AI fool-proofing method and system

Abstract

The present invention provides an AI fool-proofing method and system, and relates to the field of fool-proofing control technology. The present invention combines real-time operation data, uses a primary fool-proofing monitoring model to intervene in obvious misoperation behaviors, combines current operation behavior sequences and historical operation behavior sequences, and uses an advanced fool-proofing monitoring model to perform more detailed monitoring and intervention on misoperation behaviors. In particular, the advanced fool-proofing monitoring model uses deep learning technologies such as long short-term memory networks, which can capture complex patterns in time series, thereby improving the accuracy of misoperation prediction. While improving the judgment of current misoperation behaviors, it can also predict misoperation behaviors that may occur in the future, achieve early warning, and reduce the probability of misoperation. The combination of primary intervention and advanced intervention provides multi-level security protection, and the two complement each other and work together to reduce the risk of misoperation.

Description

Technical Field

[0001] The present invention relates to the field of fool-proof control technology, and in particular to an AI fool-proof method and system. Background Art

[0002] With the rapid development of information technology, various intelligent devices and systems have been widely used in daily life and industrial production. Users use these intelligent systems to perform various operations, from simple information queries to complex industrial control processes. However, in the process of use, due to the negligence of users, unfamiliarity with the operation process or the complexity of the system itself, misoperation often occurs. Misoperation may lead to data loss, equipment damage, reduced production efficiency and even safety accidents, causing immeasurable losses to individuals, enterprises and even society. Traditional fool-proofing measures often rely on manual inspection and simple rule-based automated inspection. These methods are not only inefficient, but also these rules are often inflexible and difficult to cope with the increasingly complex operating environment and emerging threats. Once a new situation that is not covered by the existing rules appears, misoperation is prone to occur and cannot be effectively prevented in a timely manner. Moreover, based on the existing rules, it is impossible to predict possible misoperations that have not occurred, thereby achieving the function of early prevention.

[0003] Therefore, it is necessary to provide an AI fool-proofing method and system to solve the above technical problems. Summary of the invention

[0004] In order to solve the above technical problems, the present invention provides an AI fool-proofing method and system, which can achieve the beneficial effects of flexibly and efficiently blocking the occurrence of erroneous operations and preventing the occurrence of erroneous operations in advance.

[0005] The present invention provides an AI fool-proofing method, the fool-proofing method comprising the following steps:

[0006] S1: Obtain the real-time operation data of the user's operation target, the current operation behavior sequence and the historical operation behavior sequence;

[0007] S2: In the process of error prevention, based on real-time operation data, a pre-trained primary error prevention monitoring model is used to perform primary intervention on the erroneous operation behavior, and based on the current operation behavior sequence and the historical operation behavior sequence, a pre-trained advanced error prevention monitoring model is used to perform advanced intervention on the erroneous operation behavior, wherein the erroneous operation behavior is an operation behavior that is contrary to the current operation state of the user's operation target;

[0008] S3: Combine the output results after the primary intervention and the output results after the advanced intervention, and feed back to the user through the user interface of the user operation target.

[0009] Preferably, the current operation behavior sequence is data of all operation behaviors executed within a preset time window including the time when the current operation behavior occurs, and the historical operation behavior sequence is data of all operation behaviors executed within a time window in the past.

[0010] Preferably, the primary intervention comprises the following steps:

[0011] Obtain real-time operation data of user operation targets;

[0012] Based on the acquired real-time operation data, the primary error-proofing monitoring model is used to identify the set of operation behaviors that are contrary to the current user operation target operation status;

[0013] Based on the set of conflicting operational behaviors, the corresponding intervention measures are determined and executed in combination with the predefined intervention measure mapping table.

[0014] Preferably, the advanced intervention comprises the following steps:

[0015] Preprocessing the acquired current operation behavior sequence, and extracting key features from the preprocessed current operation behavior sequence;

[0016] The extracted key features, current operation behavior sequence, and historical operation behavior sequence are input into the pre-trained advanced error-proofing monitoring model to obtain the current risk score of the current operation behavior and the predicted risk score of the potential future erroneous operation behavior;

[0017] Intervene in current operating behaviors based on current risk scores, and intervene in potential future misoperation behaviors based on predicted risk scores.

