CN113361751A - Firefly algorithm optimization-based product quality prediction method and device for RBF model - Google Patents

Firefly algorithm optimization-based product quality prediction method and device for RBF model Download PDF

Info

Publication number
CN113361751A
CN113361751A CN202110536540.1A CN202110536540A CN113361751A CN 113361751 A CN113361751 A CN 113361751A CN 202110536540 A CN202110536540 A CN 202110536540A CN 113361751 A CN113361751 A CN 113361751A
Authority
CN
China
Prior art keywords
product quality
firefly
quality prediction
model
clustering
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Pending
Application number
CN202110536540.1A
Other languages
Chinese (zh)
Inventor
韩慧慧
王坚
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Tongji University
Original Assignee
Tongji University
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Tongji University filed Critical Tongji University
Priority to CN202110536540.1A priority Critical patent/CN113361751A/en
Publication of CN113361751A publication Critical patent/CN113361751A/en
Pending legal-status Critical Current

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q10/00—Administration; Management
    • G06Q10/04—Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/004—Artificial life, i.e. computing arrangements simulating life
    • G06N3/006—Artificial life, i.e. computing arrangements simulating life based on simulated virtual individual or collective life forms, e.g. social simulations or particle swarm optimisation [PSO]
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/08—Learning methods
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q10/00—Administration; Management
    • G06Q10/06—Resources, workflows, human or project management; Enterprise or organisation planning; Enterprise or organisation modelling
    • G06Q10/063—Operations research, analysis or management
    • G06Q10/0639—Performance analysis of employees; Performance analysis of enterprise or organisation operations
    • G06Q10/06395—Quality analysis or management
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q10/00—Administration; Management
    • G06Q10/06—Resources, workflows, human or project management; Enterprise or organisation planning; Enterprise or organisation modelling
    • G06Q10/067—Enterprise or organisation modelling
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q50/00—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
    • G06Q50/04—Manufacturing
    • Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
    • Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
    • Y02P—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN THE PRODUCTION OR PROCESSING OF GOODS
    • Y02P90/00—Enabling technologies with a potential contribution to greenhouse gas [GHG] emissions mitigation
    • Y02P90/30—Computing systems specially adapted for manufacturing

Landscapes

  • Engineering & Computer Science (AREA)
  • Business, Economics & Management (AREA)
  • Human Resources & Organizations (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Strategic Management (AREA)
  • Economics (AREA)
  • General Physics & Mathematics (AREA)
  • Entrepreneurship & Innovation (AREA)
  • Development Economics (AREA)
  • Marketing (AREA)
  • General Business, Economics & Management (AREA)
  • Tourism & Hospitality (AREA)
  • Game Theory and Decision Science (AREA)
  • Quality & Reliability (AREA)
  • Operations Research (AREA)
  • Educational Administration (AREA)
  • Health & Medical Sciences (AREA)
  • General Health & Medical Sciences (AREA)
  • Evolutionary Computation (AREA)
  • Molecular Biology (AREA)
  • Biophysics (AREA)
  • Computational Linguistics (AREA)
  • Data Mining & Analysis (AREA)
  • Artificial Intelligence (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Biomedical Technology (AREA)
  • Computing Systems (AREA)
  • General Engineering & Computer Science (AREA)
  • Mathematical Physics (AREA)
  • Software Systems (AREA)
  • Manufacturing & Machinery (AREA)
  • Primary Health Care (AREA)
  • Management, Administration, Business Operations System, And Electronic Commerce (AREA)

Abstract

The invention relates to a firefly algorithm optimization-based product quality prediction method and a firefly algorithm optimization-based product quality prediction device for an RBF model, wherein the method comprises the following steps: obtaining quality data of a product to be tested, loading the data into a product quality prediction model, and obtaining the prediction results of qualification and unqualified products; the product quality prediction model adopts an RBF network with a Gaussian kernel function as a basic framework; the method comprises the steps of inputting a product quality data set into a product quality prediction model in the training process of the product quality prediction model, outputting a prediction result by the product quality prediction model in each input process to obtain a model loss function value, optimizing the number of neurons in the RBF network by adopting a fuzzy clustering FCM (fuzzy clustering algorithm) according to the model loss function value, and updating a clustering center and a target function value of the fuzzy clustering FCM by adopting a Firefly group optimization algorithm based on a Levy mechanism. Compared with the prior art, the method has the advantages of high accuracy, high speed, good approximation capability, global optimum performance and the like.

