Data analysis processing method based on intelligent AI

Abstract

The invention relates to the technical field of data analysis and processing, and discloses a data analysis and processing method based on intelligent AI, which comprises the following steps of: 1, constructing a knowledge graph model, modeling a data entity and a relationship, mapping a knowledge graph to a vector space by adopting a graph embedding method, and constraining a semantic relationship of data by utilizing a first-order predicate logic rule; 2, on the basis of the semantic relation of the knowledge graph, data correction constraints are constructed by means of S5 modal logical reasoning, necessary conditions and possible conditions of data are defined, and correction values of data points under the logical constraints are calculated through Kripke semantics; and step 3, on the basis of the data correction constraint, combining neural logic reasoning and a variational auto-encoder to optimize the data. According to the method, the technical scheme that intelligent map construction is combined with S5 modal logical reasoning is adopted, the data semantic rigorous correction effect is achieved, and compared with an existing traditional data cleaning scheme, the problems that logic constraints are not tight and correction precision is insufficient are solved.

Description

Technical Field

[0001] The present invention relates to the technical field of data analysis and processing, and particularly to a data analysis and processing method based on intelligent AI. Background Art

[0002] With the explosive growth of data volume, how to efficiently and accurately process and analyze these massive data has become an urgent problem to be solved. Most of the existing technologies adopt traditional data cleaning methods, mainly through rule-driven methods for data correction. However, the complex semantic relationships behind the data are ignored, resulting in inaccurate and incomplete data correction results. Traditional methods lack the ability of deep reasoning when facing data semantic understanding and complex data relationships. Especially when dealing with data involving multi-level logical constraints, problems such as incomplete data correction and inaccurate correction of data errors are likely to occur.

[0003] Most of the existing technologies rely on rule-driven cleaning strategies. The method can handle simple outliers, but has poor effects when facing complex data semantics and correlation relationships. Especially in large-scale data processing, the limitations of rule formulation make data correction inflexible and unable to meet the requirements of accurate correction, resulting in problems of insufficient accuracy.

[0004] The existing technologies identify abnormal data based on rule presets or simple models, and cannot detect and respond to adversarial samples in data in real time. With the continuous upgrading of attack means, traditional detection mechanisms are easily bypassed, resulting in data security problems, and lack of robustness optimization, unable to ensure the stability of the corrected data under external perturbations.

[0005] Most of the existing data correction methods rely on static display modes. Users can see the results but cannot quickly give feedback and make adjustments, resulting in users being unable to intervene and optimize the results in a timely manner during the process of correcting data, affecting the final data quality and processing efficiency.

[0006] Therefore, the present invention effectively solves the above deficiencies in the existing technologies by introducing intelligent graph construction and S5 modal logic reasoning technology, combining the optimization methods of neural logic reasoning and variational autoencoders, and integrating an interactive display and adaptive feedback mechanism, improving the accuracy, efficiency and flexibility of data correction.

[0007] For this reason, the present invention proposes a data analysis and processing method based on intelligent AI to solve the above-mentioned problems. Summary of the Invention

[0008] Aiming at the deficiencies of the existing technologies, the present invention provides a data analysis and processing method based on intelligent AI to solve the problems mentioned in the above background art.

[0009] To achieve the above objectives, the present invention is implemented through the following technical solutions: A data analysis and processing method based on intelligent AI, including:

[0010] Step 1: Construct a knowledge graph model, model data entities and relationships, map the knowledge graph to a vector space using graph embedding methods, and use first-order predicate logic rules to constrain the semantic relationships of data;

[0011] Step 2: Based on the semantic relationships of the knowledge graph, use S5 modal logic reasoning to construct data correction constraints, define the necessary and possible conditions of the data, and calculate the corrected values of data points under logical constraints through Kripke semantics;

[0012] Step 3: On the basis of data correction constraints, combine neural logic reasoning and variational autoencoders to optimize data, train a neuro-symbolic reasoning model to learn correction rules, and generate corrected data that conforms to logical constraints through variational inference;

[0013] Step 4: Conduct a security assessment on the corrected data, and use adversarial sample detection and robustness optimization strategies to detect abnormal inputs;

[0014] Step 5: Display the corrected data and the reasoning process based on an intelligent human-computer interaction interface, provide an interactive feedback mechanism, enable users to adjust data correction rules and model parameters, and optimize the data processing flow.

