Technology development data processing method based on artificial intelligence

Abstract

The invention discloses a technology development data processing method based on artificial intelligence, and belongs to the technical field of data processing, and the method comprises the steps: obtaining a technology development data processing task input by a technician, taking the verified technology development data processing task as the basis, and carrying out the task learning analysis through employing an artificial intelligence algorithm, thereby achieving the technical development data processing. According to the technical development data processing method and device, the technical development data processing rule is determined, and finally the real-time technical development data can be processed through the technical development data processing rule, so that the processing efficiency of the technical development data can be effectively improved, and the problem that errors are likely to occur due to manual processing of the technical development data can be effectively solved.

Description

Technical Field

[0001] The present invention belongs to the technical field of data processing, and particularly relates to a data processing method for technology development based on artificial intelligence. Background Art

[0002] With the rapid development of information technology, the big data era has arrived. How to extract valuable information from massive data has become an urgent problem to be solved. Technology development data refers to various types of data generated during the process of technology research and development. These data are of great significance for technological progress, product innovation, and process optimization. In the prior art, often by the staff themselves to analyze the technology development data or wait for the failure to occur, and then analyze the technology development data. The manual processing of technology development data has problems such as low efficiency, low accuracy, and easy errors. Summary of the Invention

[0003] The present invention provides a data processing method for technology development based on artificial intelligence to solve the problems existing in the prior art.

[0004] A data processing method for technology development based on artificial intelligence includes: Obtain a technology development data processing task input by a technician, and verify the technology development data processing task to obtain a verified technology development data processing task; Based on the verified technology development data processing task, use an artificial intelligence algorithm to perform task learning analysis to determine technology development data processing rules; Collect real-time technology development data, and process the real-time technology development data through the technology development data processing rules to obtain processed technology development data.

[0005] Further, obtaining a technology development data processing task input by a technician includes: obtaining a batch of technology development data input by a technician and data labels corresponding to each technology development data to obtain a technology development data processing task.

[0006] Further, verifying the technology development data processing task to obtain a verified technology development data processing task includes: For each type of technology development data in the technology development data processing task, verify whether it is numerical data. If so, perform data length verification; otherwise, convert the text data into numerical data and then perform data length verification; For each type of technology development data in the technology development data processing task, verify whether the lengths are the same. If so, obtain a verified technology development data processing task; otherwise, perform length normalization processing to obtain a verified technology development data processing task.

[0007] Further, based on the verified technical development data processing tasks, artificial intelligence algorithms are used for task learning and analysis to determine technical development data processing rules, including: Initialize the hyperparameters of the artificial intelligence algorithm using a uniform initial strategy to obtain multiple hyperparameter encodings corresponding to the artificial intelligence algorithm; Based on the verified technical development data processing tasks, obtain the fitness corresponding to each hyperparameter encoding, and divide the multiple hyperparameter encodings into optimal encodings, elite encodings, general encodings, and inferior encodings according to the fitness of each hyperparameter encoding; For the optimal encoding, perform global greedy search on the optimal encoding using a global mutation optimization strategy to obtain the optimal encoding after global greedy search; For the elite encoding, perform random local search on the elite encoding using an adaptive fluctuation optimization strategy to obtain the elite encoding after random local search; For the general encoding, perform joint local search on the general encoding using a multi-encoding cooperation optimization strategy to obtain the general encoding after joint local search; For the inferior encoding, perform fast-guided search on the inferior encoding using an adaptive sine fast guidance strategy to obtain the inferior encoding after fast-guided search; Re-fuse the optimal encoding after global greedy search, the elite encoding after random local search, the general encoding after joint local search, and the inferior encoding after fast-guided search into a population; Judge whether the current learning times are greater than or equal to the maximum learning times. If so, obtain the final hyperparameters corresponding to the artificial intelligence algorithm according to the re-fused population, and use the artificial intelligence algorithm after applying the final hyperparameters as the technical development data processing rules. Otherwise, based on the re-fused population, return to the step of hyperparameter encoding division.