Description

Firefly algorithm optimization-based product quality prediction method and device for RBF model
Technical Field
The invention relates to the technical field of product quality prediction, in particular to a firefly algorithm optimization-based product quality prediction method and a firefly algorithm optimization-based product quality prediction device for an RBF model.
Background
The core content of the intelligent manufacturing execution system is quality management, which is the key for enterprises to gain profits and long-term competitive advantages. Product quality issues often stem from the manufacturing process. With the progress of production technology and the increasingly finer demands of people, the product manufacturing process becomes more complicated, so that the factors affecting the product quality are greatly increased. With the continuous development of the information-based construction of manufacturing enterprises, a large amount of data in the processes of product part design, production, quality detection and the like appears in an enterprise database after long-time accumulation. However, most enterprises simply make statistics and visual display of data in the quality control process. These operations neither use new generation data mining techniques nor exploit the underlying knowledge behind the data, resulting in rich product quality data, but poor application of the knowledge behind the data.
The knowledge discovery is a new generation technology combined with technologies in various fields such as probability theory, statistics and the like, can effectively discover potential knowledge behind data, helps enterprises to mine useful knowledge, and helps technicians to make decisions and determine reasonable processing technological parameters. The rule knowledge mined through knowledge discovery, namely the association rule between the part processing data and the product quality, can be used as the basis for quality prediction. The introduction of knowledge discovery can enable the production of parts to be more efficient and effective, and has great application value in enterprises. The classical algorithms for extracting association rules are Apriori, Eclat, FP-tree, etc.
In recent years, studies on product quality prediction have been widely conducted. In order to achieve real-time prediction of product quality in batch operations, Wang et al introduced a data-driven multi-model approach in 2011. Mao et al used a fuzzy analytic hierarchy process in 2016 to find out key factors affecting product quality. And a mechanical assembly precision prediction model is established by using a state space equation so as to improve the quality stability of the processed product. In order to improve the accuracy of product quality prediction, a product quality prediction model based on a BP neural network and a rough set theory is designed. To accurately predict the product quality and improve the interpretability of the model, Yeh et al applied a matrix decomposition technique and an attention mechanism in 2019 and proposed a multi-task learning model of an encoder-decoder structure. Yao et al propose a hierarchical quality monitoring algorithm based on a distributed parallel semi-supervised Gaussian mixture model to monitor product quality.
Although the performance of the previously proposed quality prediction models has improved to some extent, some challenges remain. First, many researchers do not consider the quality of product quality data to also affect the prediction effect of the model. Secondly, common product quality prediction models, such as BP neural networks, logistic regression, RF and SVM, have limitations of high complexity, low accuracy, slow convergence rate, and the like.
Disclosure of Invention
The invention aims to overcome the defects of high complexity, low precision, low convergence speed and the like in the prior art and provide a firefly algorithm-optimized RBF model-based product quality prediction method and device.
The purpose of the invention can be realized by the following technical scheme:
the invention provides a firefly algorithm optimization-based product quality prediction method for an RBF model, which comprises the following steps:
obtaining quality data of a product to be tested, loading the data into a product quality prediction model which is constructed in advance and trained, and obtaining the prediction results of qualified products and unqualified products; the product quality prediction model adopts an RBF network with a Gaussian kernel function as a basic framework;
the training process of the product quality prediction model comprises the following steps:
a1: acquiring a product quality data set, and preprocessing the product quality data set;
a2: encoding the preprocessed product quality data set by using an auto-encoder to obtain reconstruction characteristics;
a3: initializing an RBF network, sequentially inputting the reconstruction characteristics into a product quality prediction model, outputting a prediction result by the product quality prediction model in each input process to obtain a model loss function value, optimizing the number of neurons in the RBF network by adopting a fuzzy clustering FCM mechanism according to the loss function value of the model, and updating a clustering center and an objective function value of the fuzzy clustering FCM mechanism by adopting a Firefly group optimization algorithm based on a Levy mechanism;
a4: and repeating the step A3 until a preset training stopping condition is reached, and obtaining a trained product quality prediction model.
Further, initializing the RBF network comprises initializing a neuron center, a neuron base width, a neuron number and a connection weight of the RBF network;
the optimizing the number of the neurons in the RBF network by adopting a fuzzy clustering FCM mechanism specifically comprises the following steps: and iteratively optimizing the clustering centers and the clustering numbers by adopting a fuzzy clustering FCM mechanism so as to determine the neuron centers, the neuron base widths, the neuron numbers and the connection weights of the RBF network.
Further, the updating of the cluster center and the objective function of the fuzzy clustering FCM mechanism by using a Firefly cluster optimization algorithm based on a Levy mechanism specifically comprises the following steps:
a301: initializing basic parameters of the fuzzy clustering FCM mechanism;
a302: calculating the brightness of each firefly;
a303: updating the positions of the fireflies according to the brightness sequence of the fireflies;
a304: selecting the brightest firefly as a center, and calculating an objective function value of a fuzzy clustering FCM mechanism;
a305: calculating a membership function;
a306: updating the clustering center of the fuzzy clustering FCM mechanism according to the membership function;
a307: and repeating the steps A302 to A306 until a preset firefly algorithm stop condition is reached, and obtaining the updated cluster center and membership matrix of the fuzzy cluster FCM mechanism.