[0015] Preferably, in step S1, when the data standardization conditions are met and the data is from a preprocessed structured database, it further includes:

[0016] Step 1.1, construct a basic knowledge graph model, and set the data entity set as ε a and the relationship set as R b Define the knowledge graph as:

[0017] KG = {(ε, ρ, γ)},

[0018] where KG is the knowledge graph set, ε is the data entity, ρ is the semantic association between data, and γ is another data entity;

[0019] Step 1.2, perform graph embedding learning, and use the TransE method to construct the embedding function defined as:

[0020] F = ∥Λ·α + β - Λ·δ∥,

[0021] where F is the output of the graph embedding function, Λ is the embedding transformation matrix, α is the vector representation of the first data entity in the feature space, β is the relationship vector, and δ is the vector representation of the second data entity in the feature space;

[0022] Step 1.3. Establish data semantic constraint conditions. Based on the knowledge graph constructed in Step 1.1 and the output results of the embedding function in Step 1.2, use first-order predicate logic rules to define the constraints on data entities and their relationships. The logical constraint expression is:

[0023] C = {φ(ε, ρ, γ) | φ: ε → γ},

[0024] where C is the set of logical constraints and φ is the logical mapping rule.

[0025] Preferably, in Step S2, further include under the condition of referring to the data semantic constraint expression C obtained in Step S1:

[0026] Step 2.1. Define the necessity and possibility operators. Let the necessity operator be B a , and let the possibility operator be P b . The data semantic constraint condition to be corrected is denoted as ζ c ;

[0027] Step 2.2. Construct the necessary condition correction formula. According to the S5 modal logic definition, assume that the necessity logical constraint is satisfied Introduce the correction function R d , and the defined formula is:

[0028]

[0029] where U is the set of data correction candidates and u is the specific correction status;

[0030] Step 2.3. Construct the possibility condition correction formula. According to the S5 modal logic definition, assume that the possibility logical constraint is satisfied, and introduce the correction function R e , and the defined formula is:

[0031]

[0032] where V is the set of data correction candidates and v is the correction candidate status.

[0033] Preferably, in Step S3, further include under the condition of referring to the necessary condition correction formula R d (ζ c ) and the possibility condition correction formula R e (ζ c ) results:

[0034] Step 3.1. Adopt a neural logic inference training model. Let the neural theorem proving function be:

[0035]

[0036] Among them, x is the corrected data feature vector of the output, y is the corrected data vector output by the model, k is the neural network weight matrix, λ is the bias vector, and Y is the set of corrected data candidates;

[0037] Step 3.2, use the variational autoencoder to optimize the data. Let the function for generating corrected data be:

[0038]

[0039] Among them, m is the candidate vector for correcting the data to be generated in the output, n is the latent variable, P(m|n) is the conditional probability of data generation, Q(n|m) is the conditional probability of data encoding, P ′ (n) is the prior probability, Δ is an index for measuring the difference between the data encoding probability and the prior probability, and N is the set of latent variable candidates;

[0040] Step 3.3, fuse the results of neural logical reasoning and the variational autoencoder. Let the fusion output function be:

[0041] E(p) = ω1·Φ n (q) + ω2·Γ v (r),

[0042] Among them, p is the finally generated corrected data vector, Φ n (q) is the output of the neuro-symbolic reasoning model, Γ v (r) is the data generated by the variational autoencoder, ω1 is the neural logical reasoning weight factor, ω2 is the variational autoencoder weight factor, q is the input data feature vector, and r is the variational autoencoder input data vector.

