Planning data processing method, device, equipment and medium based on artificial intelligence

Abstract

The present invention is applicable to the field of artificial intelligence, and in particular to a method, device, equipment and medium for processing plan data based on artificial intelligence. The present invention obtains N plan subvectors with a time series relationship by inputting plan data and a preset initial subvector into a trained plan generation model, and inputs the ith plan subvector and the i+1th plan subvector into a trained probability prediction model to obtain a first probability that the ith plan subvector and the i+1th plan subvector meet the time series relationship, thereby obtaining N-1 first probabilities, judging whether the N-1 first probabilities meet preset conditions, determining N target plan subvectors, and splicing the N target plan subvectors to obtain a target plan vector, and finally determining a target plan text, and obtaining N plan subvectors with a time series relationship according to the plan data, thereby improving the time series of the final complete target plan text and improving the feasibility of the target plan text.

Description

Technical Field

[0001] The present invention is applicable to the field of artificial intelligence technology, and in particular to a planning data processing method, device, equipment and medium based on artificial intelligence. Background Art

[0002] With the development of deep learning technology, the existing technology uses deep neural networks to learn the alignment relationship between plan data and plan texts to achieve the purpose of directly generating complete plan texts based on plan data. When faced with complex and scattered plan data and target plan texts, the internal timing of the complete plan texts directly generated by deep neural networks is poor, which greatly reduces the feasibility of the plan texts.

[0003] Therefore, in the field of artificial intelligence, how to improve the feasibility of complex planning texts generated based on complex and scattered planning data has become an urgent problem to be solved. Summary of the invention

[0004] In view of this, embodiments of the present invention provide a method, device, equipment and medium for processing planning data based on artificial intelligence to solve the problem of poor feasibility of complex planning texts generated by existing deep neural networks based on complex and scattered planning data.

[0005] In a first aspect, an embodiment of the present invention provides a method for processing planning data based on artificial intelligence, the method comprising:

[0006] Acquire planning data, input the planning data and a preset initial sub-vector into a trained planning generation model, and obtain N planning sub-vectors with a time series relationship, wherein the planning data includes personal information of the personnel to be trained and preset training business information, and N is an integer greater than 1;

[0007] Among the N plan sub-vectors, the i-th plan sub-vector and the i+1-th plan sub-vector are sequentially input into the trained probability prediction model to obtain the first probabilities that the N-1 i-th plan sub-vectors and the i+1-th plan sub-vectors conform to the temporal relationship, where i=1, 2, ..., N-1;

[0008] Determine whether N-1 of the first probabilities meet a preset condition, and if N-1 of the first probabilities meet the preset condition, use the N plan sub-vectors as N target plan sub-vectors;

[0009] The N target plan sub-vectors are spliced, a splicing result is determined as a target plan vector, and a target plan text is determined according to the target plan vector.

[0010] In a second aspect, an embodiment of the present invention provides a planning data processing device based on artificial intelligence, the planning data processing device comprising:

[0011] A plan sub-vector generation module is used to obtain plan data, input the plan data and the preset initial sub-vector into the trained plan generation model, and obtain N plan sub-vectors with a time series relationship, wherein the plan data includes the personal information of the personnel to be trained and the preset training business information, and N is an integer greater than 1;

[0012] A first probability prediction module is used to input the ith plan subvector and the i+1th plan subvector in the N plan subvectors into the trained probability prediction model in sequence to obtain a first probability that the N-1 ith plan subvectors and the i+1th plan subvectors conform to the temporal relationship, where i=1, 2, ..., N-1;

[0013] a target plan sub-vector determining module, configured to determine whether N-1 of the first probabilities meet a preset condition, and if N-1 of the first probabilities meet the preset condition, use the N plan sub-vectors as N target plan sub-vectors;

[0014] The target plan text determination module is used to splice the N target plan sub-vectors, determine the splicing result as the target plan vector, and determine the target plan text according to the target plan vector.

[0015] In a third aspect, an embodiment of the present invention provides a computer device, comprising a processor, a memory, and a computer program stored in the memory and executable on the processor, wherein the processor implements the planning data processing method as described in the first aspect when executing the computer program.

[0016] In a fourth aspect, an embodiment of the present invention provides a computer-readable storage medium, wherein the computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the planning data processing method as described in the first aspect is implemented.

[0017] Compared with the prior art, the embodiments of the present invention have the following beneficial effects: by inputting plan data and a preset initial sub-vector into a trained plan generation model, N plan sub-vectors having a time series relationship are obtained, and the i-th plan sub-vector and the i+1-th plan sub-vector are input into a trained probability prediction model, and the first probability that the i-th plan sub-vector and the i+1-th plan sub-vector meet the time series relationship is obtained, thereby obtaining N-1 first probabilities of the N plan sub-vectors, and by judging whether the N-1 first probabilities meet preset conditions, N target plan sub-vectors are determined, and the N target plan sub-vectors are spliced ​​to determine the splicing result as the target plan vector, and finally the target plan text is determined according to the target plan vector, and N plan sub-vectors having a time series relationship are obtained according to the plan data, thereby improving the time series of the complete target plan text finally obtained and improving the feasibility of the target plan text. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.