[0008] Further, initialize the hyperparameters of the artificial intelligence algorithm using a uniform initial strategy to obtain multiple hyperparameter encodings corresponding to the artificial intelligence algorithm, including: For the hyperparameters of the artificial intelligence algorithm, perform random initialization within the upper and lower limits of the hyperparameters, and encode the initialized hyperparameters as vectors to obtain the basic encoding; Use the basic encoding as the first hyperparameter encoding to obtain N hyperparameter encodings as: ; where, represents the i th hyperparameter encoding, represents the first constant coefficient, and is set to 0.5; represents the second constant coefficient, and Set to 0.2; represents pi, represents the remainder function, represents the sine function, represents the i +(M - 1) hyperparameter encodings, i m = 1, 2, …, M - 1.

[0009] Furthermore, based on the verified technology development data processing tasks, obtain the fitness corresponding to each hyperparameter encoding, and divide the multiple hyperparameter encodings into optimal encodings, elite encodings, general encodings, and inferior encodings according to the fitness of each hyperparameter encoding, including: Using the technology development data in the verified technology development data processing task as input data and the data labels corresponding to the technology development data in the verified technology development data processing task as the expected output, obtain the error function value corresponding to the hyperparameter encoding; Take the negative of the error function value corresponding to the hyperparameter encoding to obtain the fitness corresponding to the hyperparameter encoding; Take the hyperparameter encoding with the maximum fitness as the optimal encoding, and arrange the remaining hyperparameter encodings in descending order of fitness to obtain the hyperparameter encoding sequence; Use a 2:6:2 ratio to sequentially divide the hyperparameter encoding sequence into elite encodings, general encodings, and inferior encodings.

[0010] Furthermore, for the optimal encoding, adopt a global mutation optimization strategy to perform global greedy search on the optimal encoding, and obtain the optimal encoding after global greedy search, including: Based on the current learning times, obtain the global mutation probability as: ; where represents the global mutation probability, represents the second type of Euler integral function, represents pi, represents the mutation probability control constant, represents the current learning times; For the optimal encoding, based on the global mutation probability, use the roulette wheel method to determine the mutation operation selection result corresponding to the optimal encoding; among them, the mutation operation selection result includes whether to mutate or not; When the mutation operation selection result is not to mutate, directly use the original optimal encoding as the optimal encoding after global greedy search; When the mutation operation selection result is to mutate, mutate the optimal encoding as: ; where represents the t optimal encoding in the m-th learning process, Denote the optimal code after mutation , Denote the first random number between (0, 1); Judge whether the fitness of the optimal code after mutation increases. If so, use the optimal code after mutation as the optimal code after global greedy search. Otherwise, directly use the original optimal code as the optimal code after global greedy search.

[0011] Furthermore, for the elite code, adopt an adaptive fluctuation optimization strategy to perform random local search on the elite code to obtain the elite code after random local search, including:[[]] Based on the current learning times, use the hyperbolic tangent function to obtain the adaptive fluctuation factor as:[[]] ; where Denote the adaptive fluctuation factor,[[]] Denote the cosine function,[[]] Denote the hyperbolic tangent function,[[]] Denote the preset maximum learning times,[[]] Denote the pi;[[]] Perform random local search on the elite code according to the adaptive fluctuation factor to obtain the elite code after random local search as:[[]] ; where Denote the t th learning process and the j th elite code,[[]] j = 1, 2, …, N1, N1 denotes the total number of elite codes,[[]] Denote the second random number between (0, 1),[[]] Denote the elite code after random local search[[]] , e denotes the natural constant, sin denotes the sine function, and cos denotes the cosine function.[[]]