[0043] Preferably, in step S4, further included under the result of referring to the fusion output function E(p) in step S3:

[0044] Step 4.1, adopt the adversarial sample detection method. Let the adversarial sample detection function be:

[0045]

[0046] Among them, is the value of the adversarial sample detection function, E(p) is the corrected data vector generated by fusing the results of neural logical reasoning and the variational autoencoder, i is the adversarial candidate sample, I is the set of adversarial candidates, is the detection input feature;

[0047] Step 4.2, adopt the robustness optimization strategy. Let the robustness optimization function be:

[0048]

[0049] Among them, Θ r(ψ) is the robust optimization function value, E(p) is the corrected data vector generated by fusing the results of neural logic reasoning and variational autoencoder, j is the robust candidate state, J is the robust candidate set, and ψ is the robust optimization input;

[0050] Step 4.3, fuse the adversarial sample detection and the robust optimization results, and set the security evaluation output function as:

[0051]

[0052] where, Σ(μ) is the final security evaluation value, ξ1 is the anti-detection weight factor, ξ2 is the robust optimization weight factor, μ is the security evaluation input variable, and ψ is the robust optimization input.

[0053] Preferably, in step S5, further include under the reference of the security evaluation output function Σ(μ) in step S4 and the result of the fusion output function E(p) in step S3:

[0054] Step 5.1, construct an interactive display function, and set the interactive display function as:

[0055] D(ω) = τ1·Σ(μ) + τ2·E(p),

[0056] where, D(ω) represents the interactive display output, Σ(μ) is the final security evaluation value, E(p) is the corrected data vector generated by fusing the results of neural logic reasoning and variational autoencoder, τ1 represents the security evaluation display weight factor, τ2 represents the data correction display weight factor, and ω represents the interface control parameter;

[0057] Step 5.2, construct an interactive feedback adjustment function, and set the feedback adjustment function as:

[0058]

[0059] where, f(l) represents the interactive feedback output, D(ω) represents the interactive display output, k n represents the interactive candidate adjustment state, K5 represents the interactive candidate set, and l represents the feedback control variable;

[0060] Step 5.3, construct an interactive parameter optimization function, and set the parameter optimization function as:

[0061]

[0062] where, represents the interactive parameter optimization output, and the necessary condition correction formula R d (ζ c ) and the possibility condition correction formula R e (ζ c) where f(l) represents the interactive feedback output, η1 represents the necessary condition correction weight factor, η2 represents the possibility condition correction weight factor, η3 represents the interactive feedback weight factor, and ζ represents the optimization control variable.

[0063] Preferably, the intelligent human - machine interaction interface uses a graph neural network for visualizing data relationships. Users can adjust the data correction logic and view the semantic consistency analysis report of the corrected data in real - time.

[0064] Preferably, the data analysis and processing method based on intelligent AI adjusts the correction strategies of different data sets through adaptive parameters, improving cross - domain applicability.

[0065] A terminal device, which includes at least one set of processing units for executing the steps of the data analysis and processing method based on intelligent AI to complete data correction operations and provide interactive operations for data correction through a user interface.

[0066] A storage medium stores a computer - executable program. When the program is executed, it can implement the data analysis and processing method based on intelligent AI to complete data correction operations and provide dynamic correction feedback through an intelligent human - machine interaction interface.

[0067] The present invention provides a data analysis and processing method based on intelligent AI, having the following beneficial effects:

[0068] 1. The present invention adopts the technical solution of combining intelligent graph construction with S5 modal logic reasoning to achieve a rigorous correction effect of data semantics. Compared with the existing traditional data cleaning solutions, it solves the problems of loose logical constraints and insufficient correction accuracy.

[0069] 2. The present invention uses neural - logic reasoning combined with variational auto - encoder to optimize the data technical solution, achieving efficient and accurate data correction. Compared with the old rule - driven methods, it overcomes the defects of lagging abnormal input detection and low robustness.

