AI information processing system based on artificial intelligence

Abstract

The invention relates to the technical field of information, and discloses an artificial intelligence-based AI information processing system, which comprises a data acquisition module, a data preprocessing module, a data feature selection module, a data analysis module, an AI model training module, a user interface module and a safety protection module, the data collection module collects data sets of the data preprocessing module, the model training module and the user interface module through a network, the data analysis module carries out calculation and analysis, the AI model training module carries out model parameter adjustment according to an analysis result, and the user interface module feeds back user opinions to the system through a user interface according to an adjusted AI model. The safety protection module is used for protecting the data acquisition module, the data preprocessing module, the data feature selection module, the data analysis module, the AI model training module and the user interface module according to an analysis result, and the system periodically detects and calculates deviation in an integrated model through the process, so that the problem of data deviation is solved in practical application.

Description

Technical Field

[0001] The present invention relates to the field of information technology, and particularly to an AI information processing system based on artificial intelligence. Background Art

[0002] In today's digital age, artificial intelligence (AI) technology has been widely applied in various fields, including data processing and storage support services, information technology consulting services, professional design services, software sales and outsourcing services, network and information security software development, animation and game development, geographic remote sensing information services, Internet security services, artificial intelligence application software development, artificial intelligence industry application system integration services, information system integration services, artificial intelligence basic software development, artificial intelligence double-creation service platforms, artificial intelligence public data platforms, advertising design agencies, technology services and development, technology consulting and exchanges, technology transfer and promotion, cloud computing equipment manufacturing and sales, Internet data services, wholesale of computer hardware and software and auxiliary equipment, Internet sales (except for goods that require permission for sales), artificial intelligence general application systems, software development, artificial intelligence basic resources and technology platforms, and technical consulting services for artificial intelligence public service platforms, etc. However, despite the many conveniences and efficiency improvements brought by AI technology, there are still some significant deficiencies in its existing technologies.

[0003] Data bias is also a challenge faced by AI algorithms. AI algorithms require a large amount of data for training, but if this data is biased, then the AI model may be affected. For example, when using a face recognition algorithm, if the dataset only contains specific types of faces, then the algorithm may not be able to accurately recognize other types of faces. This data bias may lead to unfair or discriminatory results in the actual application of AI systems, thus triggering social ethics and legal issues.

[0004] In summary, although AI technology shows great potential in multiple fields, deficiencies such as poor interpretability and data bias still need to attract our high attention. In future research and applications, we need to continuously explore and improve AI technology to enhance its transparency, fairness, and reliability, so as to better serve the development of human society. Summary of the Invention

[0005] (1) Technical Problems to be Solved

[0006] Aiming at the deficiencies of the existing technology, the present invention provides an AI information processing system based on artificial intelligence, which has the advantages of being able to automatically identify and correct biases in data, ensuring the fairness and reliability of data analysis results, and solving the problem of data bias in the existing technology.

[0007] (2) Technical Solutions

[0008] To achieve the above object, the present invention provides the following technical solution: An AI information processing system based on artificial intelligence, including a data acquisition module, a data preprocessing module, a data feature selection module, a data analysis module, an AI model training module, a user interface module, and a security protection module;

[0009] The data acquisition module collects the data sets of the data preprocessing module, the model training module, and the user interface module through the network. The data preprocessing module obtains the data set of the data preprocessing module through a program for cleaning and converting formats. The model training module combines multiple models through an ensemble learning method to predict the data set of the model training module. The user interface module obtains the data set of the user interface module by collecting user feedback, performing manual review, and establishing a feedback mechanism, and statistically numbers and indexes the data sets of the data preprocessing module, the model training module, and the user interface module;

[0010] Before training the AI model, the data preprocessing module discovers and corrects data inconsistencies or imbalances by examining and analyzing the data set of the data acquisition module;

[0011] The data feature selection module extracts useful feature data from the data set processed by the data preprocessing module, reduces the dimension, and improves the model efficiency;

[0012] The data analysis module calculates and analyzes the data collected by the data acquisition module;

[0013] The AI model training module adjusts the model parameters according to the analysis results;

[0014] The user interface module feeds back the user's opinions to the system through the user interface for the adjusted AI model;

[0015] The security protection module protects the data acquisition module, the data preprocessing module, the data feature selection module, the data analysis module, the AI model training module, and the user interface module according to the analysis results.