[0018] Preferably, the step of preprocessing the acquired current operation behavior sequence and extracting key features therefrom comprises the following steps:

[0019] Remove repeated events, invalid data and outliers in the current operation behavior sequence through noise filtering;

[0020] Perform data cleaning on the current operation behavior sequence after noise filtering;

[0021] Extract key features of the current operation behavior sequence after data cleaning, where the key features include operation frequency, operation interval, operation sequence, user habits, and action mode;

[0022] The extracted key features are normalized to obtain normalized key features.

[0023] Preferably, the acquisition of the current risk score comprises the following steps:

[0024] Input the extracted key features into the first-level classifier of the advanced foolproof monitoring model, and preliminarily classify the key features of the current operation behavior and mark the classification results. If it is marked as normal, the current risk score of the current operation behavior is directly set to the preset normal value. If it is marked as abnormal, proceed to the next step;

[0025] The classification results marked as abnormal by the first-level classifier are input into the second-level classifier of the advanced fool-proofing monitoring model for analysis to obtain the risk score of the current operation behavior;

[0026] Input the current operation behavior sequence into the pre-trained long short-term memory network model to obtain the sequence analysis score of the current operation behavior;

[0027] The risk score of the current operation behavior and the sequence analysis score are weighted and fused using the weighted fusion method to obtain the current risk score of the current operation behavior.

[0028] Preferably, the predicted risk score is obtained by inputting the historical operation sequence into a pre-trained long short-term memory network model to obtain the predicted risk score of future potential misoperation behavior.

[0029] Preferably, the intervening in the current operation behavior based on the current risk score and intervening in the future potential misoperation behavior based on the predicted risk score comprises the following steps:

[0030] Determine whether the current operation behavior is an erroneous operation based on the current risk score and a preset current risk score threshold, and if it is an erroneous operation, take a corresponding first intervention measure to intervene;

[0031] Based on the predicted risk score and the preset predicted risk score threshold, it is determined whether the potential future misoperation behavior has the risk of misoperation. If there is a risk of misoperation, the corresponding second intervention measure is taken for pre-intervention.

[0032] Preferably, the first intervention measure and the second intervention measure both include prohibiting associated operations, pausing current operations, issuing warning prompts, delaying instruction execution, requiring secondary confirmation, and suggesting alternative operations.

[0033] An AI fool-proofing system is applied to an AI fool-proofing method, and the fool-proofing system includes:

[0034] The data collection module is used to obtain the real-time operation data of the user's operation target, the current operation behavior sequence and the historical operation behavior sequence;

[0035] A foolproofing intervention module is used to perform primary intervention on erroneous operation behaviors based on real-time operation data and using a pre-trained primary foolproofing monitoring model during the foolproofing process, and to perform advanced intervention on erroneous operation behaviors based on a current operation behavior sequence and a historical operation behavior sequence using a pre-trained advanced foolproofing monitoring model, wherein the erroneous operation behaviors are operation behaviors that are contrary to the current operation state of the user's operation target;

[0036] The result feedback module is used to merge the output results after the primary intervention and the output results after the advanced intervention, and feed back to the user through the user interface of the user operation target.

[0037] Compared with the related art, the AI ​​fool-proofing method and system provided by the present invention has the following beneficial effects:

[0038] The present invention uses a primary fool-proof monitoring model to intervene in obvious misoperation behaviors by combining real-time operation data, and uses an advanced fool-proof monitoring model to conduct more detailed monitoring and intervention on misoperation behaviors by combining the current operation behavior sequence and the historical operation behavior sequence. In particular, the advanced fool-proof monitoring model uses deep learning technologies such as long short-term memory networks, which can capture complex patterns in time series, thereby improving the accuracy of misoperation prediction. In addition to improving the judgment of current misoperation behaviors, it can also predict misoperation behaviors that may occur in the future, achieve early warning, and reduce the probability of misoperation. The combination of primary intervention and advanced intervention provides multi-level security protection. Primary intervention is based on rapid response to real-time data, while advanced intervention provides a more comprehensive risk assessment through in-depth analysis of the operation behavior sequence. The two complement each other and work together to reduce the risk of misoperation. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] Figure 1 A flowchart of an AI fool-proofing method of the present invention;

[0040] Figure 2 This is a module structure diagram of an AI fool-proofing system of the present invention. DETAILED DESCRIPTION

[0041] The present invention will be further described in detail below in conjunction with the accompanying drawings and embodiments. It is to be understood that the specific embodiments described herein are only used to explain the present invention, rather than to limit the present invention. It should also be noted that, for ease of description, only the parts related to the present invention, rather than all structures, are shown in the accompanying drawings. In addition, the embodiments of the present invention and the features in the embodiments may be combined with each other without conflict.