Further, the computational expression of the luminance of the firefly is:
Figure BDA0003070064900000031
in the formula IiIs the lightness, x, of the ith fireflyiIs the position of the ith firefly, J (x)i) Is the target function of the ith firefly, and n is the total number of the fireflies;
the updated expression of the position of the firefly is as follows:
Figure BDA0003070064900000032
in the formula (I), the compound is shown in the specification,
Figure BDA0003070064900000033
position of ith firefly at time t +1, β0The attractive force at the zero distance is,
Figure BDA0003070064900000034
for the relative attraction of the ith and jth fireflies,
Figure BDA0003070064900000035
the position of the ith firefly at time t,
Figure BDA0003070064900000037
to perturb the terms to enhance global search capability, alpha is a scale factor,
Figure BDA0003070064900000038
is a random number vector;
the calculation expression of the objective function value of the fuzzy clustering FCM mechanism is as follows:
Figure BDA0003070064900000036
wherein m is a fuzzy weight index for controlling the classification degree of the fuzzy matrix, JmFor fuzzy clustering of the objective function values of the FCM mechanism under the fuzzy weighting index m, U ═ Uij]c×nIs degree of membershipMatrix, c, uij∈[0,1]Represents the membership degree of the jth target belonging to the ith cluster, | xi-vj||2Is the euclidean distance of the jth target from the ith cluster center.
Further, the calculation expression of the membership function is as follows:
Figure BDA0003070064900000041
the updating expression of the clustering center is as follows:
Figure BDA0003070064900000042
further, the firefly algorithm stopping condition is that the distance between the cluster center of the updated fuzzy clustering FCM mechanism and the original cluster center is smaller than a preset distance threshold, or the preset maximum iteration number of the firefly algorithm is met.
Further, the preprocessing the product quality data set comprises the following steps:
a11: carrying out standardization and discretization operation on the quality data set of the product;
a12: processing the data set by using an FP-Growth association rule algorithm so as to extract a rule set, and then sequencing according to the support degree and the confidence degree to obtain an ordered rule set;
a13: performing a k-means clustering algorithm on a plurality of types of samples for clustering, and reserving a part of samples in each cluster according to a preset key rule; then, calculating the number of samples needing to be reserved after sampling through a first calculation formula, wherein the key rule is used for obtaining samples ranked in the front;
the expression of the first calculation formula is:
Figure BDA0003070064900000043
in the formula, NiFor samples to be retained after sampling in the ith clusterThe number of the principal, M is the total number of most classes before sampling, MiThe total number of samples in the ith cluster;
a14: performing a k-means clustering algorithm on the minority samples for clustering, then generating a new sample in each cluster by using a GAN method, and calculating the number of the new samples in the samples to be generated by using a second calculation formula;
Figure BDA0003070064900000044
in the formula, niThe number of minority classes required to be generated in the ith cluster, m is the total number of minority classes before sampling, miIs the initial number of samples in the ith cluster.
Further, the step a2 specifically includes the following steps:
a21: in the encoding process of the encoder, input data X ═ X (X)1,x2,…,xn) Is converted into a reconstructed feature ψ ═ (ξ)1,ξ2,…,ξd) The calculation expression of the reconstruction characteristic is as follows:
ψ＝L_ReLU(WeX+be)
in the formula, WeAs weights of the encoder, beIs the basis of the encoder;
a22: during decoding of a decoder, the reconstruction characteristic is restored to X ', and the calculation expression of X' is as follows:
X'＝L_ReLU(Wdψ+bd)
in the formula, WdAs weights of the decoder, bdIs the base of the decoder;
a23: optimizing a loss function loss of the self-encoder by using a gradient descent method, wherein the calculation expression of the loss function loss of the self-encoder is as follows:
Loss＝||X-X'||2
＝||X-f[g(x)]||2
＝||X-L_ReLU(Wd(L_ReLU(WeX+be))+bd)||2
where L _ ReLU (·) is the ReLU activation function.
A24: and repeating the steps A21 to A23 until the preset first coding times are reached, and selecting the corresponding reconstruction characteristics when the loss function loss of the self-coder is minimum as the input of the product quality prediction model.
Further, the training process of the product quality prediction model further comprises: dividing the product quality data set into a training set and a test set, and executing steps A2-A4 by adopting the training set; the training process of the product quality prediction model further comprises the step A5: inputting the test set into a trained product quality prediction model, if the accuracy of the output result is greater than or equal to a preset accuracy threshold, the trained product quality prediction model is the final prediction model, otherwise, returning to the step S2 to perform model training again.
The invention also provides a firefly algorithm optimized RBF model-based product quality prediction device which is characterized by comprising a memory and a processor, wherein the memory stores a computer program, and the processor calls the computer program to execute the steps of the method.
Compared with the prior art, the invention has the following advantages:
the invention introduces an optimized data equalization method to preprocess the data, thereby reducing the adverse effect of unbalanced data on a product quality prediction model. Secondly, the RBF network with the Gaussian kernel function is used as a basic framework, and a firefly algorithm with a Levy mechanism is used for optimizing the RBF network to construct a product quality prediction model RBFFALM. The model is high in accuracy and high in speed, and has good approximation capability and global optimal performance.
Drawings
FIG. 1 is a schematic flow diagram of the process of the present invention;
FIG. 2 is a schematic diagram of the structure of a product quality prediction model in an industrial process of the present invention;
FIG. 3 is a schematic structural diagram of a self-encoder;
FIG. 4 is a graph showing the result of S448 in the example.
Detailed Description
In order to make the objects, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are some, but not all, embodiments of the present invention. The components of embodiments of the present invention generally described and illustrated in the figures herein may be arranged and designed in a wide variety of different configurations.
Thus, the following detailed description of the embodiments of the present invention, presented in the figures, is not intended to limit the scope of the invention, as claimed, but is merely representative of selected embodiments of the invention. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
