Data identification early warning system based on intelligent AI

Abstract

The invention relates to a data identification early warning system based on intelligent AI in the field of data processing, which comprises a processing terminal, the processing terminal is connected with an early warning layer and an abnormal database, the processing terminal and the early warning layer are respectively connected with a working layer, the working layer is connected with a data acquisition layer, the data acquisition layer is connected with a data processing layer, and the data processing layer is connected with an abnormal database. Data processing work is shared in the data processing layer and the processing terminal, the data processing layer processes redundant data preprocessing and clipping work, the working terminal only processes data matching and early warning index calculation, complex work is separated from the processing terminal, the work intensity of the processing terminal is reduced, the data processing efficiency is improved, and the data processing efficiency is improved. Meanwhile, composite data unit matching and single data unit matching are sequentially carried out on the collected to-be-detected data, the accumulated value of the early warning index is calculated, early warning feedback and common feedback are made according to the early warning index, the data processing cost is reduced, and the data processing timeliness is improved.

Description

Technical Field

[0001] The present invention relates to a data recognition and early warning system, and in particular to an intelligent AI-based data recognition and early warning system applied in the field of data processing. Background Art

[0002] With the rapid development of information technology, all kinds of data are growing explosively. How to effectively manage and utilize this data has become an important challenge facing enterprises and society, especially in key areas such as finance, industry, and healthcare. The timely discovery and processing of abnormal data is of great significance for avoiding losses and ensuring safety. Traditional manual review methods are not only inefficient, but also difficult to cope with large-scale and highly complex data environments. Therefore, it is particularly urgent to develop an intelligent and efficient data identification and early warning system.

[0003] The specification of invention patent CN202010390508.2 discloses an intelligent fault warning system for IT system operation and maintenance big data. The solution provided by the invention introduces AI technology into the operation and maintenance field, automatically realizes real-time and rapid identification and warning of computer system failures, greatly reduces labor costs, improves the efficiency of fault handling, and effectively ensures the continuous, healthy and rapid development of various businesses.

[0004] The invention patent CN201910715271.8 specification discloses an intelligent identification and early warning device for big data platform failures, including a platform data receiver, a platform indicator hierarchical storage device, a hierarchical indicator model microcontroller, a platform indicator judgment instrument, a frequent item set algorithm processor, and a fault alarm. It can solve the problem of untimely discovery of abnormal application operation indicators of the big data platform, make early fault warnings, and can mine patterns from historical failures of the big data platform for prediction of future failures, thereby better ensuring system stability.

[0005] The above application is mainly for IT system operation and maintenance. The data to be processed comes from the work of the IT system. The data is relatively simple and the required processing intensity is relatively low. However, with the development of information technology, various types of data on the Internet are growing explosively. If existing technology is used to comprehensively process massive amounts of data, the cost of data processing will be greatly increased. At the same time, it will easily affect the timeliness of data processing and induce safety accidents. Summary of the invention

[0006] In view of the above-mentioned prior art, the technical problem to be solved by the present invention is that when the prior art processes massive amounts of data comprehensively, the cost of data processing will be greatly increased, and it is also easy to affect the timeliness of data processing and induce safety accidents.

[0007] To solve the above problems, the present invention provides a data identification and early warning system based on intelligent AI, including a processing terminal, the processing terminal is signal-connected with an early warning layer and an abnormal database, the abnormal database is signal-connected with an Internet of Things database, the processing terminal and the early warning layer are signal-connected with a working layer respectively, the working layer is signal-connected with a data acquisition layer, the data acquisition layer is signal-connected with a data processing layer, and the data processing layer is signal-connected with the processing terminal; Its working principle includes: in the process of establishing the abnormal database, abnormal data related to early warning are collected from the Internet of Things database. After data trimming, the entered abnormal data forms multiple individual data units, and the individual data units generated from the same case are packaged into a composite data unit. At the same time, a single data unit is formed with a single hot word. The composite data unit and the single data unit together constitute the abnormal database. At the same time, the abnormal database will regularly update the composite data unit and the single data unit from the Internet of Things. The data generated by the working process of the working layer will be collected by the data collection layer. The collected data includes the production data generated by the work itself and the communication data generated by the interaction with the outside world. The collected data is summarized by the data collection layer and transmitted to the data processing layer. The data processing layer completes feature extraction through data preprocessing and data tailoring, and then transmits it to the processing terminal; The processing terminal first matches the data group to be checked with the composite data unit in the abnormal database. First, the composite data unit is searched in the abnormal database using the basic unit, and the composite data unit with multiple basic units is searched out and matched one by one. The individual data unit in the composite data unit is recorded as m, and the matched unit in the data group to be checked is recorded as n. Then the warning index N of this group of data matching is: N = (m + n) / (mn); After that, the data group to be tested and the single data unit in the abnormal database are matched. First, the units to be tested that are successfully matched in the final N value calculation are removed, and then the matching is performed again. Every time a data group to be tested is successfully matched, the N value is increased by 1. When N is greater than or equal to 5, the processing terminal sends a data warning to the warning layer, and the warning layer sends a warning feedback to the working layer. The technical personnel at the working layer will handle it according to the warning feedback.