[0012] Furthermore, for the general code, adopt a multi-code cooperation optimization strategy to perform joint local search on the general code to obtain the general code after joint local search, including:[[]] For any general code, randomly match an elite code to the general code to obtain the target elite code corresponding to the general code;[[]] Obtain the Euclidean distance between the general code and its corresponding target elite code, and obtain the random movement factor according to the Euclidean distance as:[[]] ; where Denote the random movement factor generated for the d th dimensional parameter of the general code,[[]] Denote the preset maximum learning times,[[]] Denote the random number between (-1, 1),[[]] Denote them The Euclidean distance between a general code and its corresponding target elite code d = 1, 2, …, L, where L represents the total dimension of the parameters; Based on the optimal code, the random movement factor, and the target elite code, perform a joint local search on the general code, and the general code after the joint local search is: ; where represents the t th parameter of the m th general code in the d th learning process, represents the m th parameter of the general code after the d th joint local search, represents the first learning factor, represents the second learning factor, represents the third random number between (0, 1), represents the fifth random number between (0, 1), represents the t th parameter of the target elite code corresponding to the m th general code in the d th learning process, m = 1, 2, …, N2, where N2 represents the total number of general codes.

[0013] Furthermore, for the inferior code, adopt an adaptive sine fast guidance strategy to perform a fast guidance search on the inferior code, and the inferior code after the fast guidance search includes: Generate an adaptive inertia weight and an adaptive step size search factor as: and ; where represents the adaptive inertia weight, represents the exponential function with the natural constant e as the base, represents the preset maximum number of learning times, represents the adaptive step size search factor, represents the constant factor; Based on the adaptive inertia weight and the adaptive step size search factor, perform a fast guidance search on the inferior code, and the inferior code after the fast guidance search is: ; where represents the t th inferior code in the n th learning process, represents the inferior code after the fast guidance search , represents (0, 2 the sixth random number between ( represents pi, represents the seventh random number between (0, 1).

[0014] A data processing method for technology development based on artificial intelligence provided by the present invention, by obtaining the technology development data processing task input by the technician, and then based on the verified technology development data processing task, using artificial intelligence algorithms to perform task learning and analysis to determine the technology development data processing rules. Finally, the real-time technology development data can be processed through the technology development data processing rules, which can not only effectively improve the processing efficiency of technology development data, but also effectively reduce the problem of error-proneness caused by manual processing of technology development data. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] The accompanying drawings herein are incorporated into and constitute a part of this specification, showing embodiments consistent with the present invention and used together with the specification to explain the principles of the present invention.

[0016] Figure 1 It is a flowchart of a data processing method for technology development based on artificial intelligence provided by an embodiment of the present invention.

[0017] Figure 2 It is a flowchart of determining the technology development data processing rules provided by an embodiment of the present invention.

[0018] Through the above-mentioned accompanying drawings, specific embodiments of the present invention have been shown, and there will be more detailed descriptions hereinafter. These drawings and written descriptions are not intended to limit the scope of the inventive concept in any way, but to illustrate the concept of the present invention to those skilled in the art by reference to specific embodiments. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0019] Here, the exemplary embodiments will be described in detail, and the examples are shown in the accompanying drawings. When the following description refers to the accompanying drawings, unless otherwise indicated, the same numbers in different drawings represent the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with the present invention. On the contrary, they are merely examples of devices and methods consistent with some aspects of the present invention as detailed in the appended claims.

[0020] The embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0021] As Figure 1 shown, an embodiment of the present invention provides a data processing method for technology development based on artificial intelligence, including: S11. Obtain the technology development data processing task input by the technician, and verify the technology development data processing task to obtain the verified technology development data processing task; For example, the technology development data is the design of the parameters of an electric motor. When it is necessary to simulate its operating temperature, operating speed, etc. based on the motor parameters, modeling processing is usually required, which requires technicians to have relatively high professional knowledge to achieve. However, through artificial intelligence technology, rapid prediction can be realized. When the staff adjusts the parameters to a relatively reasonable value and then conducts fine testing, the technology development efficiency can be greatly improved. Or, the technology development data is the test data of a software (which can include input parameters, software control parameters, output parameters, etc.). When it is necessary to classify the technology development data, the rapid classification of the test data can be realized through artificial intelligence technology, thereby improving the storage and search efficiency of the technology development data.

[0022] S12. Based on the technology development data processing task after verification, use an artificial intelligence algorithm to perform task learning analysis to determine the technology development data processing rules; By learning the correlation between the data and labels in the technology development data processing task, it can effectively assist technicians in quickly classifying the technology development data, thereby improving the processing efficiency of the technology development data.