It should be noted that: like reference numbers and letters refer to like items in the following figures, and thus, once an item is defined in one figure, it need not be further defined and explained in subsequent figures.
In the description of the present invention, it should be noted that the terms "center", "upper", "lower", "left", "right", "vertical", "horizontal", "inner", "outer", etc. indicate orientations or positional relationships based on the orientations or positional relationships shown in the drawings or the orientations or positional relationships that the products of the present invention are conventionally placed in use, and are only used for convenience in describing the present invention and simplifying the description, but do not indicate or imply that the devices or elements referred to must have a specific orientation, be constructed and operated in a specific orientation, and thus, should not be construed as limiting the present invention.
It should be noted that the terms "first" and "second" are used for descriptive purposes only and are not to be construed as indicating or implying relative importance or implicitly indicating the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of that feature. In the description of the present application, "a plurality" means two or more unless specifically limited otherwise.
Example 1
The embodiment provides a firefly algorithm optimization-based product quality prediction method for an RBF model, which comprises the following steps:
obtaining quality data of a product to be tested, loading the data into a product quality prediction model which is constructed in advance and trained, and obtaining the prediction results of qualified products and unqualified products; the product quality prediction model adopts an RBF network with a Gaussian kernel function as a basic framework;
the training process of the product quality prediction model comprises the following steps:
a1: acquiring a product quality data set, and preprocessing the product quality data set;
a2: encoding the preprocessed product quality data set by using an auto-encoder to obtain reconstruction characteristics;
a3: initializing an RBF network, sequentially inputting the reconstruction characteristics into a product quality prediction model, outputting a prediction result by the product quality prediction model in each input process to obtain a model loss function value, optimizing the number of neurons in the RBF network by adopting a fuzzy clustering FCM mechanism according to the loss function value of the model, and updating a clustering center and an objective function value of the fuzzy clustering FCM mechanism by adopting a Firefly group optimization algorithm based on a Levy mechanism;
a4: and repeating the step A3 until a preset training stopping condition is reached, and obtaining a trained product quality prediction model.
Specifically, the initializing the RBF network comprises initializing a neuron center, a neuron base width, a neuron number and a connection weight of the RBF network;
the optimizing the number of the neurons in the RBF network by adopting a fuzzy clustering FCM mechanism specifically comprises the following steps: and iteratively optimizing the clustering centers and the clustering numbers by adopting a fuzzy clustering FCM mechanism so as to determine the neuron centers, the neuron base widths, the neuron numbers and the connection weights of the RBF network.
The updating of the clustering center and the objective function of the fuzzy clustering FCM mechanism by adopting a Firefly group optimization algorithm based on a Levy mechanism specifically comprises the following steps:
a301: initializing basic parameters of the fuzzy clustering FCM mechanism;
a302: calculating the brightness of each firefly;
a303: updating the positions of the fireflies according to the brightness sequence of the fireflies;
a304: selecting the brightest firefly as a center, and calculating an objective function value of a fuzzy clustering FCM mechanism;
a305: calculating a membership function;
a306: updating the clustering center of the fuzzy clustering FCM mechanism according to the membership function;
a307: and repeating the steps A302 to A306 until a preset firefly algorithm stop condition is reached, and obtaining the updated cluster center and membership matrix of the fuzzy cluster FCM mechanism.
The computational expression of the brightness of the firefly is as follows:
Figure BDA0003070064900000071
in the formula IiIs the lightness, x, of the ith fireflyiIs the position of the ith firefly, J (x)i) Is the target function of the ith firefly, and n is the total number of the fireflies;
the updated expression of the position of the firefly is as follows:
Figure BDA0003070064900000072
in the formula (I), the compound is shown in the specification,
Figure BDA0003070064900000073
the position of the ith firefly at time t +1,
Figure BDA0003070064900000074
the position of the ith firefly at time t,
Figure BDA0003070064900000075
to perturb the terms to enhance global search capability, alpha is a scale factor,
Figure BDA0003070064900000076
is a random number vector;
the calculation expression of the objective function value of the fuzzy clustering FCM mechanism is as follows:
Figure BDA0003070064900000081
wherein m is a fuzzy weight index for controlling the classification degree of the fuzzy matrix, JmFor fuzzy clustering of the objective function values of the FCM mechanism under the fuzzy weighting index m, U ═ Uij]c×nIs a membership matrix, c, uij∈[0,1]Represents the membership degree of the jth target belonging to the ith cluster, | xi-vj||2Is the euclidean distance of the jth target from the ith cluster center.
The calculation expression of the membership function is as follows:
Figure BDA0003070064900000082
the updating expression of the clustering center is as follows:
Figure BDA0003070064900000083
the firefly algorithm stopping condition is that the distance between the cluster center of the updated fuzzy cluster FCM mechanism and the original cluster center is smaller than a preset distance threshold value, or the preset maximum iteration times of the firefly algorithm are met.
As a preferred embodiment, the preprocessing the product quality data set comprises the following steps:
a11: carrying out standardization and discretization operation on the quality data set of the product;
a12: processing the data set by using an FP-Growth association rule algorithm so as to extract a rule set, and then sequencing according to the support degree and the confidence degree to obtain an ordered rule set;
a13: performing a k-means clustering algorithm on a plurality of types of samples for clustering, and reserving a part of samples in each cluster according to a preset key rule; then, calculating the number of samples needing to be reserved after sampling through a first calculation formula, wherein the key rule is used for obtaining samples ranked in the front;
