A quantitative evaluation method and system for the value of industrial design professional members
By obtaining job information from the recruitment website, performing word segmentation and feature processing, and training CNN neural network, the problem of numerous indicator selection and artificial subjective influence in the quantitative evaluation of value of industrial design professional members is solved, and an automated and efficient value evaluation is achieved.
Patent Information
- Application Number
- CN202410632206.X
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2024-05-21
- Publication Date
- 2024-11-22
- Estimated Expiration
- 2044-05-21
AI Technical Summary
The existing technology lacks specific means to quantitatively evaluate the value of members of industrial design. The selection evaluation index is numerous and subjectively affected by artificial subjectiveness, and the fitting effect of conventional regression models is not good.
By obtaining the job requirements description text and job salary for industrial design positions from the recruitment website, perform word segmentation and feature vector processing, calculate the eigenvalues of the covariance matrix, select the main features to train the CNN neural network, and conduct value quantitative evaluation.
Automatically select the optimal evaluation indicators, reduce the workload of manual selection, improve evaluation efficiency, avoid inaccurate problems of routine fitting, and achieve more efficient machine learning training.
Smart Images

Figure CN118469389B_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of industrial design professional member data processing, and specifically to a method and system for quantitatively evaluating the value of industrial design professional members. Background Art
[0002] The field of industrial design engineering includes two professional directions: product design and experience design. The product design direction includes design strategy, product planning, user research, product styling design, color and material design, ergonomics design, product structure and process design, product display design, product packaging design, service process and business model design, etc. Experience design includes user experience, interaction design, product architecture, brand planning, artistic creativity, interface design, visual design, original painting design, character design, motion effect design, multimedia design, service design, etc.
[0003] With the development of social economy, after products meet basic needs, user experience becomes more and more important. There are more and more people engaged in industrial design and the number of guild members has skyrocketed. In order to better grasp industry data and better serve members of the Industrial Design Guild, it is necessary to conduct a quantitative value assessment of professional industrial design members.
[0004] However, there is no specific means to conduct quantitative value assessment of industrial design professional members, and there is no clear guidance data on which evaluation indicators to use and the weight of each indicator. Industrial design involves many professional indicators, and it is difficult to select specific representative indicators. The manual selection workload is large and is subject to human subjective influence. At the same time, the influence coefficient of each indicator does not change linearly, and it is difficult to fit a model with good results using conventional regression models. Summary of the invention
[0005] In order to solve the problems in the prior art, the present invention provides a method and system for quantitatively evaluating the value of industrial design professional members.
[0006] In one aspect of the present invention, a method for quantitatively evaluating the value of industrial design professional members is provided, comprising: obtaining job requirement description texts and job salaries of industrial design positions through recruitment websites; performing word segmentation on each of the job requirement description texts to obtain multiple first word segmentation lists; removing word segmentations whose co-occurrence rates are greater than a first threshold value from the multiple first word segmentation lists to obtain multiple second word segmentation lists; merging and removing duplicates from the multiple second word segmentation lists to obtain a third word segmentation list; performing the following operations on each second word segmentation list: checking each element in the third word segmentation list, and if it exists in the second word segmentation list, setting the value of the corresponding position to 0, otherwise setting it to 1, to obtain each second word segmentation list. The invention discloses a method for obtaining a first description vector of an industrial design position; randomly selecting a plurality of the first description vectors to form a first matrix, calculating the covariance matrix of the first matrix, and performing eigenvalue decomposition on the covariance matrix to obtain eigenvalues; taking the features of the second threshold before the eigenvalue as the main elements, and removing the non-main elements in the first description vector to obtain a second description vector; taking the second description vector as the input vector, and taking the position salary corresponding to the second description vector as the target value to train a first neural network to obtain a first model; organizing the features of the object for which the value of the industrial design professional member is to be quantitatively evaluated into a second description vector, and inputting the second description vector into the first model, and the first model outputs the quantitative evaluation result.
[0007] Furthermore, obtaining the job requirement description text and job salary of the industrial design position through the recruitment website includes: using the recruitment website API or using a crawler to obtain the job requirement description text and job salary of the industrial design position.