[0023] S13. Collect real-time technology development data, and process the real-time technology development data through the technology development data processing rules to obtain the processed technology development data.

[0024] Another example is that when developing a software for predicting the load of an electricity consumption area, the technology development data can be the influencing factors of electricity consumption. It is necessary to process these influencing factors of electricity consumption. At this time, the corresponding expected output is the actual electricity consumption corresponding to the influencing factors of electricity consumption. After learning these data, load prediction can be realized and technology development can be achieved.

[0025] A method for processing technology development data based on artificial intelligence provided by the present invention, by obtaining the technology development data processing task input by a technician, then based on the technology development data processing task after verification, using an artificial intelligence algorithm to perform task learning analysis to determine the technology development data processing rules, and finally the real-time technology development data can be processed through the technology development data processing rules, which can not only effectively improve the processing efficiency of the technology development data, but also effectively reduce the problem of error-proneness caused by manual processing of the technology development data.

[0026] In the embodiment of the present invention, obtaining the technology development data processing task input by a technician includes: obtaining the batch of technology development data input by the technician and the data label corresponding to each technology development data to obtain the technology development data processing task.

[0027] For example, the primary development data may include development data B1, development data B2, and development data B3, and the corresponding label is C. Then the technical development data processing task can be: input the technical development data [B1, B2, B3], and the expected output data should be C.

[0028] In an embodiment of the present invention, the technical development data processing task is verified to obtain the verified technical development data processing task, including: For each type of technical development data in the technical development data processing task, verify whether it is numerical data. If so, perform data length verification; otherwise, convert the text data into numerical data and then perform data length verification. For each type of technical development data in the technical development data processing task, verify whether the lengths are the same. If so, obtain the verified technical development data processing task; otherwise, perform length normalization processing (such as padding zeros in front of the data) to obtain the verified technical development data processing task.

[0029] As Figure 2 shown, based on the verified technical development data processing task, an artificial intelligence algorithm is used for task learning and analysis to determine the technical development data processing rules, including: S21. Initialize the hyperparameters of the artificial intelligence algorithm using a uniform initial strategy to obtain multiple hyperparameter encodings corresponding to the artificial intelligence algorithm. S22. Based on the verified technical development data processing task, obtain the fitness corresponding to each hyperparameter encoding, and divide the multiple hyperparameter encodings into optimal encodings, elite encodings, general encodings, and inferior encodings according to the fitness of each hyperparameter encoding. S23. For the optimal encoding, perform global greedy search on the optimal encoding using a global mutation optimization strategy to obtain the optimal encoding after global greedy search. S24. For the elite encoding, perform random local search on the elite encoding using an adaptive fluctuation optimization strategy to obtain the elite encoding after random local search. S25. For the general encoding, perform joint local search on the general encoding using a multi-encoding cooperation optimization strategy to obtain the general encoding after joint local search. S26. For the inferior encoding, perform fast guiding search on the inferior encoding using an adaptive sine fast guiding strategy to obtain the inferior encoding after fast guiding search. S27. Re-fuse the optimal encoding after global greedy search, the elite encoding after random local search, the general encoding after joint local search, and the inferior encoding after fast guiding search into a population. S28. Determine whether the current number of learning times is greater than or equal to the maximum number of learning times. If so, obtain the final hyperparameters corresponding to the artificial intelligence algorithm based on the re - fused population, and use the artificial intelligence algorithm after applying the final hyperparameters as the technical development data processing rule. Otherwise, based on the re - fused population, return to the step of hyperparameter coding division.

[0030] When using artificial intelligence algorithms for data classification or data prediction currently, the gradient descent algorithm is often used to optimize the hyperparameters corresponding to the artificial intelligence algorithm. It is usually easy to fall into local optima, resulting in poor data classification or data prediction effects. For technical development data, if the classification is incorrect, it is very likely to cause errors in the subsequent test and optimization process. Therefore, the embodiments of the present invention provide an improved hyperparameter optimization algorithm to improve the accuracy of technical development data.