the expression of the first calculation formula is:
Figure BDA0003070064900000084
in the formula, NiThe number of samples needing to be reserved after sampling in the ith cluster, M is the total number of most classes before sampling, M is the total number of the most classes before samplingiThe total number of samples in the ith cluster;
a14: performing a k-means clustering algorithm on the minority samples for clustering, then generating a new sample in each cluster by using a GAN method, and calculating the number of the new samples in the samples to be generated by using a second calculation formula;
Figure BDA0003070064900000085
in the formula, niThe number of minority classes required to be generated in the ith cluster, m is the total number of minority classes before sampling, miThe number of initial samples in the ith cluster; most of the samples in the category refer to: a certain type of data in the data is very much more than a preset majority of threshold values; the minority class samples refer to: the data of a certain class in the data set is very small and is less than a preset few threshold values.
8. The firefly algorithm optimized RBF model-based product quality prediction method according to claim 1, wherein said step A2 specifically comprises the following steps:
a21: in the encoding process of the encoder, input data X ═ X (X)1,x2,…,xn) Is converted into a reconstructed feature ψ ═ (ξ)1,ξ2,…,ξd) The calculation expression of the reconstruction characteristic is as follows:
ψ＝L_ReLU(WeX+be)
in the formula, WeAs weights of the encoder, beIs the basis of the encoder;
a22: during decoding of a decoder, the reconstruction characteristic is restored to X ', and the calculation expression of X' is as follows:
X'＝L_ReLU(Wdψ+bd)
in the formula, WdAs weights of the decoder, bdIs the base of the decoder;
a23: optimizing a loss function loss of the self-encoder by using a gradient descent method, wherein the calculation expression of the loss function loss of the self-encoder is as follows:
Loss＝||X-X'||2
＝||X-f[g(x)]||2
＝||X-L_ReLU(Wd(L_ReLU(WeX+be))+bd)||2
where L _ ReLU (·) is the ReLU activation function.
A24: and repeating the steps A21 to A23 until the preset first coding times are reached, and selecting the corresponding reconstruction characteristics when the loss function loss of the self-coder is minimum as the input of the product quality prediction model.
As a preferred embodiment, the training process of the product quality prediction model further includes: dividing the product quality data set into a training set and a test set, and executing steps A2-A4 by adopting the training set; the training process of the product quality prediction model further comprises the step A5: inputting the test set into a trained product quality prediction model, if the accuracy of the output result is greater than or equal to a preset accuracy threshold, the trained product quality prediction model is the final prediction model, otherwise, returning to the step S2 to perform model training again.
The embodiment also provides a product quality prediction device based on a firefly algorithm optimized RBF model, which is characterized by comprising a memory and a processor, wherein the memory stores a computer program, and the processor calls the computer program to execute the steps of the method.
A combination of the above preferred embodiments can provide an optimal embodiment, which will be described in detail below.
A product quality prediction method of a firefly algorithm optimized RBF model comprises the following steps:
and S1, acquiring the quality data set of the product, and carrying out data equalization preprocessing operation on the quality data set of the product.
S2, dividing the data set after data preprocessing into a training set and a testing set;
s3, coding the data by using a self-coder to obtain reconstruction characteristics;
s4, constructing a product quality prediction model, called RBFFALM for short, based on the RBF model optimized by the firefly algorithm with the Levy mechanism by using the reconstruction characteristics. Training the RBFFALM model by using a training set to obtain an optimal quality prediction model;
and S5, inputting the quality data of the product to be tested in the actual production process into the RBFFALM model, and outputting the qualified and unqualified results of the product.
In this embodiment, the hot rolling data set is from Baoku group, and includes 3580 pieces of data, each piece of data has 32 dimensions, that is, 32 factors affecting the quality of the final product in the production process.
Carrying out data equalization operation on the data set:
and S11, carrying out standardization and discretization operation on the data set.
S12, processing the data set by using the FP-Growth association rule algorithm so as to extract a rule set, and then sequencing according to the support degree and the confidence degree to obtain an ordered rule set. The key rule set is the top ranked rule.
And S13, performing a k-means clustering algorithm on the majority of samples for clustering, and keeping part of samples in each cluster according to a key rule. Then, the number of samples to be retained after sampling is calculated by equation (1).
Figure BDA0003070064900000101
Wherein N isiNumber of samples to be reserved for sampling in ith cluster, M and MiThe total number of most classes before sampling and the total number of samples in the ith cluster, respectively.
S14, performing k-means clustering algorithm on the minority samples for clustering, and then generating a new sample in each cluster by using a GAN method, wherein the number of the new samples in the samples to be generated can be calculated by formula (2).
Figure BDA0003070064900000102
Wherein n isiThe number of generations, m and m, required for the minority class in the ith clusteriThe total number of the minority classes before sampling and the initial number of samples in the ith cluster are respectively.
And then S2 takes 70% of the data set of the quality of the product in the industrial processing process after the data equalization preprocessing operation as a training set and 30% as a testing set.
Further, the step S3 specifically includes the following steps:
s31, in the process of encoding by the encoder, the input data X ═ X1,x2,…,xn) Is converted into a reconstructed feature ψ ═ (ξ)1,ξ2,…,ξd) See formula (3).
ψ＝L_ReLU(WeX+be) (3)
Wherein, WeAnd beRespectively, as weights and bases of the encoder.
S32, during decoding, the reconstruction feature is reduced to X', see equation (4).
X'＝L_ReLU(Wdψ+bd) (4)
Wherein, WdAnd bdRespectively expressed as weights and bases of the decoder
S33, optimizing the loss function loss by using a gradient descent method, see formula (5).
Figure BDA0003070064900000111
S34, when the loss function is minimum, changing the reconstruction characteristic ψ (ξ)1,ξ2,…,ξd) As input to the RBF model.
The data dimension reduction is realized by reconstructing the features through the encoder, the original data is represented by the features after dimension reduction, and the essence is to perform feature extraction and feature selection on the data.