[0042] It should also be noted that, for ease of description, only the parts related to the present invention, but not all of the contents, are shown in the accompanying drawings. Before discussing the exemplary embodiments in more detail, it should be mentioned that some exemplary embodiments are described as processes or methods depicted as flow charts. Although the flow charts describe the operations (or steps) as sequential processes, many of the operations therein can be implemented in parallel, concurrently or simultaneously. In addition, the order of the operations can be rearranged. The process can be terminated when its operation is completed, but it can also have additional steps not included in the accompanying drawings. The process can correspond to methods, functions, procedures, subroutines, subprograms, etc.

[0043] Embodiment 1

[0044] The present invention provides an AI fool-proofing method. In a specific implementation process, Figure 1 As shown, it shows a flow chart of an AI fool-proofing method provided by the present invention, including:

[0045] Step S1: Acquire real-time operation data, current operation behavior sequence and historical operation behavior sequence of the user operation target.

[0046] Specifically, the current operation behavior sequence is data of all operation behaviors executed within a preset time window including the time when the current operation behavior occurs, and the historical operation behavior sequence is data of all operation behaviors executed within a window of a period of time in the past.

[0047] In the specific implementation process, for example, the user operation target can be a system or a device, and real-time operation data can be obtained through sensors. The real-time operation data includes but is not limited to CPU usage, memory usage, disk I / O rate, network connection status, bandwidth usage, latency, application operation status, response time, and error log. The current operation behavior sequence is the data of all operation behaviors performed within a preset time window containing the time when the current operation behavior occurs. The time window can be set according to the application scenario. For example, it can be a record of operation behaviors within the past 1 minute, 5 minutes or longer time period. The historical operation behavior sequence is the data of all operation behaviors performed within a window of a period of time in the past. According to actual needs, determine which historical operation behavior data needs to be included in the analysis scope. These data are recorded in log files or databases for system or device calls.

[0048] Step S2: In the error prevention process, primary intervention is performed on erroneous operation behaviors based on real-time operation data and using a pre-trained primary error prevention monitoring model, and advanced intervention is performed on erroneous operation behaviors based on current operation behavior sequences and historical operation behavior sequences using a pre-trained advanced error prevention monitoring model, wherein erroneous operation behaviors are operation behaviors that are contrary to the current operation status of the user's operation target.

[0049] Specifically, primary intervention includes the following steps:

[0050] Obtain real-time operation data of user operation targets;

[0051] Based on the acquired real-time operation data, the primary error-proofing monitoring model is used to identify the set of operation behaviors that are contrary to the current user operation target operation status;

[0052] Based on the set of conflicting operational behaviors, the corresponding intervention measures are determined and executed in combination with the predefined intervention measure mapping table.

[0053] In the specific implementation process, the error-proofing model is divided into a primary error-proofing monitoring model and an advanced error-proofing monitoring model. The primary error-proofing monitoring model intervenes in obvious erroneous operation behaviors. The primary error-proofing monitoring model identifies a set of operation behaviors that are inconsistent with the current user's operation target operation status based on real-time operation data. The inconsistent operation behavior set is pre-defined in the mapping table. Each inconsistent operation behavior has a corresponding disposal measure. Combined with the mapping table, the corresponding intervention measures are directly executed to prevent the occurrence of erroneous operations. The system will also update the mapping table based on the historical data of erroneous operations to expand the scope of intervention for erroneous operations.

[0054] Specifically, advanced interventions include the following steps:

[0055] Preprocessing the acquired current operation behavior sequence, and extracting key features from the preprocessed current operation behavior sequence;

[0056] The extracted key features, current operation behavior sequence, and historical operation behavior sequence are input into the pre-trained advanced error-proofing monitoring model to obtain the current risk score of the current operation behavior and the predicted risk score of the potential future erroneous operation behavior;

[0057] Intervene in current operating behaviors based on current risk scores, and intervene in potential future misoperation behaviors based on predicted risk scores.