[0031] In the embodiments of the present invention, a uniform initial strategy is used to initialize the hyperparameters of the artificial intelligence algorithm, and multiple hyperparameter encodings corresponding to the artificial intelligence algorithm are obtained, including: For the hyperparameters of the artificial intelligence algorithm, randomly initialize within the upper and lower limits of the hyperparameters, and encode the initialized hyperparameters as vectors to obtain the basic encoding; Taking the basic encoding as the first hyperparameter encoding, N hyperparameter encodings are obtained as: ; where represents the i th hyperparameter encoding, represents the first constant coefficient, and is set to 0.5; represents the second constant coefficient, and is set to 0.2; represents pi, represents the remainder function, represents the sine function, represents the i +1 th hyperparameter encoding, i = 1, 2, …, M - 1.

[0032] The uniform initial strategy provided by the embodiments of the present invention can make the hyperparameter encodings more evenly distributed in the initial solution space, thus effectively improving the optimization speed of the algorithm.

[0033] In the embodiments of the present invention, based on the verified technical development data processing tasks, obtain the fitness corresponding to each hyperparameter encoding, and divide the multiple hyperparameter encodings into optimal encodings, elite encodings, general encodings, and inferior encodings according to the fitness of each hyperparameter encoding, including: Using the technology development data in the technology development data processing task after verification as the input data, and using the data label corresponding to the technology development data in the technology development data processing task after verification as the expected output, obtain the error function value corresponding to the hyperparameter encoding; Take the negative of the error function value corresponding to the hyperparameter encoding to obtain the fitness corresponding to the hyperparameter encoding; Take the hyperparameter encoding with the maximum fitness as the optimal encoding, and arrange the remaining hyperparameter encodings in descending order of fitness to obtain a hyperparameter encoding sequence; Divide the hyperparameter encoding sequence into elite encodings, general encodings, and inferior encodings in a ratio of 2:6:2. To ensure integer division, the number of hyperparameter encodings should be a multiple of 10 plus one.

[0034] In the embodiment of the present invention, for the optimal encoding, a global mutation optimization strategy is used to perform a global greedy search on the optimal encoding, and the optimal encoding after the global greedy search is obtained, including: Based on the current learning times, obtain the global mutation probability as: ; where, represents the global mutation probability, represents the second type of Euler integral function, represents pi, represents the mutation probability control constant, represents the current learning times; For the optimal encoding, based on the global mutation probability, use the roulette wheel method to determine the mutation operation selection result corresponding to the optimal encoding; among them, the mutation operation selection result includes the need to mutate or not to mutate; When the mutation operation selection result is not to mutate, then directly use the original optimal encoding as the optimal encoding after the global greedy search; When the mutation operation selection result is to mutate, then mutate the optimal encoding as: ; where, represents the optimal encoding in the t th learning process, represents the optimal encoding after mutation , represents the first random number between (0, 1); Judge whether the fitness of the optimal encoding after mutation increases. If so, use the optimal encoding after mutation as the optimal encoding after the global greedy search, otherwise directly use the original optimal encoding as the optimal encoding after the global greedy search.

[0035] The technical development data provided by the embodiments of the present invention can provide a relatively large mutation probability in the early and middle stages of the algorithm, and the decline speed is slow. After the algorithm reaches the later stage, the mutation probability drops rapidly, thereby improving the convergence accuracy of the algorithm. At the same time, combined with greedy search, the convergence speed of the algorithm is further ensured.

[0036] In the embodiments of the present invention, for elite coding, an adaptive fluctuation optimization strategy is adopted to perform random local search on the elite coding to obtain the elite coding after random local search, including: Based on the current learning times, the hyperbolic tangent function is used to obtain the adaptive fluctuation factor as: ; where represents the adaptive fluctuation factor, represents the cosine function, represents the hyperbolic tangent function, represents the preset maximum learning times, represents the pi; According to the adaptive fluctuation factor, random local search is performed on the elite coding, and the elite coding after random local search is: ; where represents the t th learning process and the j th elite coding, j = 1, 2,..., N1, N1 represents the total number of elite codings, represents the second random number between (0, 1), represents the elite coding after random local search , e represents the natural constant, sin represents the sine function, and cos represents the cosine function.