The structure of the product quality prediction model in the industrial processing process provided by the invention is shown in figure 2,
the method comprises data preprocessing, an automatic encoder and an RBF backbone network optimized by using a firefly group intelligent optimization algorithm with a Levy mechanism.
The step S4 specifically includes the following steps:
s41, configuring a model operating environment and setting training parameters;
s42, constructing a product quality prediction model in the industrial processing process: initializing a neuron center, neuron base width, neuron number and connection weight by taking an RBF model as a basic framework;
s43, optimizing the number of the neurons by using a fuzzy clustering FCM mechanism, and when the number of the neurons is optimized by the FCM, updating a target function of the FCM by using a Firefly group optimization algorithm based on a Levy mechanism to a clustering center to finally obtain an optimized RBF model.
S44, inputting the training set into the constructed product quality prediction model in the industrial processing process, and performing model training according to the set training parameters to obtain the trained product quality prediction model in the industrial processing process;
and S45, inputting the test set into the trained product quality prediction model in the industrial processing process, if the accuracy of the output result is greater than or equal to a preset accuracy threshold, the trained product quality prediction model in the industrial processing process is the final prediction model, otherwise, returning to the step S42 to perform model training again.
When the product quality prediction model is trained in a specific industrial processing process, the proposed model can be operated on an open-source Python 3.6, a Tensorflow framework, an Nvidia 1080Ti GPU and a Windows 10.
The step S43 of performing an optimization process on the FCM by using a Firefly cluster optimization algorithm based on a Levy mechanism specifically includes the following steps:
s431, initializing basic parameters of FCM
S432, calculating the brightness of each firefly, see formula (6)
Figure BDA0003070064900000112
Wherein, IiAnd xiRespectively, the brightness and position of firefly, J (x)i) Representing the objective function of firefly.
S433, according to the brightness sequence, updating the position of the firefly, see formula (7)
Figure BDA0003070064900000121
Wherein the content of the first and second substances,
Figure BDA0003070064900000122
position of ith firefly at time t +1, β0The attractive force at the zero distance is,
Figure BDA0003070064900000123
for the relative attraction of the ith and jth fireflies,
Figure BDA0003070064900000124
the position of the ith firefly at time t,
Figure BDA0003070064900000129
to perturb the terms to enhance global search capability, alpha is a scale factor,
Figure BDA00030700649000001210
is a random number vector;
and S434, selecting the brightest firefly as a center, and calculating the FCM objective function value.
Figure BDA0003070064900000125
Where m represents a fuzzy weighting index that controls the degree of classification of the fuzzy matrix. U ═ Uij]c×nIs a membership matrix, uij∈[0,1]Representing the degree of membership that the jth target belongs to the ith cluster. | xi-vj||2Representing the euclidean distance of the jth target from the ith cluster center.
S435, calculating membership function, see formula (9)
Figure BDA0003070064900000126
S436, updating the clustering center, see formula (10)
Figure BDA0003070064900000127
And S437, stopping iteration when the distance between the updated clustering center and the original clustering center is smaller than a given threshold or meets the maximum iteration number, and obtaining the clustering center and the membership matrix.
The model training process in step S44 specifically includes the following steps:
s441, initializing neuron centers, neuron base widths, neuron numbers and connection weights of RBFs;
s442, the reconstruction feature ψ (#) obtained by the self-encoder1,ξ2,…,ξd) As input of a product quality prediction model in an industrial processing process.
S443, with the input of data, the model outputs the corresponding prediction result, and the RMSE loss changes accordingly.
And S444, optimizing the number of the neurons in the RBF by using a fuzzy clustering FCM (fuzzy C-means) mechanism according to the change of the loss function. And the Firefly cluster optimization algorithm based on the Levy mechanism can adjust the FCM clustering center and update the target function J of the FCMm，
Figure BDA0003070064900000128
Where m denotes a fuzzy weight index for controlling the classification degree of the fuzzy matrix, and U ═ Uij]c×nIs a membership matrix, uij∈[0,1]Representing the degree of membership that the jth target belongs to the ith cluster. | xi-vj||2Representing the euclidean distance of the jth target from the ith cluster center.
S445, iteratively optimizing the clustering center c and the clustering number K, thereby determining the neuron center Q, the neuron base width, the neuron number and the connection weight.
And S446, when the root mean square error meets the threshold value, the RMSE loss is minimized, and further the optimal overlapping factor alpha and the optimal neuron number Q are obtained.
And S447, finally obtaining the trained quality prediction model.
And S448, sending the test sample into the trained quality prediction model, and checking whether the performance of the model is optimal.
And after the training of the product quality prediction model in the industrial processing process is completed according to the process, inputting the test set into the trained product quality prediction model in the industrial processing process, if the accuracy of the output result is greater than or equal to the preset accuracy threshold, obtaining the optimal product quality prediction model, and if not, returning to perform model training again.
The present embodiment predicts the product quality data, and the experimental effect is shown in fig. 4. By adopting the method provided by the invention and combining with the newly provided data equalization processing method, the accuracy rate is high, the speed is high, and the method has good approximation capability and global optimal performance in the aspect of predicting the product quality in the factory processing process.
The foregoing detailed description of the preferred embodiments of the invention has been presented. It should be understood that numerous modifications and variations could be devised by those skilled in the art in light of the present teachings without departing from the inventive concepts. Therefore, the technical solutions available to those skilled in the art through logic analysis, reasoning and limited experiments based on the prior art according to the concept of the present invention should be within the scope of protection defined by the claims.