[0058] Specifically, the obtained current operation behavior sequence is preprocessed and key features are extracted therefrom, including the following steps:

[0059] Remove repeated events, invalid data and outliers in the current operation behavior sequence through noise filtering;

[0060] Perform data cleaning on the current operation behavior sequence after noise filtering;

[0061] Extract key features of the current operation behavior sequence after data cleaning, where the key features include operation frequency, operation interval, operation sequence, user habits, and action mode;

[0062] The extracted key features are normalized to obtain normalized key features.

[0063] In the specific implementation process, in order to be able to intervene in erroneous operations and potential erroneous operations in real time, more detailed monitoring is required. First, the current operation behavior sequence obtained is preprocessed. For example, noise filtering is used to remove repeated events, invalid data and outliers in the current operation behavior sequence. Then, data cleaning, feature extraction and normalization are performed to obtain the normalized key features. The advanced fool-proofing monitoring system analyzes the extracted key features, the current operation behavior sequence and the historical operation behavior sequence to obtain the current risk score of the current operation behavior and the predicted risk score of potential erroneous operations in the future. Through these two scores, the current operation is intervened and potential erroneous operations in the future are pre-intervened.

[0064] Specifically, obtaining the current risk score includes the following steps:

[0065] Input the extracted key features into the first-level classifier of the advanced foolproof monitoring model, and preliminarily classify the key features of the current operation behavior and mark the classification results. If it is marked as normal, the current risk score of the current operation behavior is directly set to the preset normal value. If it is marked as abnormal, proceed to the next step;

[0066] The classification results marked as abnormal by the first-level classifier are input into the second-level classifier of the advanced fool-proofing monitoring model for analysis to obtain the risk score of the current operation behavior;

[0067] Input the current operation behavior sequence into the pre-trained long short-term memory network model to obtain the sequence analysis score of the current operation behavior;

[0068] The risk score of the current operation behavior and the sequence analysis score are weighted and fused using the weighted fusion method to obtain the current risk score of the current operation behavior.

[0069] In the specific implementation process, the functions of the advanced error-proofing monitoring model are divided into two parts. The first part is to analyze the current operation behavior, obtain the current risk score, and intervene in the current operation behavior according to the current risk score. The other part is to predict potential erroneous operation behaviors in the future and pre-intervene in possible erroneous operations according to the predicted risk score. Among them, the acquisition of the current risk score needs to be obtained through the two-level classifier screening of the advanced fool-proof monitoring model. First, the extracted key features are input into the first-level classifier. The model will judge whether the current operation behavior is within the normal range based on the previously trained data set. If the classification result shows that it is normal, the current risk score of the current operation behavior is directly set to the preset normal value; then, if it is marked as abnormal, it needs to be processed by the second-level classifier. The second-level classifier will conduct a more detailed analysis of the abnormal operation behavior and output a specific risk score for the current operation behavior, indicating the degree to which the behavior deviates from normal operation; secondly, the pre-trained long short-term memory network model is used. This model is particularly suitable for processing time series data and can capture the long-term dependencies between operation behaviors. The current operation behavior sequence is input into the long short-term memory network model. The model will analyze the entire sequence and give a sequence analysis score to reflect the abnormality of the current operation behavior in the entire sequence; finally, the risk score and the sequence analysis score are weighted and fused to obtain the current risk score.

[0070] Specifically, the predicted risk score is obtained by inputting the historical operation sequence into a pre-trained long short-term memory network model to obtain the predicted risk score of future potential misoperation behaviors.

[0071] In the specific implementation process, the acquisition of predicted risk scores is another part of the function of the advanced error-proofing monitoring model. The risk scores of future potential erroneous operations are predicted through the pre-trained long short-term memory network model. The historical operation sequence is input into the model, and the predicted risk scores of future potential erroneous operations are output to predict the possibility of potential erroneous operations in the future.

[0072] Specifically, the current operation behavior is intervened based on the current risk score, and the future potential misoperation behavior is intervened based on the predicted risk score, including the following steps:

[0073] Determine whether the current operation behavior is an erroneous operation based on the current risk score and a preset current risk score threshold, and if it is an erroneous operation, take a corresponding first intervention measure to intervene;

[0074] Based on the predicted risk score and the preset predicted risk score threshold, it is determined whether the potential future misoperation behavior has the risk of misoperation. If there is a risk of misoperation, the corresponding second intervention measure is taken for pre-intervention.