[0019] Figure 1 This is a schematic diagram of an application environment of a planning data processing method provided by the first embodiment of the present invention;

[0020] Figure 2 It is a flowchart of a planning data processing method provided in the first embodiment of the present invention;

[0021] Figure 3 It is a structural diagram of a planning data processing device provided in Embodiment 2 of the present invention;

[0022] Figure 4 It is a structural diagram of a computer device provided in Embodiment 3 of the present invention. DETAILED DESCRIPTION

[0023] In the following description, specific details such as specific system structures, technologies, etc. are provided for the purpose of illustration rather than limitation, so as to provide a thorough understanding of the embodiments of the present invention. However, it should be clear to those skilled in the art that the present invention may be implemented in other embodiments without these specific details. In other cases, detailed descriptions of well-known systems, devices, circuits, and methods are omitted to prevent unnecessary details from obstructing the description of the present invention.

[0024] It should be understood that when used in the present specification and the appended claims, the term "comprising" indicates the presence of described features, integers, steps, operations, elements and / or components, but does not exclude the presence or addition of one or more other features, integers, steps, operations, elements, components and / or combinations thereof.

[0025] It should also be understood that the term "and / or" used in the present description and the appended claims refers to and includes any and all possible combinations of one or more of the associated listed items.

[0026] As used in the present specification and the appended claims, the term "if" may be interpreted as "when" or "upon" or "in response to determining" or "in response to detecting", depending on the context. Similarly, the phrase "if it is determined" or "if [described condition or event] is detected" may be interpreted as meaning "upon determination" or "in response to determining" or "upon detection of [described condition or event]" or "in response to detecting [described condition or event]", depending on the context.

[0027] In addition, in the description of the present specification and the appended claims, the terms "first", "second", "third", etc. are only used to distinguish the descriptions and cannot be understood as indicating or implying relative importance.

[0028] References to "one embodiment" or "some embodiments" etc. described in the present specification mean that one or more embodiments of the present invention include specific features, structures or characteristics described in conjunction with the embodiment. Therefore, the statements "in one embodiment", "in some embodiments", "in some other embodiments", "in some other embodiments", etc. that appear in different places in this specification do not necessarily refer to the same embodiment, but mean "one or more but not all embodiments", unless otherwise specifically emphasized in other ways. The terms "including", "comprising", "having" and their variations all mean "including but not limited to", unless otherwise specifically emphasized in other ways.

[0029] The embodiments of the present invention can acquire and process relevant data based on artificial intelligence technology. Artificial Intelligence (AI) is the theory, method, technology and application system that uses digital computers or machines controlled by digital computers to simulate, extend and expand human intelligence, perceive the environment, acquire knowledge and use knowledge to obtain the best results.

[0030] AI basic technologies generally include sensors, dedicated AI chips, cloud computing, distributed storage, big data processing technology, operation / interaction systems, mechatronics, etc. AI software technologies mainly include computer vision technology, robotics technology, biometrics technology, speech processing technology, natural language processing technology, and machine learning / deep learning.

[0031] It should be understood that the order of execution of the steps in the following embodiments does not imply a precedence of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present invention.

[0032] In order to illustrate the technical solution of the present invention, specific embodiments are provided below for illustration.

[0033] A planning data processing method provided in the first embodiment of the present invention can be applied in the following aspects: Figure 1 In an application environment, a client communicates with a server. The client includes but is not limited to a PDA, a desktop computer, a laptop computer, an ultra-mobile personal computer (UMPC), a netbook, a cloud computer device, a personal digital assistant (PDA), and other computer devices. The server can be implemented by an independent server or a server cluster consisting of multiple servers.

[0034] See also Figure 2 , is a flow chart of a method for processing planning data provided by the first embodiment of the present invention. The method for processing planning data can be applied to Figure 1 In the client, the planning data processing method may include the following steps:

[0035] S201: Acquire planning data, input the planning data and a preset initial sub-vector into a trained planning generation model, and obtain N planning sub-vectors with a time series relationship.

[0036] Among them, the planning data includes the personal information of the trainees and the preset training business information. The personal information of the trainees includes the position information, age information, and education information of the trainees. The preset training business information includes the preset training direction information and training intensity information of the trainees. In order to improve the training effect of the trainees, it is necessary to formulate targeted training directions and training intensities according to the different position information, age information, and education information of the trainees, that is, to generate targeted training plans for each trainee.

[0037] The present application processes the plan data according to the trained plan generation model. First, for each person to be trained, the corresponding plan data of each person to be trained is obtained, including personal information and preset training business information, and an initial sub-vector with all vector parameters as 0 is preset. The plan data and the preset initial sub-vector are input into the trained plan generation model to obtain N plan sub-vectors with a time series relationship, where N is an integer greater than 1.

[0038] Among them, firstly, N planning steps for each person to be trained are preset according to the actual training situation, and the specific value of N can be set according to the actual situation. According to the trained plan generation model, N planning sub-vectors with a time sequence relationship are obtained, and the 1st, 2nd, ..., N planning sub-vectors correspond to the text vectors of the 1st, 2nd, ..., N planning steps respectively, so as to obtain the target plan text according to the 1st, 2nd, ..., N planning sub-vectors with a time sequence relationship as a targeted training plan for each person to be trained, so as to improve the training effect of each person to be trained.