[0016] Preferably, the data preprocessing module numbers the minority group data set and the basic data set according to the characteristics of the data set of the data preprocessing module. The numbers of the minority group data set are R1, R2, R3,... R n , and the numbers of the basic data set are S1, S2, S3,... S m .

[0017] Preferably, the model training module numbers the prediction results of the model according to the characteristics of the data set of the model training module. The numbers of the prediction results of the model are o1, o2, o3,... o n .

[0018] Preferably, the user interface module numbers the prediction results after model feedback adjustment according to the characteristics of the user interface module dataset. The prediction results after model feedback adjustment are numbered as q1, q2, q3, … q n .

[0019] Preferably, the data analysis module includes a diverse data source acquisition unit, an AI model integrated prediction unit, and an AI data deviation processing unit. The diverse data source acquisition unit calculates the increased sample data source Hj according to the dataset of the data preprocessing module. The AI model integrated prediction unit calculates the AI model integrated prediction result Gz according to the dataset of the model training module. The AI data deviation processing unit calculates the data deviation Pl after AI model update according to the dataset of the model training module and the dataset of the user interface module.

[0020] Preferably, the diverse data source acquisition unit calculates the increased sample data source Hj according to the dataset of the data preprocessing module, and its calculation formula is:

[0021] Hj = (R1, R2, R3, … R n ) ∪ (S1, S2, S3, … S m )

[0022] In the formula, Hj represents the increased sample data source, R1, R2, R3, … R n represents the minority group dataset, n represents the types of minority group data, S1, S2, S3, … S m represents the basic dataset, and m represents the types of basic data.

[0023] Preferably, the AI model integrated prediction unit calculates the AI model integrated prediction result Gz according to the dataset of the model training module, and its calculation formula is:

[0024]

[0025] In the formula, Gz represents the AI model integrated prediction result, o1, o2, o3, … o n represents the prediction result of the model, o i represents the prediction result of the i-th model, w1, w2, w3, … w n represents the weight of the prediction result of the model, n represents the number of models counted, w i represents the weight of the prediction result of the i-th model.

[0026] Preferably, the AI data deviation processing unit calculates the data deviation Pl after AI model update according to the dataset of the model training module and the dataset of the user interface module, and its calculation formula is:

[0027]

[0028] In the formula, Pl represents the data deviation after the AI model is updated, o1, o2, o3, … o n represents the prediction result of the model, o i represents the prediction result of the i-th model, n represents the number of models counted, represents the average value of the prediction results of n models, q1, q2, q3, … q n represents the prediction result after the model feedback adjustment, q i represents the prediction result after the feedback adjustment of the i-th model, represents the average value of the prediction results after the feedback adjustment of n models, represents the data deviation of the old model, represents the data deviation of the model adjusted according to the feedback, and a represents the learning rate.

[0029] Preferably, the AI model training module adjusts the model parameters of the AI model according to the increased sample data source Hj. When new sample data is added to the system, the system uses the new data to update and adjust the parameters of the AI model, adjusts or retrains the sub-models according to the integrated prediction result Gz of the AI model, and re-evaluates the feature engineering, adjusts the learning rate, changes the model structure or introduces new regularization techniques according to the data deviation Pl after the AI model is updated.

[0030] Preferably, the security protection module performs privacy protection on the data acquisition module, data preprocessing module, data feature selection module, data analysis module, AI model training module and user interface module according to the increased sample data source Hj.

[0031] Compared with the prior art, the present invention provides an AI information processing system based on artificial intelligence, having the following beneficial effects:

[0032] 1. By calculating the increased sample data source Hj, the present invention enables the system to introduce diverse data sources in the dynamic data acquisition function. Through the increased data sources, the model can learn more extensive features and patterns, thereby improving its performance on unseen data. Moreover, diverse data sources not only help balance the biases in the training data, reduce prediction errors caused by insufficient and unbalanced data, but also provide information from different perspectives, making the model more robust. In applications involving sensitive or important decisions, such as in the financial, medical, and legal fields, diverse data can improve the accuracy of decisions. BRIEF DESCRIPTION OF THE DRAWINGS

[0033] Figure 1 It is a schematic structural diagram of the present invention. DETAILED DESCRIPTION OF THE INVENTION