Claims (10)

1. A product quality prediction method of a firefly algorithm optimized RBF model is characterized by comprising the following steps:
obtaining quality data of a product to be tested, loading the data into a product quality prediction model which is constructed in advance and trained, and obtaining the prediction results of qualified products and unqualified products; the product quality prediction model adopts an RBF network with a Gaussian kernel function as a basic framework;
the training process of the product quality prediction model comprises the following steps:
a1: acquiring a product quality data set, and preprocessing the product quality data set;
a2: encoding the preprocessed product quality data set by using an auto-encoder to obtain reconstruction characteristics;
a3: initializing an RBF network, sequentially inputting the reconstruction characteristics into a product quality prediction model, outputting a prediction result by the product quality prediction model in each input process to obtain a model loss function value, optimizing the number of neurons in the RBF network by adopting a fuzzy clustering FCM mechanism according to the loss function value of the model, and updating a clustering center and an objective function value of the fuzzy clustering FCM mechanism by adopting a Firefly group optimization algorithm based on a Levy mechanism;
a4: and repeating the step A3 until a preset training stopping condition is reached, and obtaining a trained product quality prediction model.
2. The firefly algorithm-optimized RBF model-based product quality prediction method according to claim 1, wherein the initializing RBF network comprises initializing neuron centers, neuron base widths, neuron numbers and connection weights of the RBF network;
the optimizing the number of the neurons in the RBF network by adopting a fuzzy clustering FCM mechanism specifically comprises the following steps: and iteratively optimizing the clustering centers and the clustering numbers by adopting a fuzzy clustering FCM mechanism so as to determine the neuron centers, the neuron base widths, the neuron numbers and the connection weights of the RBF network.
3. The Firefly algorithm optimization-based RBF model product quality prediction method according to claim 1, wherein said updating the cluster center and objective function of the fuzzy clustering FCM mechanism by using a Firefly swarm optimization algorithm based on a Levy mechanism specifically comprises the following steps:
a301: initializing basic parameters of the fuzzy clustering FCM mechanism;
a302: calculating the brightness of each firefly;
a303: updating the positions of the fireflies according to the brightness sequence of the fireflies;
a304: selecting the brightest firefly as a center, and calculating an objective function value of a fuzzy clustering FCM mechanism;
a305: calculating a membership function;
a306: updating the clustering center of the fuzzy clustering FCM mechanism according to the membership function;
a307: and repeating the steps A302 to A306 until a preset firefly algorithm stop condition is reached, and obtaining the updated cluster center and membership matrix of the fuzzy cluster FCM mechanism.
4. The firefly algorithm-optimized RBF model-based product quality prediction method according to claim 3, characterized in that the computational expression of the firefly brightness is as follows:
Figure FDA0003070064890000021
in the formula IiIs the ith fireflyLightness of fire, xiIs the position of the ith firefly, J (x)i) Is the target function of the ith firefly, and n is the total number of the fireflies;
the updated expression of the position of the firefly is as follows:
Figure FDA0003070064890000022
in the formula (I), the compound is shown in the specification,
Figure FDA0003070064890000023
position of ith firefly at time t +1, β0The attractive force at the zero distance is,
Figure FDA0003070064890000024
for the relative attraction of the ith and jth fireflies,
Figure FDA0003070064890000025
the position of the ith firefly at time t,
Figure FDA00030700648900000210
to perturb the terms to enhance global search capability, alpha is a scale factor,
Figure FDA0003070064890000029
is a random number vector;
the calculation expression of the objective function value of the fuzzy clustering FCM mechanism is as follows:
Figure FDA0003070064890000026
wherein m is a fuzzy weight index for controlling the classification degree of the fuzzy matrix, JmFor fuzzy clustering of the objective function values of the FCM mechanism under the fuzzy weighting index m, U ═ Uij]c×nIs a membership matrix, uij∈[0,1]Represents the jth eyeDegree of membership, | x, indicating the ith clusteri-vj||2Is the euclidean distance of the jth target from the ith cluster center.
5. The firefly algorithm-optimized RBF model-based product quality prediction method according to claim 4, characterized in that the calculation expression of the membership function is as follows:
Figure FDA0003070064890000027
the updating expression of the clustering center is as follows:
Figure FDA0003070064890000028
6. the firefly algorithm optimized RBF model-based product quality prediction method according to claim 5, wherein the firefly algorithm stopping condition is that the distance between the cluster center of the updated fuzzy clustering FCM mechanism and the original cluster center is smaller than a preset distance threshold, or meets a preset firefly algorithm maximum iteration number.
7. The firefly algorithm optimized RBF model-based product quality prediction method according to claim 1, wherein said preprocessing the product quality data set comprises the steps of:
a11: carrying out standardization and discretization operation on the quality data set of the product;
a12: processing the data set by using an FP-Growth association rule algorithm so as to extract a rule set, and then sequencing according to the support degree and the confidence degree to obtain an ordered rule set;
a13: performing a k-means clustering algorithm on a plurality of types of samples for clustering, and reserving a part of samples in each cluster according to a preset key rule; then, calculating the number of samples needing to be reserved after sampling through a first calculation formula, wherein the key rule is used for obtaining samples ranked in the front, and the majority of samples are certain data samples of which the data number is greater than a preset majority of threshold values;
the expression of the first calculation formula is:
Figure FDA0003070064890000031
in the formula, NiThe number of samples needing to be reserved after sampling in the ith cluster, M is the total number of most classes before sampling, M is the total number of the most classes before samplingiThe total number of samples in the ith cluster;
a14: performing a k-means clustering algorithm on a minority sample for clustering, then generating a new sample in each cluster by using a GAN method, and calculating the number of the new samples in the samples to be generated by using a second calculation formula, wherein the minority sample is a certain type of data sample of which the data number is smaller than a preset minority threshold value;
Figure FDA0003070064890000032
in the formula, niThe number of minority classes required to be generated in the ith cluster, m is the total number of minority classes before sampling, miIs the initial number of samples in the ith cluster.
8. The firefly algorithm optimized RBF model-based product quality prediction method according to claim 1, wherein step A2 specifically comprises the following steps:
a21: in the encoding process of the encoder, input data X ═ X (X)1,x2,…,xn) Is converted into a reconstructed feature ψ ═ (ξ)1,ξ2,…,ξd) The calculation expression of the reconstruction characteristic is as follows:
ψ＝L_ReLU(WeX+be)
in the formula, WeAs weights of the encoder, beIs the basis of the encoder;
a22: during decoding of a decoder, the reconstruction characteristic is restored to X ', and the calculation expression of X' is as follows:
X'＝L_ReLU(Wdψ+bd)
in the formula, WdAs weights of the decoder, bdIs the base of the decoder;
a23: optimizing a loss function loss of the self-encoder by using a gradient descent method, wherein the calculation expression of the loss function loss of the self-encoder is as follows:
Loss＝||X-X'||2
＝||X-f[g(x)]||2
＝||X-L_ReLU(Wd(L_ReLU(WeX+be))+bd)||2
wherein, L _ ReLU (-) is a ReLU activation function;
a24: and repeating the steps A21 to A23 until the preset first coding times are reached, and selecting the corresponding reconstruction characteristics when the loss function loss of the self-coder is minimum as the input of the product quality prediction model.
9. The firefly algorithm optimized RBF model-based product quality prediction method according to claim 1, wherein the training process of the product quality prediction model further comprises: dividing the product quality data set into a training set and a test set, and executing steps A2-A4 by adopting the training set; the training process of the product quality prediction model further comprises the step A5: inputting the test set into a trained product quality prediction model, if the accuracy of the output result is greater than or equal to a preset accuracy threshold, the trained product quality prediction model is the final prediction model, otherwise, returning to the step S2 to perform model training again.
10. A firefly algorithm-optimized RBF model-based product quality prediction device, comprising a memory and a processor, wherein the memory stores a computer program, and the processor invokes the computer program to perform the steps of the method according to any one of claims 1 to 9.
CN202110536540.1A 2021-05-17 2021-05-17 Firefly algorithm optimization-based product quality prediction method and device for RBF model Pending CN113361751A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202110536540.1A CN113361751A (en) 2021-05-17 2021-05-17 Firefly algorithm optimization-based product quality prediction method and device for RBF model

