dimensional parameter weighting method is obtained, comprising: The multi-dimensional parameter database module is used to store the inherent attribute parameters, performance parameters, application scenario parameters, and user requirement-related parameters of electronic components. This database module also has a parameter dynamic update function, which can crawl and update the latest parameter information according to a preset cycle, and establish a parameter version control mechanism and alternative model association. The parameter preprocessing module connects to the multi-dimensional parameter database module and is used to perform data cleaning, outlier handling, and standardization transformation on the multi-dimensional parameters in the database. Data cleaning is used to remove parameter records with more than 20% missing values, and outlier handling adopts an improved Z-score method. The weight calculation module, connected to the parameter preprocessing module, is used to determine the combined weights of parameters in each dimension. This module includes a subjective weight calculation submodule and an objective weight calculation submodule. The subjective weight calculation submodule uses the Analytic Hierarchy Process (AHP) to obtain the subjective weight vector by constructing a hierarchical structure model, expert judgment matrix, consistency check and eigenvector calculation. The objective weight calculation submodule uses the entropy weight method to obtain the objective weight vector by calculating information entropy and objective weights. The combined weight vector is obtained by linear fusion. The requirement acquisition module is used to receive target requirement parameters input by users, including performance index range, application scenario description, budget constraints and delivery time requirements, and convert them into standardized requirement vectors. The similarity calculation module is connected to the parameter preprocessing module, the weight calculation module, and the demand acquisition module, respectively, and is used to calculate the similarity between electronic component samples and target demand vectors. The recommendation list generation module is connected to the similarity calculation module. It is used to generate an initial recommendation list based on the similarity score, select the top N components, and optimize the initial list by incorporating user historical behavior feedback factors and real-time market supply information factors. The final recommendation list is generated by sorting the components according to the score. The output module, connected to the recommendation list generation module, is used to display the final recommendation list to the user in a visual manner, including parameter details, similarity scores, feedback factors, and market supply information for each component.

[0028] The subjective weight calculation submodule in the weight calculation module also includes an expert weight adjustment unit.

[0029] The similarity calculation module also includes a hard constraint filtering unit and a parameter importance adjustment unit.

[0030] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0031] Finally, it should be noted that the above descriptions are merely preferred embodiments of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or make equivalent substitutions for some of the technical features. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. An electronic component automatic recommendation method based on multi-dimension parameter weight, characterized in that, Includes the following steps: Step 1: Construct a multi-dimensional parameter database for electronic components. This database includes the inherent attribute parameters, performance parameters, application scenario parameters, and user requirement-related parameters of electronic components. Step two, preprocessing the multi-dimensional parameters, including data cleaning, outlier processing and standardization conversion, wherein the data cleaning step is used to eliminate parameter records with missing values accounting for more than a preset threshold in the database, the preset threshold being 20% of the total number of parameter items in the parameter record, the outlier processing step adopts an improved Z-score method, for each parameter item, the Z-score value is calculated as wherein is the parameter value, is the mean value of the parameter item, is the standard deviation of the parameter item, when the absolute value is greater than 3, it is determined to be an outlier, and the median of the parameter item is used to replace the outlier, the standardization conversion step adopts a min-max standardization method to convert all parameter values to the [0, 1] interval, and the conversion formula is wherein is the standardized parameter value, is the minimum value of the parameter item, is the maximum value of the parameter item; Step three, determining the weight value of each dimension parameter, the determination of the weight value adopts the combination weight calculation method combining subjective weight and objective weight, wherein the subjective weight is determined by analytic hierarchy process, specifically including constructing a hierarchical structure model, dividing the multi-dimensional parameters into target layer, criterion layer and index layer, the target layer is the electronic component recommendation fitness, the criterion layer is four categories of inherent attribute, performance, application scene and user demand, the index layer is each specific parameter item, by inviting at least five experts in the field of electronic components to compare the parameter items in the same level two by two, a judgment matrix is constructed wherein represents the importance degree of the th parameter item relative to the th parameter item, the value of is an integer in the 1-9 scale method, and satisfies , =1, the maximum eigenvalue of the judgment matrix and the corresponding eigenvector are calculated , the eigenvector is normalized to obtain the subjective weight vector, and its expression is ;in This represents the weight vector of sub-level factors. For the first The importance weight of each sub-level factor relative to its parent-level goal. The number of factors in each sub-level is determined, and a consistency test is performed. The consistency index is... When the consistency ratio When <0.1, the judgment matrix has satisfactory consistency, where The average random consistency index; Based on matrix order Select the corresponding values; the objective weights are determined using the entropy weight method, specifically including calculating the... Information entropy of each parameter item ,in The number of component samples. For the first The sample at the th The proportion of the standardized value of each parameter to the sum of all standardized values ​​of that parameter, i.e. ,when When =0, define Then calculate the objective weight. To obtain the objective weight vector Combined weight vector It is obtained by linearly fusing subjective weights and objective weights, i.e. ; Step 4: Obtain the target requirement parameters input by the user. These target requirement parameters include the performance index range specified by the user, application scenario description, budget constraints, and delivery time requirements. Then, convert the target requirement parameters into a standardized requirement vector corresponding to the parameter items in the database. Each of them For users to the first A standardized representation of the required value or range of each parameter item; Step 5: Calculate the similarity between each electronic component sample in the database and the target demand vector. The similarity calculation uses the weighted cosine similarity algorithm, and the formula is as follows: ,in For the first The standardized values ​​of each parameter in the component sample. For the first The combined weights of the parameter items; Step 6: Sort all electronic component samples according to the similarity calculation results, and select the top N components with the highest similarity scores as the initial recommendation list, where N is the number of recommendations that can be set by the user, with a default value of 10. Step seven involves optimizing and adjusting the initial recommendation list. This optimization process includes introducing a user historical behavior feedback factor (β). This factor is calculated based on user clicks, favorites, purchases, and reviews of historical recommendation results. ,in This is the normalized value of the click frequency. This is a normalized value based on the number of collections. To improve conversion rates, To evaluate the score normalization values, all normalization values ​​are transformed to the [0,1] interval, and the similarity score of each component in the initial recommendation list is calculated. With the corresponding feedback factor Multiply by each product to obtain the final fitness score. The system then re-sorts the components based on their final fit scores, generating and outputting the final recommended list of electronic components.