[0037] The adaptive fluctuation optimization strategy provided by the embodiments of the present invention can make the elite coding fluctuate and search around in different directions and at different speeds, and can effectively search in the relatively optimal area, improving the possibility of the algorithm to search for the optimal solution.

[0038] In the embodiments of the present invention, for general coding, a multi-coding cooperation optimization strategy is adopted to perform joint local search on the general coding to obtain the general coding after joint local search, including: For any general coding, a target elite coding corresponding to the general coding is randomly matched for the general coding; The Euclidean distance between the general coding and its corresponding target elite coding is obtained, and the random movement factor is obtained according to the Euclidean distance as: ; where represents the random movement factor generated for the d th dimension parameter of the general coding, represents the preset maximum number of learning times, represents a random number between (-1, 1), represents the m Euclidean distance between the d -th general encoding and its corresponding target elite encoding, where = 1, 2, …, L, and L represents the total dimension of the parameters; ; where, represents the t -th parameter of the m -th general encoding in the d -th learning process, represents the m -th parameter of the d -th general encoding after joint local search, represents the first learning factor, represents the second learning factor, represents a third random number between (0, 1), represents a fifth random number between (0, 1), represents the t -th parameter of the target elite encoding corresponding to the m -th general encoding in the d -th learning process, where m = 1, 2, …, N2, and N2 represents the total number of general encodings.

[0039] The multi-encoding collaborative optimization strategy provided by the embodiments of the present invention can effectively fuse encoding information and has a certain degree of perturbation, which can more effectively search the unfamiliar areas between encoding individuals, improve the traversal of the solution space, and enhance the search ability of the algorithm.

[0040] In the embodiments of the present invention, for inferior encodings, an adaptive sine fast guidance strategy is adopted to perform fast guidance search on the inferior encodings, and the inferior encodings after fast guidance search are obtained, including: Generate an adaptive inertia weight and an adaptive step size search factor as: and ; where, represents the adaptive inertia weight, represents the exponential function with the natural constant e as the base, represents the preset maximum number of learning times, represents the adaptive step size search factor, represents a constant factor; According to the adaptive inertia weight and the adaptive step size search factor, perform a fast-guided search on the inferior coding, and the inferior coding after the fast-guided search is obtained as follows: ; where represents the t th inferior coding in the n th learning process, represents the inferior coding after the fast-guided search , represents the sixth random number between (0, 2 ), represents pi, represents the seventh random number between (0, 1).

[0041] The adaptive sine fast-guided strategy provided by the embodiment of the present invention advances rapidly towards the optimal region in a sine waveform, which can not only effectively improve the search ability of the algorithm, but also improve the global search ability.

[0042] Optionally, after each update of the coding, the coding can be processed for out-of-bounds to ensure the validity of the parameters.

[0043] After considering the specification and the practice of the invention disclosed herein, those skilled in the art will readily conceive of other embodiments of the present invention. The present invention is intended to cover any variations, uses, or adaptations of the present invention, which follow the general principles of the present invention and include the common general knowledge or conventional technical means in the technical field not disclosed by the present invention. It should be understood that the present invention is not limited to the exact structures described above and shown in the drawings, and various modifications and changes can be made without departing from its scope. The scope of the present invention is only limited by the appended claims.

Claims

1. A data processing method for technology development based on artificial intelligence, characterized in that, Including: Obtain the technical development data processing task input by the technical personnel, and verify the technical development data processing task to obtain the verified technical development data processing task; Based on the verified technical development data processing task, use artificial intelligence algorithms to perform task learning and analysis to determine the technical development data processing rules; Collect real-time technical development data, and process the real-time technical development data through the technical development data processing rules to obtain the processed technical development data.

2. The data processing method for technology development based on artificial intelligence according to claim 1, wherein Obtain the technical development data processing task input by the technical personnel, including: obtaining the batch of technical development data input by the technical personnel and the data label corresponding to each technical development data to obtain the technical development data processing task.