[0039] Optionally, the plan data and the preset initial sub-vector are input into the trained plan generation model to obtain N plan sub-vectors with a time series relationship, including:

[0040] Taking the plan class data and the initial sub-vector as input, initializing the number of iterations to 0, inputting the input into the trained plan generation model, and outputting the first plan sub-vector;

[0041] The planning class data and the first planning sub-vector are used as the pre-planning information, and the number of iterations is increased by one;

[0042] The previous plan information is used as the updated input, and the updated input is input into the trained plan generation model again, and the second plan sub-vector is output, until the number of iterations reaches N, and the Nth plan sub-vector is output, thereby obtaining N plan sub-vectors with a time sequence relationship.

[0043] Among them, the N plan sub-vectors obtained according to the trained plan generation model have a temporal relationship, corresponding to N training plan steps in succession.

[0044] Specifically, the generation process of N plan sub-vectors with a time-series relationship is as follows: first, the plan data and the initial sub-vector are used as input, the number of iterations is initialized to 0, the input is input into the trained plan generation model, and the first plan sub-vector is output, wherein the vector parameters in the initial sub-vector are all 0. Under the premise that the plan sub-vector is not obtained, the plan data and the initial sub-vector with all 0 are used as input, so that the input data type of the plan generation model is consistent. Then, the plan data and the obtained first plan sub-vector are used as the pre-planning information, and the number of iterations is increased once, so that the pre-planning information is used as the updated input, and the updated input is input into the trained plan generation model again, and the second plan sub-vector is output, until the number of iterations reaches N, and the Nth plan sub-vector is output, thereby obtaining N plan sub-vectors with a time-series relationship, corresponding to the N training plan steps with a time-series relationship.

[0045] Optionally, the plan generation model includes a first encoder and a first decoder, and the plan generation model is trained using sample plan class data and initial sub-vectors as training samples and actual sample plan vectors as training labels.

[0046] Among them, the training samples of the plan generation model are a large amount of sample training data and initial sub-vectors, and the training labels are actual sample plan vectors, which serve as the optimization basis for model parameters in the plan generation model and improve the generation effect of the plan generation model.

[0047] Optionally, the training process of the planned generation model includes:

[0048] Inputting the sample plan class data and the initial sub-vector into the first encoder for feature extraction to obtain the sample plan features;

[0049] Performing Gaussian sampling on the sample plan feature to obtain a sample plan sampling result, and inputting the sample plan sampling result into the first decoder for reconstruction to obtain N sample plan sub-vectors with a time sequence relationship;

[0050] Concatenate the N sample plan sub-vectors, and determine the concatenation result as the sample target plan vector;

[0051] The first loss function is calculated according to the sample target plan vector and the actual sample plan vector, and the parameters of the first encoder and the first decoder are reversely corrected according to the gradient descent method until the first loss function converges to obtain a trained plan generation model.

[0052] Among them, for a large number of sample training plans and initial sub-vectors, the sample training plans and the initial sub-vectors are respectively input into the first encoder for feature extraction to obtain sample plan features, and the sample plan features are Gaussian sampled to obtain sample plan sampling results, and the sample plan sampling results are input into the first decoder for reconstruction to obtain N sample plan sub-vectors with a time series relationship.

[0053] Among them, the N sample plan sub-vectors with a time sequence relationship correspond to N training plan steps with a time sequence relationship. Therefore, after the N sample plan sub-vectors are spliced ​​in time sequence, the splicing result is determined to be a sample target plan vector, which is used to represent the complete target plan.

[0054] At the same time, during the training process of the plan generation model, the first loss function is calculated according to the difference between the obtained sample target plan vector and the actual sample plan vector, and the parameters of the first encoder and the first decoder are reversely corrected according to the gradient descent method until the first loss function converges to obtain a trained plan generation model, which is used to input the plan class data and the preset initial sub-vector into the trained plan generation model to obtain N plan sub-vectors with a time series relationship.

[0055] In one embodiment, during the training process of the plan generation model, a first loss function is obtained by calculating the mean square error between the sample target plan vector and the actual sample plan vector, the sample target plan vector is recorded as Y, the actual sample plan vector is recorded as S, and the mean square error between the sample target plan vector Y and the actual sample plan vector S is recorded as the first loss function Loss1, then the first loss function Loss1 is:

[0056]

[0057] Where Y is the sample target plan vector and S is the actual sample plan vector.

[0058] In one embodiment, based on the trained plan generation model, the acquired plan data of the person to be trained and the initial sub-vector whose preset vector parameters are all 0 are input into the trained plan generation model to obtain N plan sub-vectors with a time sequence relationship of the person to be trained, and the N plan sub-vectors are respectively recorded as J1, J2, ..., J N , respectively corresponding to the text vectors of the 1st, 2nd, ..., Nth planning steps of the targeted plan formulated for the trainee, so as to improve the training effect of the trainee according to the targeted training plan steps.

[0059] S202: Among the N plan sub-vectors, the i-th plan sub-vector and the i+1-th plan sub-vector are input into the trained probability prediction model in turn to obtain the first probability that the N-1 i-th plan sub-vectors and the i+1-th plan sub-vectors conform to the temporal relationship.