[0008] In the above-mentioned data recognition and early warning system based on intelligent AI, the data processing work is distributed between the data processing layer and the processing terminal. The data processing layer handles the redundant data preprocessing and trimming operations, and the working terminal only handles data matching and the calculation of the early warning index, separating the complex work from the processing terminal, reducing the workload of the processing terminal, and improving data processing efficiency.

[0009] As a further improvement of the present application, the number of individual data units in a composite data unit does not exceed 5. When AI is unable to reduce the number of individual data units, manual intervention is performed to delete the individual data units to make the data that conforms to the data unit more concise and reduce the computational load of the processing unit.

[0010] As a further improvement of the present application, the early warning index N is rounded up in each calculation step.

[0011] As another improvement of the present application, when the data group to be checked has not completed matching with the composite data unit in the abnormal database, N is recorded as 0 in the matching data unit matching stage.

[0012] As another improved supplement to the present application, the abnormal database signal is connected to a professional terminology database. When the abnormal database receives the alarm data triggered by the early warning layer, it will count the alarm data and feed back the high-frequency data to the professional terminology database. When the high-frequency data conforms to the professional terminology in the current field and is used normally, the current high-frequency data will be desensitized. Data desensitization can reduce the frequency of false alarms.

[0013] As another improved supplement of the present application, in the composite data unit, after the individual data unit is desensitized, it is still used as a search object in the data retrieval process, but in the data matching process, the individual data unit is no longer matched, thereby ensuring the smooth progress of the retrieval.

[0014] As another improvement of the present application, a single data unit adds domain risk hot words according to its own domain, and the screening can be made more stringent by adding domain risk hot words.

[0015] As a further improvement of the present application, the working layer includes an early warning device, the early warning device includes a communication body and an external antenna, and the external antenna is fixedly connected to the communication body; The outer side of the external antenna is sleeved with a signal shielding cap and a support unit, the support unit is fixedly connected to the communication body, a compression spring is connected between the signal shielding cap and the support unit, and one end of the compression spring is fixedly connected to the signal shielding cap, and the other end of the signal shielding cap contacts the support unit, the support unit includes a fixing ring, the fixing ring is fixedly connected to the communication body, the end of the fixing ring away from the communication body is fixedly connected to a plurality of connecting piles, and the signal shielding cap is sleeved on the outside of the plurality of connecting piles, and the end of the connecting pile away from the fixing ring is fixedly connected to a limited block, the upper end of the communication body is fixedly connected with a pressure sensor and an electromagnetic ring whose position matches the external antenna in sequence from bottom to top, the sum of the heights of the electromagnetic ring and the pressure sensor is equal to the height of the fixing ring, the pressure sensor is wired to the communication body, and a time anchor point for early warning processing is formed by receiving early warning feedback, detecting pressure by the pressure sensor, and inputting relevant information for the final processing, and the early warning feedback and the processing staff are matched one by one, and at the same time, the double insurance of biometric information recognition and pressing the signal shielding cap is not easy to cause the early warning equipment to be misconnected during the process of carrying the task, and it is not easy to cause the early warning feedback to exist for a long time, and at the same time, an early warning processing file can be formed through the operation of the staff, which is convenient for subsequent tracing work.