[0008] Furthermore, the covariance matrix is calculated by the following formula to obtain the covariance matrix C:
[0009]
[0010] where c i,j represents the element in the i-th row and j-th column of the covariance matrix, x ki represents the i-th feature of the k-th position, represents the mean of the i-th feature, x kj represents the jth feature of the kth position, represents the mean of the jth feature, and N represents the number of rows in the original matrix.
[0011] Furthermore, by C = VΛV -1 Perform eigendecomposition on the covariance matrix C to determine the eigenvalues of the covariance matrix C, where V is a matrix containing eigenvectors, Λ is a diagonal matrix, and the elements on the diagonal are the eigenvalues of C, V -1 Represents the inverse matrix of matrix V.
[0012] Furthermore, the first neural network is CNN.
[0013] Another aspect of the present invention is to provide a quantitative evaluation system for the value of industrial design professional members, characterized in that the system comprises the following modules:
[0014] The acquisition module is used to obtain the job requirements description text and job salary of industrial design positions through recruitment websites;
[0015] A processing module is used to segment each of the job requirement description texts to obtain multiple first segmentation lists; remove the segmentations whose co-occurrence rate is greater than a first threshold in the multiple first segmentation lists to obtain multiple second segmentation lists; merge and remove duplicates of the multiple second segmentation lists to obtain a third segmentation list; perform the following operations on each second segmentation list: check each element in the third segmentation list, if it exists in the second segmentation list, set the value of the corresponding position to 0, otherwise set it to 1, to obtain a first description vector for each industrial design position; randomly select multiple first description vectors to form a first matrix, calculate the covariance matrix of the first matrix, and perform eigenvalue decomposition on the covariance matrix to obtain eigenvalues; take the features of the second threshold before the eigenvalue as the main elements, and remove the non-main elements in the first description vector to obtain the second description vector;
[0016] A training module, used for training the first neural network with the second description vector as an input vector and the job salary corresponding to the second description vector as a target value to obtain a first model;
[0017] The evaluation module is used to organize the features of the object to be quantitatively evaluated for the value of industrial design professional members into a second description vector and input it into the first model, and the first model outputs the quantitative evaluation result.
[0018] Furthermore, obtaining the job requirement description text and job salary of the industrial design position through the recruitment website includes: using the recruitment website API or using a crawler to obtain the job requirement description text and job salary of the industrial design position.
[0019] Furthermore, the covariance matrix is calculated by the following formula to obtain the covariance matrix C:
[0020]
[0021] where c i,j represents the element in the i-th row and j-th column of the covariance matrix, x ki represents the i-th feature of the k-th position, represents the mean of the i-th feature, x kj represents the jth feature of the kth position, represents the mean of the jth feature, and N represents the number of rows in the original matrix.
[0022] Furthermore, by C = VΛV -1 Perform eigendecomposition on the covariance matrix C to determine the eigenvalues of the covariance matrix C, where V is a matrix containing eigenvectors, Λ is a diagonal matrix, and the elements on the diagonal are the eigenvalues of C, V -1 Represents the inverse matrix of matrix V.
[0023] Furthermore, the first neural network is CNN.
[0024] The present invention can produce the following beneficial effects through the above technical solution:
[0025] Through the above technical solution, the optimal evaluation indicators can be automatically selected, and the number of evaluation indicators can be controlled to a minimum, avoiding the technical problems of large workload and subjective influence during manual selection, and using the optimal evaluation indicators to train machine learning models to avoid the inaccuracy of conventional fitting. At the same time, since the number of indicators is controlled, the efficiency is also higher when conducting machine learning training. BRIEF DESCRIPTION OF THE DRAWINGS
[0026] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.
[0027] Figure 1 is a flow chart of the method of the present invention. DETAILED DESCRIPTION
[0028] Below, the invention is preferably described in conjunction with the accompanying drawings and specific implementation methods.
[0029] This embodiment solves the above problem through the following steps:
[0030] In one embodiment, reference Figure 1 The present invention provides a method for quantitatively evaluating the value of industrial design professional members, which specifically includes:
[0031] Obtain job description text and salary for industrial design positions through recruitment websites.
[0032] Due to the lack of industry evaluation standards and the need to use machine learning models in the future, a large amount of training data is required. In order to obtain objective data, this implementation obtains job description texts and job salaries for industrial design positions from recruitment websites. Based on the existing recruitment information, a quantitative assessment of the value of industrial design professional members is conducted.