[0075] During the specific implementation process, the current risk score and predicted risk score are obtained through the processing of the advanced error-proofing monitoring model. According to business needs and historical data analysis, specific current risk score thresholds and predicted risk score thresholds are set. If the current risk score is higher than the current risk score threshold, the current operation behavior is considered to be an erroneous operation, and corresponding intervention measures are taken to intervene. Through the processing of the advanced error-proofing monitoring model, a predicted risk score representing potential erroneous operation behavior in the future is output, and the predicted risk score is compared with the preset predicted risk score threshold to determine whether potential erroneous operation behavior will occur in the future, and perform corresponding pre-intervention.

[0076] Specifically, the first intervention measure and the second intervention measure both include prohibiting associated operations, suspending current operations, issuing warning prompts, delaying instruction execution, requiring secondary confirmation, and suggesting alternative operations.

[0077] In the specific implementation process, including but not limited to these measures, these measures aim to reduce the occurrence of misoperation through various means and adopt corresponding intervention strategies under different risk levels.

[0078] Step S3: The output result after the primary intervention and the output result after the advanced intervention are combined and fed back to the user through the user interface of the user operation target.

[0079] In the specific implementation process, after the primary error-proofing monitoring model identifies the operation behavior that is inconsistent with the current operating state and executes the corresponding intervention measures, the operation target will generate a report containing the primary intervention results; at the same time, the advanced error-proofing monitoring model generates the risk score of the current operation and the predicted risk score of the potential future misoperation through in-depth analysis of the current operation behavior sequence and the historical operation behavior sequence, and takes corresponding advanced intervention measures based on these scores, and also generates an advanced intervention result report. The operation target merges these two reports to form a comprehensive intervention result report, which contains all the identified misoperation behaviors, the intervention measures that have been executed, and the early warning information of the potential future misoperation. Subsequently, this comprehensive report is presented to the user in an easy-to-understand form through the user interface of the user operation target.

[0080] The working principle of an AI fool-proofing method provided by the present invention is as follows:

[0081] The present invention uses two different levels of foolproofing models, namely, primary foolproofing monitoring model and advanced foolproofing monitoring model, to monitor the operation target in real time, and comprehensively prevent the influence of misoperation on the operation target. First, the real-time operation data, current operation behavior sequence and historical operation behavior sequence of the user's operation target are obtained, and these data respectively represent the current state of the system, the operation record of the user in the recent period of time and the operation history of the user in the past for a long time. Then, based on these data, the present invention adopts two different levels of intervention measures, primary intervention and advanced intervention. Primary intervention uses the pre-trained primary foolproofing monitoring model to analyze the real-time operation data, quickly identify the operation behavior that is contrary to the current system operation state, and execute the corresponding intervention measures in combination with the pre-defined intervention measure mapping table, so as to achieve the purpose of rapid identification and rapid intervention; advanced intervention is more in-depth, and after pre-processing and feature extraction of the current operation behavior sequence, the key features are input into the pre-trained advanced foolproofing monitoring model together with the current operation behavior sequence and the historical operation behavior sequence, which can generate the risk score of the current operation behavior and the predicted risk score of the potential misoperation behavior in the future, so as to allow the system to judge whether to take intervention measures immediately according to the current risk score, and to prevent future misoperation in advance according to the predicted risk score. During the feature extraction process, the system removes noise, cleans data, and extracts key features such as operation frequency, operation interval, and operation sequence to ensure the accuracy of the analysis results. The calculation of risk scores involves multi-level classifiers and long short-term memory network models, in which the first-level classifier is responsible for preliminary classification, the second-level classifier further refines the risk assessment, and the long short-term memory network is used for sequence analysis. Finally, a comprehensive risk score is obtained through a weighted fusion method. Based on these scores, the operation target can decide to take intervention measures such as prohibiting associated operations, suspending current operations, issuing warning prompts, delaying instruction execution, requiring secondary confirmation, or providing alternative operation suggestions. Finally, the results of the entire process will be fed back to the user through the user interface so that the user can understand the system's intervention and make corresponding adjustments. This multi-level and multi-dimensional fool-proofing mechanism aims to effectively reduce the occurrence of misoperations and improve the safety of the system through intelligent analysis and timely intervention.