[0016] As a further improvement supplement to the present application, a hysteresis unit is fixedly connected to one end of the connecting pile close to the signal shielding cap, and the hysteresis unit is filled with a buffer solution. A plurality of elastic balls are suspended in the buffer solution, and the volume of the elastic balls accounts for 80% of the volume of the inner cavity of the hysteresis unit. The end of the signal shielding cap close to the communication body contacts the hysteresis unit and compresses and deforms the hysteresis unit. On the one hand, the resistance to pressing the signal shielding cap can be increased to reduce the probability of false touch. On the other hand, the separation speed of the signal shielding cap under the action of the compression spring can be reduced, so that the connection structure between the signal shielding cap and the supporting unit is not easily collapsed due to impact.

[0017] In summary, in this application, the data processing work is shared between the data processing layer and the processing terminal, wherein the data processing layer handles the complicated data preprocessing and trimming operations, and the working terminal only handles the data matching and the calculation of the warning index, separating the complex work from the processing terminal, reducing the workload of the processing terminal, and improving the data processing efficiency. At the same time, the collected data to be inspected are matched with composite data units and single data units in turn, and the cumulative value of its warning index is calculated. Warning feedback and general feedback are made according to the warning index, thereby reducing the data processing cost and increasing the timeliness of data processing.

[0018] At the same time, the industry's hot words are used to desensitize the abnormal database or add field risk hot words, making the abnormal database more suitable for the current industry. On the one hand, data desensitization can reduce the frequency of false alarms, and on the other hand, adding field risk hot words can make the screening more stringent.

[0019] Finally, a time anchor point for early warning processing is formed by receiving early warning feedback, detecting pressure via the pressure sensor and inputting relevant information. The early warning feedback and the processing staff are matched one by one. At the same time, the double insurance of biometric information recognition and pressure signal shielding cap makes it less likely that the early warning device will misconnect to tasks while being carried around, and it is less likely to cause early warning feedback to exist for a long time. At the same time, an early warning processing file can be formed through staff operation to facilitate subsequent traceability work. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] Figure 1 This is a simplified structural diagram of the intelligent AI data recognition and early warning system of the first embodiment of this application; Figure 2 This is a simplified diagram of data flow within the data recognition and early warning system of the intelligent AI in the first embodiment of the present application; Figure 3 This is a schematic diagram of a method for identifying a composite data unit according to a first embodiment of the present application; Figure 4 This is a schematic diagram of a method for identifying a single data unit according to a first embodiment of the present application; Figure 5 This is a simplified structural diagram of the intelligent AI data recognition and early warning system of the second implementation mode of this application; Figure 6 This is a simplified diagram of data flow within the data recognition and early warning system of the intelligent AI in the second embodiment of the present application; Figure 7 This is a schematic diagram of composite data desensitization according to the second implementation mode of the present application; Figure 8 This is a schematic diagram of a method for identifying a composite data unit after data desensitization according to the second embodiment of the present application; Fig. 9 A structural schematic diagram of the working status of the early warning device of the working-level technician in the third implementation mode of the present application; Fig.10 A partial structural diagram of an external antenna of the working state of the early warning device of the working-level technician in the third embodiment of the present application; Fig.11 This is a schematic structural diagram of a support unit for the working status of an early warning device for working-level technicians in the third embodiment of the present application.

[0021] Reference numerals: 1 early warning device, 101 communication body, 102 external antenna, 2 signal shielding cap, 3 support unit, 301 fixing ring, 302 connecting pile, 303 limit block, 4 electromagnetic ring, 5 pressure sensor, 6 hysteresis unit, 7 elastic ball. DETAILED DESCRIPTION

[0022] Two implementation modes of the present application are described in detail below with reference to the accompanying drawings.

[0023] The first implementation method: Figure 1 A data identification and early warning system based on intelligent AI is shown, including a processing terminal, the processing terminal is signal-connected to an early warning layer and an abnormal database, the abnormal database is signal-connected to an Internet of Things database, the processing terminal and the early warning layer are signal-connected to a working layer respectively, the working layer is signal-connected to a data acquisition layer, the data acquisition layer is signal-connected to a data processing layer, and the data processing layer is signal-connected to the processing terminal; In the process of establishing the abnormal database, abnormal data related to the early warning are collected from the Internet of Things database. Taking fraud monitoring as an example, fraud cases existing on the Internet are collected and their authenticity is identified: isolated data is discarded, and cases that have been debunked are not entered. Related data can be screened through AI. The entered abnormal data is trimmed to form multiple individual data units, and the individual data units generated from the same case are packaged to form a composite data unit. In particular, the number of individual data units in a composite data unit does not exceed 5. When AI cannot reduce the number of individual data units, manual intervention can be performed to delete the individual data units. At the same time, taking fraud monitoring as an example, single data units are formed based on frequently appearing single hot words such as "account", "remittance" and "a person". The abnormal database is composed of composite data units and single data units.