[0033] Choose a well-known recruitment website, such as 51jbo, Liepin, etc., or a specialized recruitment website for the design industry. Obtain information about industrial design positions on the recruitment website through the professional API provided by the recruitment website or by using crawlers and other technologies under the premise of complying with legal provisions. You can use a computer program to obtain information by entering corresponding keywords, such as visual design, color and material design, etc., obtain the return results of the website, and extract the job description text and job salary, and store them in local data for subsequent use. Preferably, the job salary uniformly uses the minimum salary provided on the website.
[0034] For example, the description of a job requirement is as follows:
[0035] 1. Visual design related majors are preferred, and work experience is preferred;
[0036] 2. The portfolio includes previous design works and AIGC works. Applicants with AIGC commercialization cases are preferred, including but not limited to the food and beverage industry, 3C product industry, beauty industry, etc.;
[0037] 2. Proficient in using Adobe software, and have more than 3 months of design work experience in visual design / product design;
[0038] 3. Strong learning ability, in-depth understanding of AIGC products on the market, and tracking of AI technology development trends.
[0039] Use Python and other computer programs to obtain the above description file through the feature word "[Job Requirements]", and obtain the salary of 8000-10000, and use 8000 as the salary feature of the above position.
[0040] Each of the job requirement description texts is segmented to obtain a plurality of first segmentation lists.
[0041] In order to obtain the characteristic information of each position, you can use natural language processing tools or libraries to process text. For example, use the NLTK library in Python to segment the job description text, and output a list of segmented words for the job description text.
[0042] For example, the following results can be obtained by performing word segmentation on the text in the aforementioned job requirement description example:
[0043] [Visual design, related major, preferred, work experience, preferred, portfolio, including, previous design works, AIGC, works, commercialization...]
[0044] A large amount of industrial design-related job information can be obtained through recruitment websites. If similar processing is performed on each job information, a corresponding segmented word list can be obtained for each job information.
[0045] Job descriptions usually have many basic skill requirements, such as design majors, experience, etc., and the value of talent is usually evaluated based on their unique abilities. Therefore, in order to highlight the unique abilities of professionals, it is necessary to remove the basic requirements and only keep the bonus items. Therefore, the words with a co-occurrence rate greater than the first threshold in the multiple first word segmentation lists are removed to obtain multiple second word segmentation lists.
[0046] The co-occurrence rate refers to the probability or frequency of simultaneous occurrence between two or more objects or features. For example, if the word "experienced" appears in 95% of the first segmentation lists, its co-occurrence rate is 95%. Similarly, the co-occurrence rate of each segmentation in multiple first segmentation lists can be calculated. If the co-occurrence rate is large, it means that it is a basic requirement and cannot produce discrimination, so it should be removed. The first threshold can be determined based on experimental data, and the value set for different data sources can be different. This embodiment does not make a specific limitation. For example, the first threshold is set to 80%, then the segmentations with a co-occurrence rate greater than 80% in multiple first lists are removed, and the second segmentation list is obtained.
[0047] In order to determine the overall characteristics of industrial design professional members, that is, the labels that an industrial design professional member may have, the multiple second segmentation lists are combined and duplicated to obtain a third segmentation list. Each segmentation in the second segmentation list is a description of the industrial design professional member. After all descriptions are collected and duplicated, a total feature can be obtained. Usually, each member will have some of these features.
[0048] For example, if the second segmentation words are [bachelor's degree, AIGC, Adobe], [AIGC, animation], and [CG, animation], then the third segmentation word list obtained by merging them is [bachelor's degree, AIGC, Adobe, CG, animation]. The above is just a simple example. In actual operation, the number of feature words and lists will be much larger than the above example.
[0049] In order to facilitate the subsequent machine learning, the features of each position are represented by binary features. For each second word list, the following operations are performed: check each element in the third word list, if it exists in the second word list, set the value of the corresponding position to 0, otherwise set it to 1, and obtain the first description vector of each industrial design position.