[0070] 3. The present invention integrates the technical solutions of interactive display and adaptive feedback adjustment mechanism, obtaining the advantages of real - time parameter optimization and improved user experience. Compared with the traditional static display mode, it eliminates the disadvantages of slow feedback response and poor interaction flexibility. BRIEF DESCRIPTION OF THE DRAWINGS

[0071] Figure 1 is a flowchart of the present invention. DETAILED DESCRIPTION OF THE INVENTION

[0072] To enable those skilled in the art to understand the solution of the present invention, the following will clearly and completely describe the technical solution in the embodiments of the present invention with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are some, but not all, of the embodiments of the present invention. Based on the embodiments of the present invention, other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the scope of protection of the present invention.

[0073] The following will describe the present invention in detail with reference to the accompanying drawings:

[0074] Embodiment:

[0075] Please refer to the attached Figure 1 , the embodiment of the present invention provides a data analysis and processing method based on intelligent AI, including:

[0076] Step 1: Construct a knowledge graph model, model data entities and relationships, map the knowledge graph to a vector space using a graph embedding method, and use first-order predicate logic rules to constrain the semantic relationships of the data;

[0077] Step 1.1, construct a basic knowledge graph model, let the data entity set be ε a and the relationship set be R b Define the knowledge graph as:

[0078] KG = {(ε, ρ, γ)},

[0079] where KG is the knowledge graph set, ε is the data entity, ρ is the semantic association between data, and γ is another data entity;

[0080] Step 1.2, perform graph embedding learning, and use the TransE method to construct an embedding function defined as:

[0081] F = ∥Λ·α + β - Λ·δ∥,

[0082] where F is the output of the graph embedding function, Λ is the embedding transformation matrix, α is the vector representation of the first data entity in the feature space, β is the relationship vector, and δ is the vector representation of the second data entity in the feature space;

[0083] Step 1.3, establish data semantic constraint conditions, and use first-order predicate logic rules to constrain data entities and their relationships according to the knowledge graph constructed in Step 1.1 and the output result of the embedding function in Step 1.2. Define the logical constraint expression as:

[0084] C = {φ(ε, ρ, γ)∣φ: ε → γ},

[0085] where C is the logical constraint set and φ is the logical mapping rule;

[0086] Step 2: Based on the semantic relationships in the knowledge graph, use S5 modal logic reasoning to construct data correction constraints, define the necessary and possible conditions of the data, and calculate the corrected values of data points under logical constraints through Kripke semantics;

[0087] Step 2.1, Define the necessity and possibility operators. Let the necessity operator be B a , and let the possibility operator be P b . Denote the semantic constraint conditions of the data to be corrected as ζ c ;

[0088] Step 2.2, Construct the necessary condition correction formula. According to the definition of S5 modal logic, assume that the necessity logical constraint is satisfied Introduce the correction function R d , and the defined formula is:

[0089]

[0090] where U is the data correction candidate set and u is the specific correction state;

[0091] Step 2.3, Construct the possibility condition correction formula. According to the definition of S5 modal logic, assume that the possibility logical constraint is satisfied, and introduce the correction function R e , and the defined formula is:

[0092]

[0093] where V is the data correction candidate set and v is the correction candidate state;

[0094] Step 3: On the basis of the data correction constraints, combine neural logic reasoning and variational autoencoders to optimize the data, train the neuro-symbolic reasoning model to learn the correction rules, and generate corrected data that meets the logical constraints through variational inference;

[0095] Step 3.1, Use neural logic reasoning to train the model. Let the neural theorem proving function be:

[0096]

[0097] where x is the output corrected data feature vector, y is the output corrected data vector of the model, k is the neural network weight matrix, λ is the bias vector, and Y is the corrected data candidate set;