[0024] See also Figure 2 The data generated by the working process of the working layer will be collected by the data collection layer. The collected data includes the production data generated by the work itself and the communication data generated by the interaction with the outside world. The collected data is summarized by the data collection layer and transmitted to the data processing layer. The data processing layer completes feature extraction after data preprocessing and data tailoring, and then transmits it to the processing terminal; See also Figure 3 , where data preprocessing includes data cleaning, conversion, normalization and other preprocessing operations on the data from individuals in different working layers obtained by the data collection layer, and then the preprocessed data is trimmed to the smallest unit to form a unit to be inspected, and the same group of units to be inspected are packaged to form a data group to be inspected, feature extraction is completed, and the data group to be inspected is sent to the processing terminal.

[0025] The processing terminal first matches the data group to be checked with the composite data unit in the abnormal database, first uses the basic unit to search for the composite data unit in the abnormal database, searches for the composite data unit with multiple basic units, and matches them one by one. Figure 3As an example, there are 5 individual data units in the composite data unit in the figure, which is recorded as m, and there are 2 units to be checked that have completed matching in the data group to be checked, which is n. Then the warning index N of this group of data matching is: N = (m + n) / (mn); When N is greater than or equal to 5, the processing terminal sends a data warning to the warning layer, and the warning layer sends a warning feedback to the working layer. The technical staff at the working layer handles it according to the warning feedback. Taking fraud monitoring as an example, the technical staff at the working layer contacts the working end where the warning data comes from to remind them of the risk of fraud. If necessary, they go offline to stop the dangerous behavior of the working end personnel.

[0026] Figure 3 The medium warning index N = (5 + 2) / (5 - 2) = 3.5, where N is an upward integer, that is, the value of N is 4. When the data group to be checked matches multiple composite data units in the abnormal database, the warning index matching the composite data unit takes the maximum value, and when the data group to be checked does not complete the match with the composite data unit in the abnormal database, N is recorded as 0.

[0027] After that, the data group to be tested is matched with the single data unit in the abnormal database. First, the successfully matched units to be tested in the final N value calculation are removed, and then the matching is performed again. For each successful match of a data group to be tested, the N value is increased by 1. Figure 3 and Figure 4 For example, in the current data group to be inspected, unit A and unit F to be inspected have been matched in the composite data unit retrieval, so they are removed in the subsequent single data unit matching, and finally unit D to be inspected is matched, and the N value increases from 4 to 5, triggering the early warning work, and when the N value is finally less than 5, the anomaly is eliminated, normal feedback is performed, and the working layer works normally.

[0028] In the present application, the data processing work is shared between the data processing layer and the processing terminal, wherein the data processing layer handles the complicated data preprocessing and trimming operations, and the working terminal only handles the data matching and the calculation of the warning index, separating the complex work from the processing terminal, reducing the workload of the processing terminal, and improving the data processing efficiency. At the same time, the collected data to be inspected are matched with composite data units and single data units in turn, and the cumulative value of its warning index is calculated. Warning feedback and general feedback are made according to the warning index, thereby reducing the data processing cost and increasing the timeliness of data processing.

[0029] The second implementation method: Figure 5-6, showing a data recognition and early warning system based on intelligent AI. The abnormal database signal is connected to the professional terminology database. Since different industries have their own professional terms, the professional terms existing in some industries may become abnormal data in the general abnormal database, which will cause frequent false alarms in the processing terminal and the early warning layer. When the abnormal database receives the alarm data triggered by the early warning layer, it will count the alarm data and feed back the high-frequency data to the professional terminology database. When the high-frequency data conforms to the professional terminology of the current field and is used normally, the current high-frequency data will be desensitized.