2. The automatic recommendation method for electronic components based on multi-dimensional parameter weights according to claim 1, characterized in that, The construction of the multi-dimensional parameter database also includes a dynamic parameter update mechanism. This mechanism automatically crawls the latest parameter information from the official websites of electronic component manufacturers, industry databases, and e-commerce platforms according to a preset time period of 7 days. For newly added component models, its multi-dimensional parameters are automatically extracted, standardized, and then entered into the database. For existing component models, when a change in parameter information is detected, such as performance parameter adjustments or price changes, the corresponding record in the database is automatically updated, and the parameter change time and content are marked. Simultaneously, a version control mechanism is established for each parameter item, saving at least the most recent three parameter change histories. This allows for matching calculations of parameter versions at different time points based on user needs during the recommendation process. Furthermore, the database also stores information on alternative models for each component, including direct replacement models, functionally compatible replacement models, and performance-upgraded replacement models. It records the parameter difference between the replacement model and the original model. The parameter difference is obtained by calculating the weighted Euclidean distance between the two models on core parameter items, including rated voltage, rated current, power, and size. The weighted Euclidean distance formula is... ,in The weights of the core parameter items, The first of the original models One core parameter value, The replacement model number One core parameter value.

3. The automatic recommendation method for electronic components based on multi-dimensional parameter weights according to claim 1, characterized in that, The analytic hierarchy process (AHP) also incorporates an expert weight adjustment mechanism in the construction of the judgment matrix. Expert weights are assigned to each expert based on their years of industry experience, past parameter evaluation accuracy, and professional field matching degree. Among them, experts with more than 10 years of experience in the industry The value is 1.0 for periods of 5-10 years, 0.8 for periods of less than 5 years, and 0.6 for periods of less than 5 years. The accuracy of past parameter assessments is calculated based on the consistency between historical expert assessments and feedback from actual applications, with an accuracy exceeding 90%. An extra 0.2 points are awarded for scores between 70% and 90%, and an extra 0.1 points are awarded for scores below 70%. The professional field matching degree is determined by the overlap between the expert's research field and the field of the currently recommended component. A perfect match is awarded... An extra 0.3 is awarded for a partial match, an extra 0.15 for a partial match, and no extra points are awarded for a non-match. Expert weighting applies. The total value range is [0.6, 1.5], and for all experts... Normalization is performed to make ,in , For the first The original weights of the experts, For the first The normalized weights of the experts are used to determine the final judgment matrix, which is obtained by weighted averaging of the judgment matrices of all experts. ,in For the first A judgment matrix constructed by experts is used to improve the objectivity and reliability of subjective weights.