[0098] Step 3.2, Use the variational autoencoder to optimize the data. Let the function for generating corrected data be:

[0099]

[0100] Among them, m is the candidate vector for the corrected data to be generated in the output, n is the latent variable, P(m|n) is the data generation conditional probability, Q(n|m) is the data encoding conditional probability, and P ′ (n) is the prior probability, Δ is an index for measuring the difference between the data encoding probability and the prior probability, and N is the set of latent variable candidates;

[0101] Step 3.3: Integrate the results of neural-logical reasoning and variational autoencoder. Let the integrated output function be:

[0102] E(p) = ω1·Φ n (q) + ω2·Γ v (r),

[0103] Among them, p is the final corrected data vector, Φ n (q) is the output of the neuro-symbolic reasoning model, Γ v (r) is the data generated by the variational autoencoder, ω1 is the neural-logical reasoning weight factor, ω2 is the variational autoencoder weight factor, q is the input data feature vector, and r is the variational autoencoder input data vector;

[0104] Step 4: Conduct a security assessment on the corrected data, and adopt an adversarial sample detection and robustness optimization strategy to detect abnormal inputs;

[0105] Step 4.1: Adopt an adversarial sample detection method. Let the adversarial sample detection function be:

[0106]

[0107] Among them, is the adversarial sample detection function value, E(p) is the corrected data vector generated by integrating the results of neural-logical reasoning and variational autoencoder, i is the adversarial candidate sample, I is the set of adversarial candidates, is the detected input feature;

[0108] Step 4.2: Adopt a robustness optimization strategy. Let the robustness optimization function be:

[0109]

[0110] Among them, Θ r (ψ) is the robustness optimization function value, E(p) is the corrected data vector generated by integrating the results of neural-logical reasoning and variational autoencoder, j is the robustness candidate state, J is the set of robustness candidates, and ψ is the robustness optimization input;

[0111] Step 4.3: Integrate the results of adversarial sample detection and robustness optimization. Let the security assessment output function be:

[0112]

[0113] Among them, Σ(μ) is the final security evaluation value, ξ1 is the anti-detection weight factor, ξ2 is the robust optimization weight factor, μ is the security evaluation input variable, and ψ is the robust optimization input;

[0114] Step 5: Based on the intelligent human-computer interaction interface, display the corrected data and the reasoning process, provide an interactive feedback mechanism, enable users to adjust the data correction rules and model parameters, and optimize the data processing flow;

[0115] Step 5.1, construct an interactive display function, and set the interactive display function as:

[0116] D(ω) = τ1·Σ(μ) + τ2·E(p),

[0117] Among them, D(ω) represents the interactive display output, Σ(μ) is the final security evaluation value, E(p) is the corrected data vector generated by fusing neural logic reasoning and variational autoencoder results, τ1 represents the security evaluation display weight factor, τ2 represents the data correction display weight factor, and ω represents the interface control parameter;

[0118] Step 5.2, construct an interactive feedback adjustment function, and set the feedback adjustment function as:

[0119]

[0120] Among them, f(l) represents the interactive feedback output, D(ω) represents the interactive display output, k n represents the interactive candidate adjustment state, K5 represents the interactive candidate set, and l represents the feedback control variable;

[0121] Step 5.3, construct an interactive parameter optimization function, and set the parameter optimization function as:

[0122]

[0123] Among them, represents the interactive parameter optimization output, the necessary condition correction formula R d (ζ c ) and the possibility condition correction formula R e (ζ c ), f(l) represents the interactive feedback output, η1 represents the necessary condition correction weight factor, η2 represents the possibility condition correction weight factor, η3 represents the interactive feedback weight factor, and ζ represents the optimization control variable.

[0124] Benefits of Step 1: By adopting the knowledge graph model construction technology, it realizes the accurate characterization of data entities and relationships, improving the accuracy of data semantic expression; obtaining high-dimensional feature representations through graph embedding learning to ensure the stability of the data structure after mapping to the vector space; the first-order predicate logic rules provide a strict semantic guarantee for data constraints, effectively solving the problem of loose traditional rules.