[0030] See also Figure 7-8 In the composite data unit, data A is desensitized. In the data retrieval process, data A is still the object of retrieval, but in the data matching process, data A is no longer matched. Figure 8 Taking the data as an example, there are 4 individual data units in the composite data unit in the figure, and there is 1 unit to be checked that has completed the match in the data group to be checked. Then the warning index N of this group of data matching is: N = (4 + 1) / (4 - 1) = 1.67 (save two decimal places), round up, the final value of N is set to 2. If data desensitization is not performed, there are 5 individual data units in this group of composite data units, and there are 2 units to be checked that have completed matching in the data group to be checked. Then the warning index N of this group of data matching is: N=(5+2) / (5-2)=2.33 (keep two decimal places), round up, the final value of N is set to 3. The existence of data desensitization can reasonably reduce the warning index N in the current field, reduce the number of false alarms, and improve the processing efficiency of the system.

[0031] Taking the financial field, network security, industrial monitoring and medical safety as examples, we can desensitize the hot words in this field, or add field risk hot words to a single data unit, such as: Financial risk control: In financial institutions such as banks and insurance companies, the system can automatically monitor customers' transaction behaviors and communication data, promptly detect risk events such as fraudulent transactions and credit defaults, and effectively prevent financial risks.

[0032] Network security: In the field of network security, the system can be used to monitor network traffic, log data, etc., quickly identify abnormal behaviors such as network attacks and malware intrusions, and ensure network security and stability.

[0033] Industrial monitoring: During the industrial production process, the system can monitor the operating status of equipment, product quality and other data in real time, promptly detect equipment failures, product defects and other problems, and improve production efficiency and product quality.

[0034] Healthcare: In the healthcare field, the system can be used to analyze patients’ medical records, test results and other data, identify potential disease risks, adverse drug reactions and other abnormalities, and provide strong support for clinical decision-making.

[0035] In this embodiment, the general-purpose database in the first embodiment is departed from, and the industry's hot words are used to desensitize the abnormal database or add domain risk hot words, so that the abnormal database is more suitable for the current industry. On the one hand, data desensitization can reduce the frequency of false alarms, and on the other hand, adding domain risk hot words can make the screening more stringent.

[0036] The third implementation method Figure 9-11 The warning device of the technicians at the working layer is shown, including the warning device 1, the warning device 1 includes a communication body 101 and an external antenna 102, and the external antenna 102 is fixedly connected to the communication body 101, wherein the communication body 101 of the present application has the same structure as the existing smart phone, and can realize the functions of communication, navigation positioning and data transmission through the built-in antenna, and the external antenna 102 is specially used to receive the warning feedback transmitted by the warning layer; A signal shielding cap 2 and a support unit 3 are sleeved on the outside of the external antenna 102. The support unit 3 is fixedly connected to the communication body 101. A compression spring is connected between the signal shielding cap 2 and the support unit 3, and one end of the compression spring is fixedly connected to the signal shielding cap 2, and the other end of the signal shielding cap 2 is in contact with the support unit 3. The support unit 3 includes a fixing ring 301, which is fixedly connected to the communication body 101. One end of the fixing ring 301 away from the communication body 101 is fixedly connected to a plurality of connecting piles 302, and the signal shielding cap 2 is sleeved on the outside of the plurality of connecting piles 302. One end of the connecting pile 302 away from the fixing ring 301 is fixedly connected to a limiting block 303. The upper end of the communication body 101 is fixedly connected with a pressure sensor 5 and an electromagnetic ring 4 whose positions match the external antenna 102 in sequence from bottom to top. The sum of the heights of the electromagnetic ring 4 and the pressure sensor 5 is equal to the height of the fixing ring 301. The pressure sensor 5 is wired to the communication body 101, and the outer wall of the pressure sensor 5 is covered with a signal shielding shell.