4. The automatic recommendation method for electronic components based on multi-dimensional parameter weights according to claim 1, characterized in that, In the similarity calculation step, when the user sets hard constraints on certain parameters, such as requiring a minimum or maximum value for a performance parameter to be met, a two-step calculation method is used: The first step is to screen all component samples and remove those that do not meet the hard constraints. Meeting the hard constraints is determined by the lower limit of the parameters specified by the user. The corresponding parameter value y in the component sample is ≥ For user-specified parameter upper limit The corresponding parameter value y in the component sample is ≤ For user-specified mandatory parameter attributes, such as specific certification standards, the component sample must include that attribute; The second step involves calculating the weighted cosine similarity of the selected samples, while simultaneously introducing a dynamic adjustment factor for parameter importance into the similarity calculation. When a user explicitly emphasizes the importance of a certain parameter, such as marking it as "important" or "very important" when inputting their requirements, the combined weight of that parameter will be adjusted. Multiply by the corresponding Among them, "important" corresponds to =1.5, "very important" corresponds to =2.0, unmarked parameter item =1.0, the adjusted weight is The adjusted weights are then normalized to ensure that... This is to ensure the accuracy of similarity calculations.

5. The automatic recommendation method for electronic components based on multi-dimensional parameter weights according to claim 1, characterized in that, The optimization and adjustment of the recommended list also includes the introduction of a real-time market supply information factor. This factor is calculated based on the current inventory quantity of each component, supplier delivery cycle, and price fluctuations. Specifically... ,in To ensure sufficient inventory, when the inventory quantity is more than 5 times the amount the user may purchase, =1.0, 2-5 times =0.7, 1-2 times =0.4, less than 1 time =0.1, For delivery cycle satisfaction, when the delivery cycle is less than 80% of the user's required time, =1.0, t=0.8 when 80%-100%, t=1.0 when 100%-120% =0.5, when it is greater than 120%, t=0.

2. For price stability, when the price fluctuation range is less than 5% in the past 30 days =1.0, 5%-10% =0.7, greater than 10% =0.4, all factor values ​​were converted to the [0,1] interval, and the final fit score was adjusted to The recommended list also displays market supply information for each component. This value is provided for users to refer to comprehensively.

6. A system applied to the automatic electronic component recommendation method based on multi-dimensional parameter weights as described in any one of claims 1-5, characterized in that, include: The multi-dimensional parameter database module is used to store the inherent attribute parameters, performance parameters, application scenario parameters, and user requirement-related parameters of electronic components. This database module also has a parameter dynamic update function, which can crawl and update the latest parameter information according to a preset cycle, and establish a parameter version control mechanism and alternative model association. The parameter preprocessing module is connected to the multi-dimensional parameter database module and is used to perform data cleaning, outlier handling and standardization transformation on the multi-dimensional parameters in the database. Data cleaning is used to remove parameter records with missing values ​​exceeding 20%, and outlier handling adopts the improved Z-score method. The weight calculation module, connected to the parameter preprocessing module, is used to determine the combined weights of parameters in each dimension. This module includes a subjective weight calculation submodule and an objective weight calculation submodule. The subjective weight calculation submodule uses the Analytic Hierarchy Process (AHP) to obtain the subjective weight vector by constructing a hierarchical structure model, expert judgment matrix, consistency check and eigenvector calculation. The objective weight calculation submodule uses the entropy weight method to obtain the objective weight vector by calculating information entropy and objective weights. The combined weight vector is obtained by linear fusion. The requirement acquisition module is used to receive target requirement parameters input by users, including performance index range, application scenario description, budget constraints and delivery time requirements, and convert them into standardized requirement vectors. The similarity calculation module is connected to the parameter preprocessing module, the weight calculation module, and the demand acquisition module, respectively, and is used to calculate the similarity between the electronic component samples and the target demand vector. The recommendation list generation module is connected to the similarity calculation module. It is used to generate an initial recommendation list based on the similarity score, select the top N components, and optimize the initial list by introducing user historical behavior feedback factors and real-time market supply information factors. The final recommendation list is generated by sorting the components according to the score. The output module, connected to the recommendation list generation module, is used to display the final recommendation list to the user in a visual manner, including parameter details, similarity scores, feedback factors, and market supply information for each component.

7. The automatic electronic component recommendation system based on multi-dimensional parameter weights according to claim 6, characterized in that, The subjective weight calculation submodule in the weight calculation module also includes an expert weight adjustment unit.

8. The automatic electronic component recommendation system based on multi-dimensional parameter weights according to claim 6, characterized in that, The similarity calculation module also includes a hard constraint screening unit and a parameter importance adjustment unit.

Images