[0125] Benefits of Step 2: Using S5 modal logic reasoning to construct data correction constraints, clearly defining the conditions of necessity and possibility, ensuring the logical rigor of the data correction process; the correction values obtained through Kripke semantic calculation make the data adjustment accurate, helping to compensate for the deficiency of insufficient logical constraints in traditional methods.

[0126] Benefits of Step 3: Combining neural logic reasoning and variational autoencoder technology to achieve intelligent optimization of data; the neural-symbolic reasoning model learns correction rules, making full use of the advantages of deep learning and logical reasoning; variational inference generates corrected data that conforms to logical constraints, improving data processing efficiency and correction accuracy, and making up for the problem of a single data optimization process in traditional methods.

[0127] Benefits of Step 4: Adopting adversarial sample detection and robustness optimization strategies to conduct security evaluations on the corrected data; identifying abnormal inputs in advance to ensure the reliability and stability of the data in actual applications, effectively preventing potential attacks, and overcoming the problems of untimely anomaly detection and system vulnerability in existing technologies.

[0128] Benefits of Step 5: Based on an intelligent human-computer interaction interface to display the data correction and reasoning process, providing instant interactive feedback; interactive display, feedback adjustment, and parameter optimization functions enable users to adjust data correction rules and model parameters in real time; the overall design improves the flexibility of the data processing process and the user experience, avoiding the deficiencies of single display results and slow feedback response in traditional methods.

[0129] In summary, the present invention constructs a set of data analysis and processing methods with high precision, high robustness, and high user participation through knowledge graph construction, S5 modal logic reasoning, neural logic reasoning and variational autoencoder optimization, security evaluation, and interactive feedback mechanism. Each step is interconnected, and the overall solution effectively makes up for the problems existing in the prior art in data semantic understanding, logical constraints, anomaly detection, and interactive feedback, realizing the intelligence and high efficiency of the data processing process.

[0130] The intelligent human-computer interaction interface uses a graph neural network for data relationship visualization. Users can adjust the data correction logic and view the semantic consistency analysis report of the corrected data in real time.

[0131] The data analysis and processing method based on intelligent AI adjusts the correction strategies for different data sets through adaptive parameters, improving cross-domain applicability.

[0132] A terminal device, which includes at least one set of processing units, is used to execute the steps of a data analysis and processing method based on intelligent AI to complete a data correction operation, and provides an interactive operation for data correction through a user interface.

[0133] The terminal device adopts advanced data processing algorithms to accurately execute data correction operations, overcoming the problem of low accuracy in complex data processing of traditional devices; it integrates an interactive user interface to provide real-time feedback on data correction results, making up for the shortcoming of the old system that can only display statically and lacks interactivity; the efficient processing units support parallel computing, greatly improving the data processing speed and solving the limitations of slow response and high latency.

[0134] A storage medium stores a computer-executable program, which can implement a data analysis and processing method based on intelligent AI to complete a data correction operation, and provides dynamic correction feedback through an intelligent human-machine interaction interface.

[0135] The computer-executable program stored in the storage medium can implement an intelligent AI data analysis and processing method, automatically execute data correction operations, improve the accuracy and efficiency of the correction process, and avoid errors during manual intervention; the program provides dynamic correction feedback through an intelligent human-machine interaction interface, allowing users to adjust correction rules and parameters in real time, enhancing the flexibility of the system and the user's sense of control; the stored program supports diverse data processing tasks and has efficient algorithm optimization to ensure that complex data correction tasks can respond quickly, solving the problem of slow response speed of traditional processing methods.