[0060] Among them, after obtaining the planning data of the personnel to be trained, N planning sub-vectors with time series relationship of the personnel to be trained are obtained based on the preset initial sub-vector and the trained planning generation model, which respectively correspond to the text vectors of the 1st, 2nd, ..., N planning steps of the targeted plan formulated for the personnel to be trained.

[0061] Therefore, the i-th (i=1, 2, ..., N-1) plan subvector and the i+1-th plan subvector meet the corresponding timing relationship. In order to characterize the degree of timing between the N plan subvectors generated by the plan generation model and to characterize the generation effect of the trained plan generation model, the i-th plan subvector and the i+1-th plan subvector are input into the trained probability prediction model to obtain the first probability that the i-th plan subvector and the i+1-th plan subvector meet the timing relationship. The larger the first probability, the greater the probability. The higher the degree of temporal sequence between the quantity and the i+1th plan sub-vector, the better the generation effect of the trained plan generation model. Conversely, the smaller the first probability, the lower the degree of temporal sequence between the i-th plan sub-vector and the i+1th plan sub-vector, indicating that the generation effect of the trained plan generation model is worse. It is necessary to adjust the parameters of the trained plan generation model to optimize the generation effect of the plan generation model and obtain N plan sub-vectors with stronger temporal sequence to improve the quality of the targeted plan formulated for the trainee.

[0062] Among them, the value of i is 1, 2, ..., N-1, and based on the trained probability prediction model, N-1 first probabilities of N plan sub-vectors can be obtained to characterize the generation effect of the trained plan generation model.

[0063] Optionally, the probability prediction model includes a second encoder and a second decoder, and the probability prediction model is trained using two sample plan subvectors in adjacent time sequence as training samples and using the timing degree value between the two sample plan subvectors in adjacent time sequence as training labels.

[0064] Among them, the temporal degree value between two sample plan subvectors of adjacent temporal order corresponds to the probability value that the two sample plan subvectors of adjacent temporal order meet the temporal relationship. The training samples of the probability prediction model are a large number of sample plan subvectors of adjacent temporal order, and the training labels are the temporal degree values ​​between two sample plan subvectors of adjacent temporal order, which also serve as the optimization basis of the model parameters in the probability prediction model to improve the prediction effect of the probability prediction model.

[0065] Optionally, the training process of the probability prediction model includes:

[0066] Input two sample plan subvectors in adjacent time sequence into a second encoder for feature extraction to obtain sample probability features;

[0067] Performing Gaussian sampling on the sample probability feature to obtain a sample probability sampling result, and inputting the sample probability sampling result into the second decoder for reconstruction to obtain a prediction value of the degree of temporal sequence between two sample plan subvectors in adjacent temporal order;

[0068] The second loss function is calculated according to the predicted value of the degree of temporal sequence and the value of the degree of temporal sequence between two sample plan subvectors in adjacent temporal sequence, and the parameters of the second encoder and the second decoder are reversely corrected according to the gradient descent method until the second loss function converges to obtain a trained probability prediction model.

[0069] Among them, during the training process of the probability prediction model, the second loss function is calculated according to the difference between the predicted value of the degree of timing between two sample plan sub-vectors of adjacent temporal order and the corresponding degree of timing, and the parameters of the second encoder and the second decoder are reversely corrected according to the gradient descent method until the second loss function converges to obtain a trained probability prediction model, which is used to input the i-th plan sub-vector and the i+1-th plan sub-vector into the trained probability prediction model to obtain the first probability that the i-th plan sub-vector and the i+1-th plan sub-vector conform to the temporal relationship.

[0070] In one embodiment, during the training process of the probability prediction model, a second loss function is obtained by calculating the mean square error between the predicted value of the degree of temporality and the corresponding degree of temporality between two sample plan subvectors of adjacent temporal order, the predicted value of the degree of temporality between two sample plan subvectors of adjacent temporal order is recorded as X, the degree of temporality between two sample plan subvectors of corresponding adjacent temporal order is recorded as Z, and the mean square error between the predicted value of the degree of temporality X and the degree of temporality Z between two sample plan subvectors of adjacent temporal order is recorded as the second loss function Loss2, then the second loss function Loss2 is:

[0071]

[0072] Wherein, X is the predicted value of the degree of temporal sequence between two sample plan subvectors of adjacent temporal sequence, and Z is the value of the degree of temporal sequence between two sample plan subvectors of corresponding adjacent temporal sequence.

[0073] In one embodiment, based on the trained probability prediction model, the i-th plan subvector J i and the i+1th plan subvector J i+1 Input it into the trained probability prediction model to obtain the i-th plan subvector J i and the i+1th plan subvector J i+1 The first probability that the time sequence relationship is met is recorded as G i(i+1) .

[0074] Based on the trained probability prediction model, N planning sub-vectors J1, J2, ..., J N The N-1 first probability G 12 , G 23 , …, G N(N-1) , which is used to characterize the generation effect of the trained plan generation model.

[0075] S203: Determine whether the N-1 first probabilities meet a preset condition. If the N-1 first probabilities meet the preset condition, use the N plan sub-vectors as N target plan sub-vectors.