3. The data processing method for technology development based on artificial intelligence according to claim 1, characterized in that, Verify the technical development data processing task to obtain the verified technical development data processing task, including: For each type of technical development data in the technical development data processing task, verify whether it is numerical data. If so, perform data length verification. Otherwise, convert the text data into numerical data and then perform data length verification; For each type of technical development data in the technical development data processing task, verify whether the lengths are the same. If so, obtain the verified technical development data processing task. Otherwise, perform length normalization to obtain the verified technical development data processing task.

4. The data processing method for technology development based on artificial intelligence according to claim 1, wherein Based on the verified technical development data processing task, use artificial intelligence algorithms to perform task learning and analysis to determine the technical development data processing rules, including: Initialize the hyperparameters of the artificial intelligence algorithm using a uniform initial strategy to obtain multiple hyperparameter encodings corresponding to the artificial intelligence algorithm; Based on the verified technical development data processing task, obtain the fitness corresponding to each hyperparameter encoding, and divide the multiple hyperparameter encodings into optimal encodings, elite encodings, general encodings, and inferior encodings according to the fitness of each hyperparameter encoding; For the optimal encoding, perform global greedy search on the optimal encoding using a global mutation optimization strategy to obtain the optimal encoding after global greedy search; For the elite encoding, perform random local search on the elite encoding using an adaptive fluctuation optimization strategy to obtain the elite encoding after random local search; For the general encoding, perform joint local search on the general encoding using a multi-encoding cooperation optimization strategy to obtain the general encoding after joint local search; For the inferior encoding, perform fast-guided search on the inferior encoding using an adaptive sine fast guidance strategy to obtain the inferior encoding after fast-guided search; Re-fuse the optimal encoding after global greedy search, the elite encoding after random local search, the general encoding after joint local search, and the inferior encoding after fast-guided search into a population; Judge whether the current learning times are greater than or equal to the maximum learning times. If so, obtain the final hyperparameters corresponding to the artificial intelligence algorithm according to the re-fused population, and use the artificial intelligence algorithm after applying the final hyperparameters as the technical development data processing rules. Otherwise, based on the re-fused population, return to the step of hyperparameter encoding division.

5. The data processing method for technology development based on artificial intelligence according to claim 4, wherein Initialize the hyperparameters of the artificial intelligence algorithm using a uniform initial strategy to obtain multiple hyperparameter encodings corresponding to the artificial intelligence algorithm, including: For the hyperparameters of the artificial intelligence algorithm, perform random initialization within the upper and lower limits of the hyperparameters, and encode the initialized hyperparameters as vectors to obtain the basic encoding; Using the said basic encoding as the first hyperparameter encoding, obtain N hyperparameter encodings as follows: ; where represents the i -th hyperparameter encoding, represents the first constant coefficient, and is set to 0.5; represents the second constant coefficient, and is set to 0.2; represents pi, represents the modulo function, represents the sine function, represents the i +1 -th hyperparameter encoding, i = 1, 2, …, M - 1.

6. The data processing method for technology development based on artificial intelligence according to claim 4, wherein Based on the verified technology development data processing task, obtain the fitness corresponding to each hyperparameter encoding, and divide the multiple hyperparameter encodings into optimal encoding, elite encoding, general encoding, and inferior encoding according to the fitness of each hyperparameter encoding, including: Using the technology development data in the verified technology development data processing task as the input data, and the data label corresponding to the technology development data in the verified technology development data processing task as the expected output, obtain the error function value corresponding to the hyperparameter encoding; Take the negative of the error function value corresponding to the hyperparameter encoding to obtain the fitness corresponding to the hyperparameter encoding; Take the hyperparameter encoding with the maximum fitness as the optimal encoding, and arrange the remaining hyperparameter encodings in descending order of fitness to obtain the hyperparameter encoding sequence; Divide the hyperparameter encoding sequence into elite encoding, general encoding, and inferior encoding in the ratio of 2:6:2 in sequence.