[0050] Let's continue with the example that the second word segmentation is [bachelor's degree, AIGC, Adobe], [AIGC, animation], and [CG, animation], and the third word segmentation list is [bachelor's degree, AIGC, Adobe, CG, animation]. The third word segmentation list [bachelor's degree, AIGC, Adobe, CG, animation] includes 5 elements, so the generated description vector is also 5 elements. Detect each second word segmentation list. For example, when detecting [bachelor's degree, AIGC, Adobe], traverse each element in [bachelor's degree, AIGC, Adobe, CG, animation]. If "bachelor's degree" is included in [bachelor's degree, AIGC, Adobe], the first element of the generated first description vector is 1, otherwise it is 0. By analogy, we can get the first description vector corresponding to [bachelor's degree, AIGC, Adobe] is [1, 1, 1, 0, 0]; the description vector corresponding to [AIGC, animation] is [0, 1, 0, 0, 1]; the description vector corresponding to [CG, animation] is [0, 0, 0, 1, 1].
[0051] By vectorizing the job documents into binarized form, we can determine the skill requirements for each job. At the same time, the training process of machine learning using binarized data will be very fast, which can greatly reduce the training time.
[0052] Since the word segmentation result data is usually large, if the first description vector is directly used for training, the dimension of the data will be relatively large. Therefore, further, a plurality of the first description vectors are randomly selected to form a first matrix, the covariance matrix of the first matrix is calculated, and the covariance matrix is subjected to eigenvalue decomposition to obtain eigenvalues.
[0053] The number of positions is huge. If all the first description vectors are used to form a matrix, the subsequent calculation process will be very complicated. In order to reduce the amount of calculation, this embodiment randomly selects multiple first description vectors to form a first matrix. The selected data should cover all the characteristics of the position as much as possible and be as much as possible under the premise of acceptable calculation time. Different numbers can be selected according to different calculation hardware and software conditions during the experiment, and this embodiment does not make specific restrictions.
[0054] Suppose the first matrix has N rows and M columns, where each row represents the characteristics of a position and each column represents a characteristic.
[0055] The covariance matrix C is obtained by calculating the covariance matrix using the following formula:
[0056]
[0057] where c i,j represents the element in the i-th row and j-th column of the covariance matrix, x ki represents the i-th feature of the k-th position, represents the mean of the i-th feature, xkj represents the jth feature of the kth position, represents the mean of the jth feature.
[0058] Through C = VΛV -1 Perform eigendecomposition on the covariance matrix C and determine the eigenvalues of the covariance matrix C.
[0059] where V is a matrix containing eigenvectors, Λ is a diagonal matrix whose diagonal elements are the eigenvalues of C, and V -1 Represents the inverse matrix of matrix V.
[0060] A large eigenvalue indicates that the degree of data change is large in the direction of the corresponding eigenvector. This means that the direction represented by the eigenvector is the main direction of data change, and the corresponding eigenvalue represents the magnitude of data change in this direction. Therefore, the eigenvector with a large eigenvalue has a greater impact on the data, and it can be said that it has a greater impact on the result. Therefore, the features of the second threshold before the eigenvalue are taken as the main elements, and the non-main elements in the first description vector are removed to obtain the second description vector. That is to say, further, the elements that have a greater impact on the results are taken out (the elements with commonalities are further removed), the dimension of the data is further reduced, and the efficiency of subsequent machine learning is improved.
[0061] Let's continue with the example of the second word list of [bachelor's degree, AIGC, Adobe], [AIGC, animation], [CG, animation], and the third word list of [bachelor's degree, AIGC, Adobe, CG, animation]. Assuming that the feature values of AIGC, Adobe, and CG are ranked in the top three, AIGC, Adobe, and CG are selected as the main elements. Then, the elements at the 1st and 5th positions in the first description vector need to be ranked, that is, the second description vector corresponding to [bachelor's degree, AIGC, Adobe] is [1, 1, 0]; the description vector corresponding to [AIGC, animation] is [1, 0, 0]; and the description vector corresponding to [CG, animation] is [0, 0, 1].
[0062] Through the above operation, the optimal evaluation index can be automatically selected, and the number of evaluation indicators can be controlled to a minimum, thereby avoiding the technical problems of large workload and subjective influence during manual selection.