[0034] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0035] Please refer to Figure 1 , an AI information processing system based on artificial intelligence, including a data acquisition module, a data preprocessing module, a data feature selection module, a data analysis module, an AI model training module, a user interface module, and a security protection module;

[0036] The data acquisition module collects the data sets of the data preprocessing module, the model training module, and the user interface module through the network. The data preprocessing module obtains the data set of the data preprocessing module through a program for cleaning and converting formats. The model training module combines multiple models through an ensemble learning method to predict the data set of the model training module. The user interface module obtains the data set of the user interface module by collecting user feedback, performing manual reviews, and establishing a feedback mechanism, and statistically numbers the data sets of the data preprocessing module, the model training module, and the user interface module;

[0037] Before training the AI model, the data preprocessing module discovers and corrects data inconsistencies or imbalances (collecting diverse face images to reduce biases in terms of race and gender) by examining and analyzing the data set of the data acquisition module;

[0038] The data feature selection module extracts useful feature data from the data set processed by the data preprocessing module, reduces the dimension, and improves the model efficiency (for example, in text analysis, using the bag-of-words model or the TF-IDF method to extract keywords);

[0039] The data analysis module calculates and analyzes the data collected by the data acquisition module;

[0040] The AI model training module adjusts the model parameters according to the analysis results;

[0041] The user interface module feeds back the user's opinions to the system through the user interface with the adjusted AI model;

[0042] The security protection module protects the data acquisition module, the data preprocessing module, the data feature selection module, the data analysis module, the AI model training module, and the user interface module according to the analysis results.

[0043] The data preprocessing module numbers the minority group dataset and the basic dataset according to the dataset characteristics of the data preprocessing module. The numbers of the minority group dataset are R1, R2, R3, … R n , and the numbers of the basic dataset are S1, S2, S3, … S m .

[0044] The model training module numbers the prediction results of the model according to the dataset characteristics of the model training module. The numbers of the prediction results of the model are o1, o2, o3, … o n .

[0045] The user interface module numbers the prediction results after model feedback adjustment according to the dataset characteristics of the user interface module. The numbers of the prediction results after model feedback adjustment are q1, q2, q3, … q n .

[0046] The data analysis module includes a diversity data source acquisition unit, an AI model integrated prediction unit, and an AI data deviation processing unit. The diversity data source acquisition unit calculates the increased sample data source Hj according to the dataset of the data preprocessing module. The AI model integrated prediction unit calculates the AI model integrated prediction result Gz according to the dataset of the model training module. The AI data deviation processing unit calculates the data deviation Pl after the AI model is updated according to the datasets of the model training module and the user interface module.

[0047] The diversity data source acquisition unit calculates the increased sample data source Hj according to the dataset of the data preprocessing module, and its calculation formula is:

[0048] Hj = (R1, R2, R3, … R n ) ∪ (S1, S2, S3, … S m )

[0049] In the formula, Hj represents the increased sample data source, R1, R2, R3, … R n represents the minority group dataset, n represents the types of minority group data, S1, S2, S3, … S m represents the basic dataset, and m represents the types of basic data.

[0050] The advantages are as follows: By calculating the increased sample data source Hj, the system introduces diverse data sources in the dynamic data collection function. Through the added data sources, the model can learn a wider range of features and patterns, thereby improving its performance on unseen data. Moreover, diverse data sources not only help balance the biases in the training data, reduce prediction errors caused by insufficient and unbalanced data, but also provide information from different perspectives, making the model more robust. In applications involving sensitive or important decisions, such as in the financial, medical, and legal fields, diverse data can improve the accuracy of decisions. At the same time, it ensures that the collection of the increased sample data source complies with relevant privacy regulations and ethical standards to protect the privacy rights and interests of users.

[0051] The AI model integration prediction unit calculates the AI model integration prediction result Gz based on the dataset of the model training module, and its calculation formula is:

[0052]

[0053] In the formula, Gz represents the AI model integration prediction result, o1, o2, o3,... o n represents the prediction result of the model, o i represents the prediction result of the i-th model, w1, w2, w3,... w n represents the weight of the prediction result of the model, n represents the number of models counted, w i represents the weight of the prediction result of the i-th model.