7. The data processing method for technology development based on artificial intelligence according to claim 4, characterized in that For the optimal encoding, use the global mutation optimization strategy to perform global greedy search on the optimal encoding to obtain the optimal encoding after global greedy search, including: Based on the current number of learning times, the global mutation probability is obtained as follows: ; where represents the global mutation probability, represents the second type of Euler integral function, represents the pi, represents the mutation probability control constant, represents the current number of learning times; For the optimal encoding, based on the global mutation probability, use the roulette wheel method to determine the mutation operation selection result corresponding to the optimal encoding; among them, the mutation operation selection result includes the need to mutate or not to mutate; When the mutation operation selection result is not to mutate, directly use the original optimal encoding as the optimal encoding after global greedy search; When the mutation operation selection result is that mutation is required, the optimal encoding is mutated as follows: ; where represents the optimal encoding in the t th learning process, represents the optimal encoding after mutation , represents the first random number between (0, 1); Judge whether the fitness of the mutated optimal encoding increases. If so, use the mutated optimal encoding as the optimal encoding after global greedy search, otherwise directly use the original optimal encoding as the optimal encoding after global greedy search.

8. The data processing method for technology development based on artificial intelligence according to claim 7, wherein For the elite encoding, use the adaptive fluctuation optimization strategy to perform random local search on the elite encoding to obtain the elite encoding after random local search, including: Based on the current number of learning times, the hyperbolic tangent function is used to obtain the adaptive fluctuation factor as follows: ; where represents the adaptive fluctuation factor, represents the cosine function, represents the hyperbolic tangent function, represents the preset maximum number of learning times, represents the pi; Perform random local search on the elite encoding according to the adaptive fluctuation factor, and the elite encoding after random local search is obtained as follows: ; where represents the t th learning process and the j th elite encoding, j = 1, 2, …, N1, where N1 represents the total number of elite encodings, represents the second random number between (0, 1), represents the elite encoding after random local search , e represents the natural constant, sin represents the sine function, and cos represents the cosine function.

9. The data processing method for technology development based on artificial intelligence according to claim 8, wherein For the general encoding, use the multi-encoding cooperation optimization strategy to perform joint local search on the general encoding to obtain the general encoding after joint local search, including: For any general encoding, randomly match an elite encoding to the general encoding to obtain the target elite encoding corresponding to the general encoding; Obtain the Euclidean distance between the general encoding and its corresponding target elite encoding, and obtain the random movement factor according to the Euclidean distance as: ; where represents the random movement factor generated by the d -th dimensional parameter of the general encoding, represents the preset maximum number of learning times, represents a random number between (-1, 1), represents the m -th Euclidean distance between the general encoding and its corresponding target elite encoding, d = 1, 2, …, L, where L represents the total dimension of the parameters; Perform joint local search on the general encoding based on the optimal encoding, random movement factor, and target elite encoding. The general encoding after joint local search is as follows: ; where represents the t -th dimension parameter of the m -th general encoding in the d -th learning process, represents the m -th dimension parameter of the d -th general encoding after joint local search, represents the first learning factor, represents the second learning factor, represents the third random number between (0, 1), represents the fifth random number between (0, 1), represents the t -th dimension parameter of the target elite encoding corresponding to the m -th general encoding in the d -th learning process, m = 1, 2, …, N2, where N2 represents the total number of general encodings.

10. The data processing method for technology development based on artificial intelligence according to claim 9, characterized in that, For the inferior encoding, use the adaptive sine fast guidance strategy to perform fast guidance search on the inferior encoding to obtain the inferior encoding after fast guidance search, including: The generation of the adaptive inertia weight and the adaptive step-size search factor is as follows: and ; where represents the adaptive inertia weight, represents the exponential function with the natural constant e as the base, represents the preset maximum number of learning times, represents the adaptive step-size search factor, represents the constant factor; According to the adaptive inertia weight and the adaptive step size search factor, perform a fast-guided search on the inferior coding, and the inferior coding after the fast-guided search is: ; among them, represents the t th inferior coding in the n th learning process, represents the inferior coding after the fast-guided search , represents the sixth random number between (0, 2 ), represents pi, represents the seventh random number between (0, 1).

Images