[0076] Among them, the N-1 first probabilities of the N plan sub-vectors obtained based on the trained probability prediction model are used to characterize the prediction effect of the trained probability prediction model. First, it is determined whether the N-1 first probabilities meet the preset conditions. If the N-1 first probabilities meet the preset conditions, it indicates that the timing between the N plan sub-vectors obtained according to the plan generation model is high, that is, the generation effect of the plan generation model is good, and there is no need to optimize the model parameters. The N plan sub-vectors corresponding to the N-1 first probabilities are used as N target plan sub-vectors to obtain the final complete target plan text.

[0077] Optionally, determining whether the N-1 first probabilities meet a preset condition further includes:

[0078] If the N-1 first probabilities do not meet the preset conditions, adjust the parameters of the plan generation model to obtain N-1 second probabilities after parameter adjustment, calculate the difference between the N-1 first probabilities and the N-1 second probabilities, determine the parameter adjustment direction of the plan generation model according to the difference, adjust the parameters of the plan generation model according to the parameter adjustment direction, until the N-1 second probabilities after parameter adjustment meet the preset conditions, and use the N plan sub-vectors corresponding to the N-1 second probabilities as the N target plan sub-vectors.

[0079] Among them, if the N-1 first probabilities do not meet the preset conditions, it indicates that the timing degree between the N plan sub-vectors obtained according to the plan generation model is low, that is, the generation effect of the plan generation model is poor, and the model parameters need to be optimized. Therefore, by adjusting the parameters of the plan generation model, the N-1 second probabilities after parameter adjustment are obtained, and the difference between the N-1 first probabilities and the N-1 second probabilities is calculated, and finally the parameter adjustment direction of the plan generation model is determined according to the difference.

[0080] Among them, the parameter adjustment direction includes the parameter increase direction and the parameter decrease direction. When optimizing the parameters of the plan generation model, first, the parameter increase direction or the parameter decrease direction is randomly set as the initial parameter adjustment direction. After adjusting the parameters of the plan generation model according to the initial parameter adjustment direction, the two plan sub-vectors in the adjacent time sequence order in the N plan sub-vectors are input into the trained probability prediction model again to obtain N-1 second probabilities after parameter adjustment. Then, the N-1 differences between the N-1 first probabilities and the N-1 second probabilities are calculated, and the sum of the N-1 differences is calculated. If the N-1 first probabilities are reduced If the sum of the N-1 differences minus the N-1 second probabilities is not less than 0, it means that after adjusting the parameters of the plan generation model according to the initial parameter adjustment direction, the generation effect of the plan generation model is improved. Therefore, the initial parameter adjustment direction is determined as the parameter adjustment direction of the plan generation model. If the sum of the N-1 differences between the N-1 first probabilities and the N-1 second probabilities is less than 0, it means that after adjusting the parameters of the plan generation model according to the initial parameter adjustment direction, the generation effect of the plan generation model is reduced. Therefore, another parameter adjustment direction other than the initial parameter adjustment direction is determined as the parameter adjustment direction of the plan generation model.

[0081] The parameter adjustment direction can be determined according to the difference between the probability result before the parameter adjustment and the probability result after the parameter adjustment, and the parameters of the plan generation model can be adjusted according to the determined parameter adjustment direction until the N-1 second probabilities after the parameter adjustment meet the preset conditions, thereby taking the N plan sub-vectors corresponding to the N-1 second probabilities after the parameter adjustment as the N target plan sub-vectors.

[0082] In one embodiment, the N-1 first probabilities are respectively recorded as Set the parameter increase direction as the initial parameter adjustment direction, increase the internal parameters of the plan generation model, and then input the two plan sub-vectors in the adjacent time sequence order in the N plan sub-vectors into the trained probability prediction model again to obtain N-1 second probabilities after parameter adjustment. The N-1 second probabilities are recorded as Then, N-1 differences between the N-1 first probabilities and the N-1 second probabilities are calculated respectively, and the sum of the N-1 differences is recorded as C. Then the sum C of the N-1 differences is:

[0083]

[0084] In the formula, is the jth first probability, is the j-th second probability, and the value of j is 1, 2, …, N-1.

[0085] If the sum C of the N-1 differences is ≥ 0, the initial parameter adjustment direction is determined, that is, the parameter increase direction is the parameter adjustment direction of the plan generation model, and the parameters of the plan generation model are adjusted to improve the generation effect of the plan generation model.

[0086] If the sum C of the N-1 differences is less than 0, then another parameter adjustment direction other than the initial parameter adjustment direction is determined, that is, the parameter reduction direction is the parameter adjustment direction of the plan generation model, and the parameters of the plan generation model are adjusted to improve the generation effect of the plan generation model.

[0087] S204: splicing the N target plan sub-vectors, determining the splicing result as the target plan vector, and determining the target plan text according to the target plan vector.

[0088] Among them, the N target plan sub-vectors obtained have a time sequence relationship and correspond to the text vectors of the N training plan steps with a time sequence relationship. Therefore, the N target plan sub-vectors are spliced ​​in time sequence, and the splicing result is determined as the target plan vector. Among them, the target plan vector is the text vector of the complete training plan, and the target plan text can be determined according to the target plan vector to obtain a targeted training plan for the trainees.

[0089] Optionally, determining the target plan text according to the target plan vector includes:

[0090] The target plan vector is input into the trained text reconstruction model to obtain the target plan text.

[0091] Among them, the text reconstruction model is used to input the target plan vector and output the target plan text to obtain a complete target plan with good time sequence, so that targeted training can be carried out for trainees according to the target plan text to improve the training effect of the trainees.