[0054] The advantages are as follows: By calculating the AI model integration prediction result Gz, combining the prediction results of n models helps reduce the biases and errors that a single model may bring, thereby improving the overall prediction accuracy of the system model. At the same time, different weights are assigned according to the performance and accuracy of each model, so that models with better and more accurate prediction performance in the system account for a larger proportion in the final prediction to reflect the importance of accurate models. When using cross-validation techniques to evaluate the performance of each model, the system can automatically select models with stronger generalization ability.

[0055] The AI data deviation processing unit calculates the data deviation Pl after the AI model is updated based on the dataset of the model training module and the dataset of the user interface module, and its calculation formula is:

[0056]

[0057] In the formula, Pl represents the data deviation after the AI model is updated, o1, o2, o3,... o n represents the prediction result of the model, o i represents the prediction result of the i-th model, n represents the number of models counted, Represents the average of the prediction results of n models, q1, q2, q3, … q n Represents the prediction result after the model feedback adjustment, q i Represents the prediction result after the feedback adjustment of the i-th model, Represents the average of the prediction results after the feedback adjustment of n models, Represents the data deviation of the old model, Represents the data deviation of the model adjusted according to the feedback. a represents the learning rate.

[0058] The advantages are as follows: By calculating the data deviation Pl after the update of the AI model, the system regularly detects and calculates the deviation in the integrated model, and adjusts the model weights or introduces new models as needed to correct the deviation. The AI data deviation processing unit can not only assist the system in improving the accuracy of prediction, but also ensure the fairness and reliability of the data analysis results, thus solving the problem of data deviation in practical applications.

[0059] The AI model training module adjusts the model parameters of the AI model according to the increased sample data source Hj. When new sample data is added to the system, the system uses the new data to update and adjust the parameters of the AI model (so as to maintain the accuracy and relevance of the model), adjusts or retrains the sub-models according to the integrated prediction result Gz of the AI model, and re-evaluates the feature engineering, adjusts the learning rate, changes the model structure or introduces new regularization techniques according to the data deviation Pl after the update of the AI model.

[0060] The security protection module performs privacy protection on the data acquisition module, data preprocessing module, data feature selection module, data analysis module, AI model training module and user interface module according to the increased sample data source Hj. The specific process is as follows:

[0061] Data acquisition module: For the increased sample data source Hj, it is necessary to ensure that the newly acquired data complies with privacy policies and legal regulations. Therefore, the principle of data minimization is implemented, only the necessary data is collected, and sensitive data is encrypted;

[0062] Data preprocessing module: For the increased sample data source Hj, protect the confidentiality and integrity of the data during the data cleaning and conversion process;

[0063] Data feature selection module: Avoid using features that may lead to discrimination or unfair treatment, and regularly update the feature selection criteria to ensure compliance with privacy regulations and ethical standards;

[0064] Data analysis module: Continue to protect the confidentiality and integrity of the data during the analysis phase;

[0065] AI model training module: Ensure that the model training activities comply with privacy regulations, especially when involving personal sensitive information;

[0066] User interface module: enables users to easily access and manage their data privacy settings.

[0067] Although the embodiments of the present invention have been shown and described, it will be understood by those of ordinary skill in the art that various changes, modifications, substitutions, and variations can be made to these embodiments without departing from the principles and spirit of the present invention, and the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. An AI information processing system based on artificial intelligence, characterized in that: It includes data acquisition module, data preprocessing module, data feature selection module, data analysis module, AI model training module, user interface module and security protection module; The data acquisition module collects data sets of the data preprocessing module, the model training module and the user interface module through the network; the data preprocessing module obtains the data set of the data preprocessing module through a cleaning and format conversion program; the model training module combines multiple models to predict the model training module data set through an integrated learning method; the user interface module obtains the user interface module data set by collecting user feedback, conducting manual review and establishing a feedback mechanism, and the data sets of the data preprocessing module, the model training module and the user interface module are counted and numbered; The data preprocessing module detects and corrects data inconsistencies or imbalances by reviewing and analyzing the data set of the data acquisition module before training the AI ​​model; The data feature selection module extracts useful feature data from the data set processed by the data preprocessing module, reduces the dimension and improves the model efficiency; The data analysis module calculates and analyzes the data collected by the data collection module; The AI ​​model training module adjusts model parameters according to the analysis results; The user interface module feeds back the user opinions to the system through the user interface using the adjusted AI model; The security protection module protects the data acquisition module, data preprocessing module, data feature selection module, data analysis module, AI model training module and user interface module according to the analysis results.