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202110536540.1A CN113361751A (en) 2021-05-17 2021-05-17 Firefly algorithm optimization-based product quality prediction method and device for RBF model

Publications (1)

Publication Number Publication Date
CN113361751A true CN113361751A (en) 2021-09-07

Family

ID=77526440

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202110536540.1A Pending CN113361751A (en) 2021-05-17 2021-05-17 Firefly algorithm optimization-based product quality prediction method and device for RBF model

Country Status (1)

Country Link
CN (1) CN113361751A (en)

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN115994578A (en) * 2022-11-23 2023-04-21 广东工业大学 Correlation method and system based on firefly algorithm
CN117150934A (en) * 2023-10-30 2023-12-01 南京中鑫智电科技有限公司 Comprehensive monitoring system for transformer bushing state and online data processing method thereof

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
KR20100136021A (en) * 2009-06-18 2010-12-28 김영일 A method for tracking signal detection of low voltage lines using fuzzy rbf neural network algorithm
CN106910337A (en) * 2017-01-19 2017-06-30 长安大学 A kind of traffic flow forecasting method based on glowworm swarm algorithm Yu RBF neural
CN107578121A (en) * 2017-08-18 2018-01-12 华北电力大学 Based on the power transformation engineering cost forecasting method for improving glowworm swarm algorithm optimization SVM
CN109671272A (en) * 2018-12-29 2019-04-23 广东工业大学 A kind of freeway traffic flow prediction technique

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
KR20100136021A (en) * 2009-06-18 2010-12-28 김영일 A method for tracking signal detection of low voltage lines using fuzzy rbf neural network algorithm
CN106910337A (en) * 2017-01-19 2017-06-30 长安大学 A kind of traffic flow forecasting method based on glowworm swarm algorithm Yu RBF neural
CN107578121A (en) * 2017-08-18 2018-01-12 华北电力大学 Based on the power transformation engineering cost forecasting method for improving glowworm swarm algorithm optimization SVM
CN109671272A (en) * 2018-12-29 2019-04-23 广东工业大学 A kind of freeway traffic flow prediction technique

Non-Patent Citations (2)

* Cited by examiner, † Cited by third party
Title
刘晓明等: "基于Levy飞行的萤火虫模糊聚类算法", 《计算机应用》 *
张永志: "基于模糊 C 均值聚类的模糊RBF神经网络预测焊接接头力学性能建模", 《机械工程学报》 *

Cited By (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN115994578A (en) * 2022-11-23 2023-04-21 广东工业大学 Correlation method and system based on firefly algorithm
CN115994578B (en) * 2022-11-23 2024-01-26 广东工业大学 Correlation method and system based on firefly algorithm
CN117150934A (en) * 2023-10-30 2023-12-01 南京中鑫智电科技有限公司 Comprehensive monitoring system for transformer bushing state and online data processing method thereof
CN117150934B (en) * 2023-10-30 2024-01-26 南京中鑫智电科技有限公司 Comprehensive monitoring system for transformer bushing state and online data processing method thereof

Similar Documents

Publication Publication Date Title
CN111967343B (en) Detection method based on fusion of simple neural network and extreme gradient lifting model
WO2022068623A1 (en) Model training method and related device
CN113361751A (en) Firefly algorithm optimization-based product quality prediction method and device for RBF model
US20220138504A1 (en) Separation maximization technique for anomaly scores to compare anomaly detection models
CN109191276B (en) P2P network lending institution risk assessment method based on reinforcement learning
Akpudo et al. Towards bearing failure prognostics: A practical comparison between data-driven methods for industrial applications
Zhang et al. Fast covariance matching with fuzzy genetic algorithm
CN113128671B (en) Service demand dynamic prediction method and system based on multi-mode machine learning
Munkhdalai et al. Advanced neural network approach, its explanation with lime for credit scoring application
Bolon-Canedo et al. A unified pipeline for online feature selection and classification
Liu et al. Comparison and evaluation of activation functions in term of gradient instability in deep neural networks
CN114037059A (en) Pre-training model, model generation method, data processing method and data processing device
Li et al. An improved genetic-XGBoost classifier for customer consumption behavior prediction
CN114399101A (en) TCN-BIGRU-based gas load prediction method and device
CN116860529A (en) Fault positioning method and device
ILTER et al. Credit scoring by artificial neural networks based cross-entropy and fuzzy relations
CN113326853A (en) Neural network based process parameter analysis method and equipment and computer storage medium
Koc et al. The impact of feature selection and transformation on machine learning methods in determining the credit scoring
Wang et al. Research on Facial Expression Recognition Based on Improved VGGNet
CN113343429B (en) Method and system for predicting adhesive force quality of inner container in industrial processing process
CN117370870B (en) Knowledge and data compound driven equipment multi-working condition identification and performance prediction method
US20230230141A1 (en) Method and system for recommending products based on collaborative filtering based on natural language programming analogy
US20230334364A1 (en) N-1 experts: model selection for unsupervised anomaly detection
CN113641907B (en) Super-parameter self-adaptive depth recommendation method and device based on evolutionary algorithm
US20230342588A1 (en) Ai system to initiate a concurrent reaction according to predicted user behavior

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
RJ01 Rejection of invention patent application after publication
RJ01 Rejection of invention patent application after publication

Application publication date: 20210907