[0092] In one embodiment, the N target plan subvectors are respectively denoted as J1′, J2′, …, J N ′, and concatenate the N target plan sub-vectors in chronological order to obtain the target plan vector J′=[J1′, J2′, …, J N′], the target plan vector J′ is input into the trained text reconstruction model to obtain the target plan text, so that the trainees can be trained in a targeted manner according to the target plan text, the feasibility of the target plan text is improved, and the training effect of the trainees is improved.

[0093] The embodiment of the present invention obtains N plan subvectors with a time series relationship by inputting plan data and a preset initial subvector into a trained plan generation model, and inputs the ith plan subvector and the i+1th plan subvector into a trained probability prediction model to obtain a first probability that the ith plan subvector and the i+1th plan subvector meet the time series relationship, thereby obtaining N-1 first probabilities of the N plan subvectors, and determining N target plan subvectors by judging whether the N-1 first probabilities meet preset conditions, and splicing the N target plan subvectors to determine the splicing result as the target plan vector, and finally determining the target plan text according to the target plan vector, and obtaining N plan subvectors with a time series relationship according to the plan data, thereby improving the time series of the final complete target plan text, improving the feasibility of the target plan text, and improving the training effect on the trainees.

[0094] Corresponding to the planning data processing method in the above embodiment, Figure 3 A structural block diagram of a planning data processing device provided in the second embodiment of the present invention is given. For the convenience of explanation, only the parts related to the embodiment of the present invention are shown.

[0095] See also Figure 3 , the data processing device of this plan includes:

[0096] The plan sub-vector generation module 31 is used to obtain plan data, input the plan data and the preset initial sub-vector into the trained plan generation model, and obtain N plan sub-vectors with a time series relationship, wherein the plan data includes the personal information of the personnel to be trained and the preset training business information, and N is an integer greater than 1;

[0097] A first probability prediction module 32 is used to sequentially input the ith plan subvector and the i+1th plan subvector among the N plan subvectors into the trained probability prediction model to obtain a first probability that the N-1 ith plan subvectors and the i+1th plan subvectors conform to the temporal relationship, where i=1, 2, ..., N-1;

[0098] A target plan sub-vector determination module 33 is used to determine whether N-1 first probabilities meet a preset condition, and if N-1 first probabilities meet the preset condition, use the N plan sub-vectors as N target plan sub-vectors;

[0099] The target plan text determination module 34 is used to splice the N target plan sub-vectors, determine the splicing result as the target plan vector, and determine the target plan text according to the target plan vector.

[0100] Optionally, the plan sub-vector generating module 31 includes:

[0101] A first plan sub-vector generating sub-module, used to take the plan class data and the preset initial sub-vector as input, initialize the number of iterations to 0, input the input to the trained plan generating model, and output the first plan sub-vector;

[0102] A pre-planning information determination submodule, used to use the planning class data and the first planning sub-vector as pre-planning information and increase the number of iterations by one;

[0103] The second plan sub-vector generating sub-module is used to take the previous plan information as the updated input, input the updated input into the trained plan generation model again, and output the second plan sub-vector until the number of iterations reaches N, and output the Nth plan sub-vector, thereby obtaining N plan sub-vectors with a time sequence relationship.

[0104] The plan generation model training submodule is used to determine that the plan generation model includes a first encoder and a first decoder, uses sample plan class data and initial word vectors as training samples, and uses actual sample plan vectors as training labels to train the plan generation model.

[0105] Optionally, the plan generation model training submodule includes:

[0106] A feature extraction unit, used for inputting the sample plan class data and the preset initial sub-vector into the first encoder for feature extraction to obtain the sample plan feature;

[0107] A sample plan sub-vector generating unit is used to perform Gaussian sampling on the sample plan feature to obtain a sample plan sampling result, and input the sample plan sampling result to the first decoder for reconstruction to obtain N sample plan sub-vectors with a time sequence relationship;

[0108] A sample target plan vector concatenation unit is used to concatenate N sample plan sub-vectors and determine the concatenation result as a sample target plan vector;

[0109] A parameter correction unit is used to calculate a first loss function based on a sample target plan vector and an actual sample plan vector, and reversely correct the parameters of a first encoder and a first decoder according to a gradient descent method until the first loss function converges to obtain a trained plan generation model.

[0110] Optionally, the target plan sub-vector determination module 33 includes:

[0111] The parameter adjustment submodule is used to adjust the parameters of the plan generation model if the N-1 first probabilities do not meet the preset conditions, obtain N-1 second probabilities after the parameter adjustment, calculate the difference between the N-1 first probabilities and the N-1 second probabilities, determine the parameter adjustment direction of the plan generation model according to the difference, adjust the parameters of the plan generation model according to the parameter adjustment direction, until the N-1 second probabilities after the parameter adjustment meet the preset conditions, and use the N plan sub-vectors corresponding to the N-1 second probabilities as the N target plan sub-vectors.