2. The AI ​​information processing system based on artificial intelligence according to claim 1, characterized in that: The data preprocessing module performs data numbering on the minority group data set and the basic data set according to the data set characteristics of the data preprocessing module, and the minority group data set is numbered R1, R2, R3, ... R n , the basic data sets are numbered S1, S2, S3, ...S m .

3. The AI ​​information processing system based on artificial intelligence according to claim 1, characterized in that: The model training module numbers the prediction results of the model according to the characteristics of the model training module data set, and the prediction results of the model are numbered o1, o2, o3, ... o n .

4. The AI ​​information processing system based on artificial intelligence according to claim 1, characterized in that: The user interface module performs data numbering on the prediction results adjusted by the model feedback according to the characteristics of the user interface module data set, and the prediction results adjusted by the model feedback are numbered as q1, q2, q3, ...q n .

5. The AI ​​information processing system based on artificial intelligence according to claim 1, characterized in that: The data analysis module includes a diversity data source acquisition unit, an AI model integrated prediction unit and an AI data deviation processing unit. The diversity data source acquisition unit calculates the added sample data source Hj according to the data preprocessing module data set, the AI ​​model integrated prediction unit calculates the AI ​​model integrated prediction result Gz according to the model training module data set, and the AI ​​data deviation processing unit calculates the data deviation Pl after the AI ​​model is updated according to the model training module data set and the user interface module data set.

6. The AI ​​information processing system based on artificial intelligence according to claim 5, characterized in that: The diversity data source acquisition unit calculates the added sample data source Hj according to the data set of the data preprocessing module, and the calculation formula is: <h2 style=";text-align:left;direction:ltr">Hj = R1, R2, R3, …R<h2 style=";text-align:left;direction:ltr"> n <h2 style=";text-align:left;direction:ltr"> ∪S1, S2, S3,…S<h2 style=";text-align:left;direction:ltr"> m In the formula, Hj represents the added sample data source, R1, R2, R3, ...R n represents the minority data set, n represents the type of minority data, S1, S2, S3, ...S m represents the basic data set, and m represents the basic data type.

7. The AI ​​information processing system based on artificial intelligence according to claim 5, characterized in that: The AI ​​model integration prediction unit calculates the AI ​​model integration prediction result Gz according to the model training module data set, and its calculation formula is: In the formula, Gz represents the prediction result of AI model integration, o1, o2, o3, ...o n Represents the prediction result of the model, o i represents the prediction result of the i-th model, w1, w2, w3, ... w n represents the weight of the model's prediction results, n represents the number of statistical models, and w i Represents the weight of the prediction result of the i-th model.

8. The AI ​​information processing system based on artificial intelligence according to claim 5, characterized in that: The AI ​​data deviation processing unit calculates the data deviation Pl after the AI ​​model is updated according to the model training module data set and the user interface module data set, and the calculation formula is: In the formula, Pl represents the data deviation after the AI ​​model is updated, o1, o2, o3, ...o n Represents the prediction result of the model, o i represents the prediction result of the i-th model, n represents the number of statistical models, Represents the average of the prediction results of n models, q1, q2, q3, ...q n represents the prediction result after model feedback adjustment, q i represents the prediction result after the feedback adjustment of the i-th model, It represents the average of the prediction results after feedback adjustment of n models. represents the data bias of the old model, It indicates the data bias of adjusting the model according to feedback, and a indicates the learning rate.

9. The AI ​​information processing system based on artificial intelligence according to claim 8, characterized in that: The AI ​​model training module adjusts the model parameters of the AI ​​model according to the added sample data source Hj. When new sample data is added to the system, the system will use the new data to update and adjust the parameters of the AI ​​model, adjust or retrain the sub-model according to the AI ​​model integrated prediction result Gz, and re-evaluate feature engineering, adjust the learning rate, change the model structure or introduce new regularization technology according to the data deviation Pl after the AI ​​model is updated.

10. The AI ​​information processing system based on artificial intelligence according to claim 9, characterized in that: The security protection module performs privacy protection on the data acquisition module, data preprocessing module, data feature selection module, data analysis module, AI model training module and user interface module according to the added sample data source Hj.

Images