[0082] Embodiment 2

[0083] The present invention also provides an AI fool-proofing system applied to an AI fool-proofing method. In a specific implementation process, Figure 2 As shown, it shows a schematic diagram of the module structure of an AI foolproofing system provided by the present invention, including:

[0084] The data collection module 100 is used to obtain the real-time operation data, current operation behavior sequence and historical operation behavior sequence of the user's operation target;

[0085] The foolproofing intervention module 200 is used to perform primary intervention on erroneous operation behaviors based on real-time operation data and using a pre-trained primary foolproofing monitoring model during the foolproofing process, and to perform advanced intervention on erroneous operation behaviors based on a current operation behavior sequence and a historical operation behavior sequence using a pre-trained advanced foolproofing monitoring model, wherein the erroneous operation behaviors are operation behaviors that are contrary to the current operation state of the user's operation target;

[0086] The result feedback module 300 is used to combine the output results after the primary intervention and the output results after the advanced intervention, and feed back to the user through the user interface of the user operation target.

[0087] The working principle of an AI fool-proofing system provided by the present invention is as follows:

[0088] The foolproofing system provided by the present invention identifies and prevents misoperation by real-time monitoring and analyzing the user's behavior on the operation target. First, the data collection module 100 obtains the real-time operation data of the user's operation target, the current operation behavior sequence and the historical operation behavior sequence. These data represent the current state of the system, the user's most recent operation record and the operation history in the past. Then, the foolproofing intervention module 200 adopts two levels of measures, primary intervention and advanced intervention, based on these data. The primary intervention uses a pre-trained primary foolproofing monitoring model to quickly identify operation behaviors that are inconsistent with the current operation state, and performs corresponding interventions in combination with a pre-defined intervention measure mapping table, while the advanced intervention is performed by After the operation behavior sequence is preprocessed and feature extracted, the key features are input into the pre-trained advanced error-proofing monitoring model together with the current and historical operation behavior sequences to generate a risk score for the current operation and a predicted risk score for potential future erroneous operations, so as to decide whether to take immediate or preventive intervention measures in advance based on the score, including prohibiting related operations, suspending the current operation, issuing warning prompts, delaying instruction execution, requiring secondary confirmation, or providing suggestions for alternative operations, etc. Finally, the result feedback module 300 feeds back the results of the entire process to the user through the user interface, forming a closed-loop intelligent error-proofing mechanism, which aims to effectively reduce erroneous operations and improve the security of the system through multi-level and multi-dimensional intelligent analysis and timely intervention.

[0089] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems) and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0090] A person skilled in the art may understand that all or part of the steps in the various methods of the above embodiments may be completed by instructing related hardware through a program, and the program may be stored in a computer-readable storage medium, the storage medium including a read-only memory (ROM), a random access memory (RAM), a programmable read-only memory (PROM), an erasable programmable read-only memory (EPROM), a one-time programmable read-only memory (OTPROM), an electronically-erasable programmable read-only memory (EEPROM), a compact disc (CD-ROM) or other optical disc storage, magnetic disk storage, magnetic tape storage, or any other computer-readable medium that can be used to carry or store data.

[0091] It should also be noted that the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, commodity or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, commodity or device. In the absence of more restrictions, the elements defined by the sentence "comprises a ..." do not exclude the existence of other identical elements in the process, method, commodity or device including the elements.

Claims

1. An AI fool-proofing method, characterized in that: The fool-proof method comprises the following steps: S1: Obtain the real-time operation data of the user's operation target, the current operation behavior sequence and the historical operation behavior sequence; S2: In the process of error prevention, based on real-time operation data, a pre-trained primary error prevention monitoring model is used to perform primary intervention on the erroneous operation behavior, and based on the current operation behavior sequence and the historical operation behavior sequence, a pre-trained advanced error prevention monitoring model is used to perform advanced intervention on the erroneous operation behavior, wherein the erroneous operation behavior is an operation behavior that is contrary to the current operation state of the user's operation target; S3: Combine the output results after primary intervention and the output results after advanced intervention, and feed them back to the user through the user interface of the user operation target; The advanced intervention includes the following steps: Preprocessing the acquired current operation behavior sequence, and extracting key features from the preprocessed current operation behavior sequence; The extracted key features, current operation behavior sequence, and historical operation behavior sequence are input into the pre-trained advanced error-proofing monitoring model to obtain the current risk score of the current operation behavior and the predicted risk score of the potential future erroneous operation behavior; Intervene in current operating behaviors based on current risk scores, and intervene in potential future misoperation behaviors based on predicted risk scores.