[0136] Although the embodiments of the present invention have been shown and described, for those of ordinary skill in the art, it can be understood that various changes, modifications, substitutions, and variations can be made to these embodiments without departing from the principles and spirit of the present invention. The scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A data analysis and processing method based on intelligent AI, characterized in that: include: Step 1: Build a knowledge graph model, model data entities and relationships, use graph embedding methods to map the knowledge graph to vector space, and use first-order predicate logic rules to constrain the semantic relationship of data; Step 2: Based on the semantic relationship of the knowledge graph, use S5 modal logic reasoning to construct data correction constraints, define the necessary and possible conditions of the data, and calculate the correction value of the data point under the logical constraints through Kripke semantics; Step 3: Based on the data correction constraints, neural logical reasoning and variational autoencoder are combined to optimize data, and the neural symbolic reasoning model is trained to learn the correction rules, and the correction data that conforms to the logical constraints is generated through variational inference; Step 4: Perform security assessment on the corrected data and use adversarial sample detection and robustness optimization strategies to detect abnormal inputs; Step 5: Display the corrected data and reasoning process based on the intelligent human-computer interaction interface, and provide an interactive feedback mechanism so that users can adjust data correction rules and model parameters to optimize the data processing process.

2. According to the data analysis and processing method based on intelligent AI according to claim 1, it is characterized in that: In the step S1, when the data standardization condition is met and the data comes from a preprocessed structured database, the step further includes: Step 1.1: Build a basic knowledge graph model, assuming that the data entity set is ε a And the relation set is R b The knowledge graph is defined as: KG={(ε,ρ,γ)}, Among them, KG is a knowledge graph set, ε is a data entity, ρ is the semantic association between data, and γ is another data entity; Step 1.2, perform graph embedding learning, and use the TransE method to construct the embedding function defined as: F=∥Λ·α+β-Λ·δ∥, Where F is the output of the graph embedding function, Λ is the embedding transformation matrix, α is the vector representation of the first data entity in the feature space, β is the relationship vector, and δ is the vector representation of the second data entity in the feature space; Step 1.3, establish data semantic constraints, and use first-order predicate logic rules to constrain data entities and their relationships based on the knowledge graph constructed in step 1.1 and the output of the embedded function in step 1.

2. The logical constraint expression is defined as: C={φ(ε,ρ,γ)∣φ:ε→γ}, Among them, C is the logical constraint set and φ is the logical mapping rule.

3. The data analysis and processing method based on intelligent AI according to claim 1 is characterized in that: In the step S2, under the condition of referencing the data semantic constraint expression C obtained in the step S1, the following is further included: Step 2.1, define the necessity and possibility operators, let the necessity operator be B a , let the probability operator be P b , the semantic constraint of the data to be corrected is denoted as ζ c ; Step 2.2, construct the necessary condition correction formula, according to the definition of S5 modal logic, assume that the necessity logic constraint satisfies Introducing the correction function R d , the definition formula is: Among them, U is the candidate set of data correction, and u is the specific correction status; Step 2.3, construct the possibility condition correction formula, according to the S5 modal logic definition, assume that the possibility logic constraint is satisfied, and introduce the correction function R e , the definition formula is: Among them, V is the data correction candidate set, and v is the correction candidate state.

4. The data analysis and processing method based on intelligent AI according to claim 1, characterized in that: In step S3, the necessary condition correction formula R in step S2 is cited. d (ζ c ) and the possibility condition correction formula R e (ζ c ) Results further include: Step 3.1, use the neural logic reasoning training model and set the neural theorem proving function as: Among them, x is the output corrected data feature vector, y is the corrected data vector output by the model, k is the neural network weight matrix, λ is the bias vector, and Y is the corrected data candidate set; Step 3.2, use variational autoencoder to optimize data, and set the function for generating corrected data as: Among them, m is the output candidate vector of data correction to be generated, n is the latent variable, P(m|n) is the conditional probability of data generation, Q(n|m) is the conditional probability of data encoding, P′(n) is the prior probability, Δ is the indicator to measure the difference between the data encoding probability and the prior probability, and N is the candidate set of latent variables; Step 3.3, fuse the neural logic reasoning and variational autoencoder results, and set the fusion output function to be: E(p)=ω1·Φ n (q)+ω2·Γ v (r), Among them, p is the final generated corrected data vector, Φ n (q) is the output of the neural symbolic reasoning model, Γ v (r) is the data generated by the variational autoencoder, ω1 is the neural logic reasoning weight factor, ω2 is the variational autoencoder weight factor, q is the input data feature vector, and r is the variational autoencoder input data vector.