[0112] Optionally, the parameter adjustment submodule includes:

[0113] A first parameter adjustment submodule unit, used to determine a parameter adjustment direction including a parameter increase direction and a parameter decrease direction;

[0114] A second parameter adjustment submodule unit is used to randomly set a parameter increase direction or a parameter decrease direction as an initial parameter adjustment direction, and adjust the parameters of the plan generation model according to the initial parameter adjustment direction to obtain N-1 second probabilities after the parameter adjustment;

[0115] A parameter adjustment direction determination unit is used to calculate N-1 differences between N-1 first probabilities and N-1 second probabilities, and calculate the sum of the N-1 differences. If the sum of the N-1 differences is not less than 0, the initial parameter adjustment direction is determined to be the parameter adjustment direction of the plan generation model. If the sum of the N-1 differences is less than 0, another parameter adjustment direction other than the initial parameter adjustment direction is determined to be the parameter adjustment direction of the plan generation model.

[0116] Optionally, the target plan text determination module 34 includes:

[0117] The target plan text determination submodule is used to input the target plan vector into the trained text reconstruction model to obtain the target plan text.

[0118] It should be noted that the information interaction, execution process and other contents between the above modules are based on the same concept as the embodiment of the method of the present invention. Their specific functions and technical effects can be found in the method embodiment part and will not be repeated here.

[0119] Figure 4 This is a schematic diagram of the structure of a computer device provided in Embodiment 3 of the present invention. Figure 4 As shown, the computer device of this embodiment includes: at least one processor ( Figure 4 Only one is shown), a memory, and a computer program stored in the memory and executable on at least one processor, wherein when the processor executes the computer program, the steps in any of the above-mentioned planning data processing method embodiments are implemented.

[0120] The computer device may include, but is not limited to, a processor and a memory. Those skilled in the art will appreciate that Figure 4 This is merely an example of a computer device and does not constitute a limitation on the computer device. The computer device may include more or fewer components than those shown in the figure, or a combination of certain components, or different components. For example, it may also include a network interface, a display screen, and an input device.

[0121] The processor may be a CPU, or other general-purpose processors, a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor may be a microprocessor or any conventional processor, etc.

[0122] The memory includes a readable storage medium, an internal memory, etc., wherein the internal memory may be the memory of a computer device, and the internal memory provides an environment for the operation of an operating system and computer-readable instructions in the readable storage medium. The readable storage medium may be a hard disk of a computer device, and in other embodiments may also be an external storage device of a computer device, for example, a plug-in hard disk, a smart memory card (Smart Media Card, SMC), a secure digital (Secure Digital, SD) card, a flash card (Flash Card), etc. equipped on a computer device. Further, the memory may also include both an internal storage unit of a computer device and an external storage device. The memory is used to store an operating system, an application program, a boot loader (BootLoader), data, and other programs, such as program codes of computer programs, etc. The memory may also be used to temporarily store data that has been output or is to be output.

[0123] The technicians in the relevant field can clearly understand that for the convenience and simplicity of description, only the division of the above-mentioned functional units and modules is used as an example. In practical applications, the above-mentioned function allocation can be completed by different functional units and modules as needed, that is, the internal structure of the device is divided into different functional units or modules to complete all or part of the functions described above. The functional units and modules in the embodiment can be integrated into a processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The above-mentioned integrated unit can be implemented in the form of hardware or in the form of software functional units. In addition, the specific names of the functional units and modules are only for the convenience of distinguishing each other, and are not used to limit the scope of protection of the present invention. The specific working process of the units and modules in the above-mentioned device can refer to the corresponding process in the above-mentioned method embodiment, which will not be repeated here. If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on such understanding, the present invention implements all or part of the processes in the above-mentioned embodiment method, which can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a computer-readable storage medium. When the computer program is executed by the processor, the steps of the above-mentioned method embodiment can be implemented. Among them, the computer program includes computer program code, and the computer program code can be in source code form, object code form, executable file or some intermediate form. The computer-readable medium may at least include: any entity or device capable of carrying computer program code, recording medium, computer memory, read-only memory (ROM), random access memory (RAM), electric carrier signal, telecommunication signal and software distribution medium. For example, a USB flash drive, a mobile hard disk, a magnetic disk or an optical disk. In some jurisdictions, according to legislation and patent practice, computer-readable media cannot be electric carrier signals and telecommunication signals.

[0124] The present invention implements all or part of the processes in the above-mentioned embodiment method, and can also be completed by a computer program product. When the computer program product runs on a computer device, the computer device can implement the steps in the above-mentioned method embodiment when executing.

[0125] In the above embodiments, the description of each embodiment has its own emphasis. For parts that are not described or recorded in detail in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0126] Those skilled in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of the present invention.

[0127] In the embodiments provided by the present invention, it should be understood that the disclosed devices / computer equipment and methods can be implemented in other ways. For example, the device / computer equipment embodiments described above are only schematic, for example, the division of modules or units is only a logical function division, and there may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms.

[0128] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0129] The above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that the technical solutions described in the aforementioned embodiments may still be modified, or some of the technical features may be replaced by equivalents. Such modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention, and should be included in the protection scope of the present invention.