2. The AI ​​foolproofing method according to claim 1, characterized in that: The current operation behavior sequence is data of all operation behaviors executed within a preset time window including the time when the current operation behavior occurs, and the historical operation behavior sequence is data of all operation behaviors executed within a time window in the past.

3. The AI ​​fool-proofing method according to claim 2, characterized in that: The primary intervention includes the following steps: Obtain real-time operation data of user operation targets; Based on the acquired real-time operation data, the primary error-proofing monitoring model is used to identify the set of operation behaviors that are contrary to the current user operation target operation status; Based on the set of conflicting operational behaviors, the corresponding intervention measures are determined and executed in combination with the predefined intervention measure mapping table.

4. The AI ​​fool-proofing method according to claim 3, characterized in that: The step of preprocessing the acquired current operation behavior sequence and extracting key features therefrom includes the following steps: Remove repeated events, invalid data and outliers in the current operation behavior sequence through noise filtering; Perform data cleaning on the current operation behavior sequence after noise filtering; Extract key features of the current operation behavior sequence after data cleaning, where the key features include operation frequency, operation interval, operation sequence, user habits, and action mode; The extracted key features are normalized to obtain normalized key features.

5. The AI ​​fool-proofing method according to claim 4, characterized in that: The acquisition of the current risk score includes the following steps: Input the extracted key features into the first-level classifier of the advanced foolproof monitoring model, and preliminarily classify the key features of the current operation behavior and mark the classification results. If it is marked as normal, the current risk score of the current operation behavior is directly set to the preset normal value. If it is marked as abnormal, proceed to the next step; The classification results marked as abnormal by the first-level classifier are input into the second-level classifier of the advanced fool-proofing monitoring model for analysis to obtain the risk score of the current operation behavior; Input the current operation behavior sequence into the pre-trained long short-term memory network model to obtain the sequence analysis score of the current operation behavior; The risk score of the current operation behavior and the sequence analysis score are weighted and fused using the weighted fusion method to obtain the current risk score of the current operation behavior.

6. The AI ​​fool-proofing method according to claim 5, characterized in that: The predicted risk score is obtained by inputting the historical operation sequence into a pre-trained long short-term memory network model to obtain the predicted risk score of future potential misoperation behaviors.

7. The AI ​​fool-proofing method according to claim 6, characterized in that: The step of intervening in the current operation behavior based on the current risk score and intervening in the future potential misoperation behavior based on the predicted risk score includes the following steps: Determine whether the current operation behavior is an erroneous operation based on the current risk score and a preset current risk score threshold, and if it is an erroneous operation, take a corresponding first intervention measure to intervene; Based on the predicted risk score and the preset predicted risk score threshold, it is determined whether the potential future misoperation behavior has the risk of misoperation. If there is a risk of misoperation, the corresponding second intervention measure is taken for pre-intervention.

8. The AI ​​fool-proofing method according to claim 7, characterized in that: The first intervention measure and the second intervention measure both include prohibiting related operations, suspending current operations, issuing warning prompts, delaying instruction execution, requiring secondary confirmation, and suggesting alternative operations.

9. An AI fool-proofing system, characterized in that: Applied to an AI fool-proofing method as claimed in any one of claims 1 to 8, the fool-proofing system comprises: The data collection module is used to obtain the real-time operation data of the user's operation target, the current operation behavior sequence and the historical operation behavior sequence; A foolproofing intervention module is used to perform primary intervention on erroneous operation behaviors based on real-time operation data and using a pre-trained primary foolproofing monitoring model during the foolproofing process, and to perform advanced intervention on erroneous operation behaviors based on a current operation behavior sequence and a historical operation behavior sequence using a pre-trained advanced foolproofing monitoring model, wherein the erroneous operation behaviors are operation behaviors that are contrary to the current operation state of the user's operation target; The result feedback module is used to merge the output results after the primary intervention and the output results after the advanced intervention, and feed back to the user through the user interface of the user operation target.

Images