5. The data analysis and processing method based on intelligent AI according to claim 1, characterized in that: In the step S4, the fusion output function E(p) result in the referenced step S3 further comprises: Step 4.1: Use the adversarial sample detection method and set the adversarial sample detection function as: in, is the adversarial sample detection function value, E(p) is the modified data vector generated by integrating the results of neural logic reasoning and variational autoencoder, i is the adversarial candidate sample, I is the adversarial candidate set, Input features for detection; Step 4.2, adopt the robust optimization strategy, and set the robust optimization function as: Among them, Θ r (ψ) is the robust optimization function value, E(p) is the corrected data vector generated by integrating the results of neural logic reasoning and variational autoencoder, j is the robust candidate state, J is the robust candidate set, and ψ is the robust optimization input; Step 4.3, integrate the adversarial sample detection and robustness optimization results, and set the security assessment output function as: Among them, Σ(μ) is the final security assessment value, ξ1 is the anti-detection weight factor, ξ2 is the robust optimization weight factor, μ is the security assessment input variable, and ψ is the robustness optimization input.

6. The data analysis and processing method based on intelligent AI according to claim 1, characterized in that: In the step S5, the safety assessment output function Σ(μ) in step S4 and the fusion output function E(p) result in step S3 are further included: Step 5.1, construct an interactive display function, assuming that the interactive display function is: D(ω)=τ1·Σ(μ)+τ2·E(p), Where D(ω) represents the interactive display output, Σ(μ) is the final safety assessment value, E(p) is the corrected data vector generated by integrating the results of neural logic reasoning and variational autoencoder, τ1 represents the safety assessment display weight factor, τ2 represents the data correction display weight factor, and ω represents the interface control parameter; Step 5.2, construct an interactive feedback adjustment function, assuming that the feedback adjustment function is: Where f(l) represents the interactive feedback output, D(ω) represents the interactive display output, and k n represents the interaction candidate adjustment state, K5 represents the interaction candidate set, and l represents the feedback control variable; Step 5.3, construct an interactive parameter optimization function, assuming that the parameter optimization function is: in, Represents the interactive parameter optimization output, necessary condition correction formula R d (ζ c ) and the possibility condition correction formula R e (ζ c ), f(l) represents the interactive feedback output, η1 represents the necessary condition correction weight factor, η2 represents the possibility condition correction weight factor, η3 represents the interactive feedback weight factor, and ζ represents the optimization control variable.

7. The data analysis and processing method based on intelligent AI according to claim 6, characterized in that: The intelligent human-computer interaction interface uses a graph neural network to visualize data relationships. Users can adjust data correction logic and view the semantic consistency analysis report of the corrected data in real time.

8. The data analysis and processing method based on intelligent AI according to claim 1, characterized in that: The data analysis and processing method based on intelligent AI improves cross-domain applicability by adjusting the correction strategies of different data sets through adaptive parameters.

9. A terminal device, characterized in that: The terminal device includes at least one group of processing units for executing the steps of the intelligent AI-based data analysis and processing method according to claims 1 to 8 to complete the data correction operation and provide interactive operations of data correction through a user interface.

10. A storage medium, characterized in that: The storage medium stores a computer executable program, which, when executed, can implement the intelligent AI-based data analysis and processing method according to claims 1 to 8 to complete data correction operations and provide dynamic correction feedback through an intelligent human-computer interaction interface.

Images