Claims

1. A planning data processing method based on artificial intelligence, characterized in that: The planning data processing method comprises: Acquire planning data, input the planning data and a preset initial sub-vector into a trained planning generation model, and obtain N planning sub-vectors with a time series relationship, wherein the planning data includes personal information of the personnel to be trained and preset training business information, and N is an integer greater than 1; Among the N plan sub-vectors, the i-th plan sub-vector and the i+1-th plan sub-vector are sequentially input into the trained probability prediction model to obtain the first probabilities that the N-1 i-th plan sub-vectors and the i+1-th plan sub-vectors conform to the temporal relationship, where i=1, 2, ..., N-1; Determine whether N-1 of the first probabilities meet a preset condition, and if N-1 of the first probabilities meet the preset condition, use the N plan sub-vectors as N target plan sub-vectors; The N target plan sub-vectors are spliced, a splicing result is determined as a target plan vector, and a target plan text is determined according to the target plan vector.

2. The planning data processing method according to claim 1, characterized in that: The plan data and the preset initial sub-vector are input into the trained plan generation model to obtain N plan sub-vectors with a time sequence relationship, including: Taking the plan data and a preset initial sub-vector as input, initializing the number of iterations to 0, inputting the input into the trained plan generation model, and outputting a first plan sub-vector; Using the plan class data and the first plan sub-vector as pre-plan information, and increasing the number of iterations by one; The pre-plan information is used as an updated input, and the updated input is input into the trained plan generation model again, and a second plan sub-vector is output, until the number of iterations reaches N, and the Nth plan sub-vector is output, thereby obtaining N plan sub-vectors with a time sequence relationship.

3. The planning data processing method according to claim 2, characterized in that: The plan generation model includes a first encoder and a first decoder. The plan generation model is trained using sample plan class data and initial word vectors as training samples and actual sample plan vectors as training labels.

4. The planning data processing method according to claim 3 is characterized in that: The training process of the plan generation model includes: Inputting the sample plan class data and the preset initial sub-vector into the first encoder for feature extraction to obtain sample plan features; Performing Gaussian sampling on the sample plan feature to obtain a sample plan sampling result, and inputting the sample plan sampling result into the first decoder for reconstruction to obtain N sample plan sub-vectors with a time sequence relationship; Splicing the N sample plan sub-vectors, and determining the splicing result as a sample target plan vector; A first loss function is calculated based on the sample target plan vector and the actual sample plan vector, and the parameters of the first encoder and the first decoder are reversely corrected according to the gradient descent method until the first loss function converges to obtain a trained plan generation model.

5. The planning data processing method according to claim 1, characterized in that: Determining whether the N-1 first probabilities meet a preset condition further includes: If the N-1 first probabilities do not meet the preset conditions, adjust the parameters of the plan generation model to obtain N-1 second probabilities after parameter adjustment, calculate the difference between the N-1 first probabilities and the N-1 second probabilities, determine the parameter adjustment direction of the plan generation model according to the difference, adjust the parameters of the plan generation model according to the parameter adjustment direction, until the N-1 second probabilities after parameter adjustment meet the preset conditions, and use the N plan sub-vectors corresponding to the N-1 second probabilities as the N target plan sub-vectors.

6. The planning data processing method according to claim 5, characterized in that: Adjusting the parameters of the plan generation model to obtain N-1 second probabilities after the parameters are adjusted, calculating the difference between the N-1 first probabilities and the N-1 second probabilities, and determining the parameter adjustment direction of the plan generation model according to the difference includes: The parameter adjustment direction includes a parameter increase direction and a parameter decrease direction; Randomly setting a parameter increase direction or a parameter decrease direction as an initial parameter adjustment direction, adjusting the parameters of the plan generation model according to the initial parameter adjustment direction, and obtaining N-1 second probabilities after the parameter adjustment; Calculate N-1 differences between N-1 first probabilities and N-1 second probabilities, and calculate the sum of the N-1 differences. If the sum of the N-1 differences is not less than 0, determine the initial parameter adjustment direction as the parameter adjustment direction of the plan generation model. If the sum of the N-1 differences is less than 0, determine another parameter adjustment direction other than the initial parameter adjustment direction as the parameter adjustment direction of the plan generation model.

7. The planning data processing method according to claim 1, characterized in that: Determining the target plan text according to the target plan vector includes: The target plan vector is input into a trained text reconstruction model to obtain a target plan text.

8. A planning data processing device based on artificial intelligence, characterized in that: The planning data processing device comprises: A plan sub-vector generation module is used to obtain plan data, input the plan data and the preset initial sub-vector into the trained plan generation model, and obtain N plan sub-vectors with a time series relationship, wherein the plan data includes personal information of the personnel to be trained and preset training business information, and N is an integer greater than 1; A first probability prediction module is used to input the ith plan subvector and the i+1th plan subvector in the N plan subvectors into the trained probability prediction model in sequence to obtain a first probability that the N-1 ith plan subvectors and the i+1th plan subvectors conform to the temporal relationship, where i=1, 2, ..., N-1; a target plan sub-vector determining module, configured to determine whether N-1 of the first probabilities meet a preset condition, and if N-1 of the first probabilities meet the preset condition, use the N plan sub-vectors as N target plan sub-vectors; The target plan text determination module is used to splice the N target plan sub-vectors, determine the splicing result as the target plan vector, and determine the target plan text according to the target plan vector.

9. A computer device, characterized in that: The computer device includes a processor, a memory, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the planning data processing method as described in any one of claims 1 to 7 is implemented.

10. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the planning data processing method as described in any one of claims 1 to 7 is implemented.

Images