[0063] Furthermore, since the present embodiment obtains a large sample of sample data through the network, it is possible to automatically fit by using a machine learning model to avoid human interference in selecting a fitting model. The second description vector is used as the input vector, and the salary of the position corresponding to the second description vector is used as the target value to train the first neural network to obtain the first model. Since each description vector corresponds to a position, and each position corresponds to a salary, the salary of the position corresponding to each second description vector can be obtained. Exemplarily, the second description vector corresponding to [undergraduate, AIGC, Adobe] is [1, 1, 0] and the corresponding salary is 8000; the description vector corresponding to [AIGC, animation] is [1, 0, 0] and the corresponding salary is 10000; the description vector corresponding to [CG, animation] is [0, 0, 1] and the corresponding salary is 12000, then [1, 1, 0]: 8000; [1, 0, 0]: 10000; [0, 0, 1]: 12000 are three samples, where the second description vector is the input vector and the salary is the target value. The first neural network can be any predictive neural network in the prior art, such as CNN. This embodiment does not specifically limit the structure of the neural network.
[0064] When conducting a specific evaluation, the features of the object to be evaluated for the value of industrial design professional membership are organized into a second description vector and input into the first model, and the first model outputs a quantitative evaluation result. For example, if the main elements are AIGC, Adobe, and CG, and the skills of the object to be evaluated for the value of industrial design professional membership include Adobe and CG, then they are vectorized to [0,1,1], and the vector [0,1,1] is input into the trained neural network model, and the output prediction value is the quantitative evaluation result of its value.
[0065] Through the above technical solution, the optimal evaluation indicators can be automatically selected and the number of evaluation indicators can be controlled to a minimum, avoiding the technical problems of large workload and subjective influence during manual selection, and using the optimal evaluation indicators to train machine learning models to avoid the inaccuracy of conventional fitting.
[0066] On the other hand, the present invention also provides a quantitative evaluation system for industrial design professional member value, comprising:
[0067] The acquisition module is used to obtain the job requirements description text and job salary of industrial design positions through recruitment websites;
[0068] A processing module is used to segment each of the job requirement description texts to obtain multiple first segmentation lists; remove the segmentations whose co-occurrence rate is greater than a first threshold in the multiple first segmentation lists to obtain multiple second segmentation lists; merge and remove duplicates of the multiple second segmentation lists to obtain a third segmentation list; perform the following operations on each second segmentation list: check each element in the third segmentation list, if it exists in the second segmentation list, set the value of the corresponding position to 0, otherwise set it to 1, to obtain a first description vector for each industrial design position; randomly select multiple first description vectors to form a first matrix, calculate the covariance matrix of the first matrix, and perform eigenvalue decomposition on the covariance matrix to obtain eigenvalues; take the features of the second threshold before the eigenvalue as the main elements, and remove the non-main elements in the first description vector to obtain the second description vector;
[0069] A training module, used for training the first neural network with the second description vector as an input vector and the job salary corresponding to the second description vector as a target value to obtain a first model;
[0070] The evaluation module is used to organize the features of the object to be quantitatively evaluated for the value of industrial design professional members into a second description vector and input it into the first model, and the first model outputs the quantitative evaluation result.
[0071] Furthermore, the specific implementation methods of the above-mentioned industrial design professional member value quantitative evaluation system are the same as an industrial design professional member value quantitative evaluation method, and all further technical solutions in an industrial design professional member value quantitative evaluation method are fully introduced into an industrial design professional member value quantitative evaluation system.
[0072] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the above embodiments, ordinary technicians in the relevant field should understand that the specific implementation methods of the present invention can still be modified or replaced by equivalents. Any modification or equivalent replacement that does not depart from the spirit and scope of the present invention should be covered within the scope of protection of the claims of the present invention.
[0073] The present invention does not particularly specify the structure of some modules, which shall be subject to the contents recorded in the prior art. The prior art mentioned in the above background technology section and the specific embodiment section of the present invention can be used as part of the present invention to understand the meaning of some technical features or parameters. The scope of protection of the present invention shall be subject to the contents actually recorded in the claims.
Claims
1. A method for quantitatively evaluating the value of industrial design professional members, characterized by: Obtain job descriptions and salaries for industrial design positions from recruitment websites; Segmenting each of the job requirement description texts to obtain a plurality of first segmentation lists; Remove the words whose co-occurrence rate is greater than a first threshold from the plurality of first word segmentation lists to obtain a plurality of second word segmentation lists; Merge and remove duplicates from the multiple second word segmentation lists to obtain a third word segmentation list; Perform the following operations on each second word list: Check each element in the third word list. If it exists in the second word list, set the value of the corresponding position to 0, otherwise set it to 1, and obtain the first description vector of each industrial design position; Randomly selecting a plurality of the first description vectors to form a first matrix, calculating a covariance matrix of the first matrix, and performing eigenvalue decomposition on the covariance matrix to obtain eigenvalues; Taking the features with the second threshold before the eigenvalue as the main elements, and removing the non-main elements in the first description vector, to obtain the second description vector; The first neural network is trained using the second description vector as an input vector and the job salary corresponding to the second description vector as a target value to obtain a first model; The features of the object for which the value of industrial design professional members is to be quantitatively evaluated are organized into a second description vector and input into the first model, and the first model outputs a quantitative evaluation result.
2. A method for quantitatively evaluating the value of industrial design professional members according to claim 1, characterized in that Obtaining the job requirement description text and job salary of an industrial design position through a recruitment website includes: using the recruitment website API or using a crawler to obtain the job requirement description text and job salary of the industrial design position.
3. The method for quantitatively evaluating the value of industrial design professional members according to claim 1 is characterized in that The covariance matrix C is obtained by calculating the covariance matrix using the following formula: where c i,j represents the element in the i-th row and j-th column of the covariance matrix, x ki represents the i-th feature of the k-th position, represents the mean of the i-th feature, x kj represents the jth feature of the kth position, represents the mean of the jth feature, and N represents the number of rows in the original matrix.
4. The method for quantitatively evaluating the value of industrial design professional members according to claim 1 is characterized by: Through C = VΛV -1 Perform eigendecomposition on the covariance matrix C and determine the eigenvalues of the covariance matrix C, where V is a matrix containing eigenvectors, Λ is a diagonal matrix, and the elements on the diagonal are the eigenvalues of C, V -1 Represents the inverse matrix of the matrix V.
5. The method for quantitatively evaluating the value of industrial design professional members according to claim 1 is characterized by: The first neural network is CNN.
6. A quantitative evaluation system for the value of industrial design professional members, characterized by The system includes the following modules: The acquisition module is used to obtain the job requirements description text and job salary of industrial design positions through recruitment websites; A processing module is used to segment each of the job requirement description texts to obtain multiple first segmentation lists; remove the segmentations whose co-occurrence rate in the multiple first segmentation lists is greater than a first threshold to obtain multiple second segmentation lists; merge and remove duplicates of the multiple second segmentation lists to obtain a third segmentation list; perform the following operations on each second segmentation list: check each element in the third segmentation list, if it exists in the second segmentation list, set the value of the corresponding position to 0, otherwise set it to 1, to obtain a first description vector for each industrial design position; randomly select multiple first description vectors to form a first matrix, calculate the covariance matrix of the first matrix, and perform eigenvalue decomposition on the covariance matrix to obtain eigenvalues; Taking the features with the second threshold before the eigenvalue as the main elements, and removing the non-main elements in the first description vector, to obtain the second description vector; A training module, used for training the first neural network with the second description vector as an input vector and the job salary corresponding to the second description vector as a target value to obtain a first model; The evaluation module is used to organize the features of the object to be quantitatively evaluated for the value of industrial design professional members into a second description vector and input it into the first model, and the first model outputs the quantitative evaluation result.
7. The industrial design professional member value quantitative evaluation system according to claim 6 is characterized in that Obtaining the job requirement description text and job salary of an industrial design position through a recruitment website includes: using the recruitment website API or using a crawler to obtain the job requirement description text and job salary of the industrial design position.
8. The industrial design professional member value quantitative evaluation system according to claim 6 is characterized in that The covariance matrix C is obtained by calculating the covariance matrix using the following formula: where c i,j represents the element in the i-th row and j-th column of the covariance matrix, x ki represents the i-th feature of the k-th position, represents the mean of the i-th feature, x kj represents the jth feature of the kth position, represents the mean of the jth feature, and N represents the number of rows in the original matrix.
9. The industrial design professional member value quantitative evaluation system according to claim 6 is characterized by: Through C = VΛV -1 Perform eigendecomposition on the covariance matrix C and determine the eigenvalues of the covariance matrix C, where V is a matrix containing eigenvectors, Λ is a diagonal matrix, and the elements on the diagonal are the eigenvalues of C, V -1 Represents the inverse matrix of the matrix V.
10. The industrial design professional member value quantitative evaluation system according to claim 6 is characterized by: The first neural network is CNN.