[0037] When the present application is in standby state, the electromagnetic ring 4 is in a power-off state, and the signal shielding cap 2 moves in a direction away from the communication body 101 under the action of the compression spring in a compressed state, and at this time, the external antenna 102 can achieve communication connection with the warning layer through the gap between the multiple connection piles 302; When the early warning layer sends an early warning feedback to the early warning device 1, the early warning device 1 will sound an alarm and vibrate. After the staff confirms the login through the fingerprint and face recognition of the early warning device 1, the electromagnetic ring 4 switches to the standby state and forms a corresponding work log to confirm that the staff has received the early warning feedback. If the staff does not claim the early warning feedback for a long time, the work log generated this time will be deleted. After confirming the warning information, when the staff needs to claim the warning processing, they press the signal shielding cap 2 to make the signal shielding cap 2 contact with the electromagnetic ring 4. At this time, the pressure sensor 5 detects the pressure generated by the signal shielding cap 2 on the electromagnetic ring 4, and transmits the corresponding data back to the communication body 101 to confirm the claim. The electromagnetic ring 4 switches to the working state and adsorbs the signal shielding cap 2 to form a signal closed area for the external antenna 102. The warning layer will not issue the warning processing task to the warning device 1 with the task again, and at the same time, it also reduces the impact of the electromagnetic ring 4 on the external antenna 102.

[0038] After the early warning processing task is completed, the staff needs to input the text, image and video information related to the processing task through the early warning device 1 and combine it into the work log, and then transmit it to the data acquisition layer, and the data acquisition layer will feedback to the processing terminal layer by layer to form an early warning processing file. At this time, the electromagnetic ring 4 resumes the standby state, and under the action of the compression spring in the compressed state, the signal shielding cap 2 is re-loaded, and the external antenna 102 is exposed again, and the communication connection with the early warning layer is restored.

[0039] In the early warning device 1 of the present application, a time anchor point for early warning processing is formed by receiving early warning feedback, detecting pressure via the pressure sensor 5 and finally processing the input of relevant information. At the same time, the double insurance of biometric information recognition and the pressure signal shielding cap 2 makes it less likely that the early warning device 1 will misconnect to a task while being carried around, and it is less likely that the early warning feedback will exist for a long time. An early warning processing file can be formed through staff operation to facilitate subsequent tracing work.

[0040] In particular, the present application is applicable to staff members who have multiple early warning feedback processing at the work end. When a staff member is in the process of processing an early warning feedback task, the early warning feedback task will no longer be sent to the staff member. By accepting early warning feedback tasks in batches, the efficiency of early warning task processing can be improved to avoid the accumulation of multiple tasks on the same staff member. At the same time, based on the processing data automatically generated by the early warning device 1 and the processing data input by the staff member, the staff member's response speed and early warning feedback processing efficiency can be obtained according to the work log, which is convenient for subsequent early warning system management work.

[0041] The end of the connecting pile 302 close to the signal shielding cap 2 is fixedly connected to the hysteresis unit 6, the hysteresis unit 6 is filled with a buffer solution, and a plurality of elastic balls 7 are suspended in the buffer solution, the volume of the elastic balls 7 accounts for 80% of the volume of the inner cavity of the hysteresis unit 6, and the end of the signal shielding cap 2 close to the communication body 101 is in contact with the hysteresis unit 6, and the hysteresis unit 6 is compressed and deformed, which can increase the resistance of the signal shielding cap 2 to be pressed and reduce the probability of false touch on the one hand, and reduce the separation speed of the signal shielding cap 2 under the action of the compression spring on the other hand, so that the connection structure between the signal shielding cap 2 and the supporting unit 3 is not easily collapsed due to impact.

[0042] In view of current practical needs, the above-mentioned implementation mode adopted in this application is not limited to the scope of protection. Various changes made within the knowledge scope of technical personnel in this field without departing from the concept of this application still fall within the scope of protection of the present invention.

Claims

1. A data recognition and early warning system based on intelligent AI, including a processing terminal, characterized in that: The processing terminal is signal-connected with the early warning layer and the abnormal database, the abnormal database is signal-connected with the Internet of Things database, the processing terminal and the early warning layer are signal-connected with the working layer respectively, the working layer is signal-connected with the data acquisition layer, the data acquisition layer is signal-connected with the data processing layer, and the data processing layer is signal-connected with the processing terminal; Its working principle includes: in the process of establishing the abnormal database, abnormal data related to early warning are collected from the Internet of Things database. After data trimming, the entered abnormal data forms multiple individual data units, and the individual data units generated from the same case are packaged into a composite data unit. At the same time, a single data unit is formed with a single hot word. The composite data unit and the single data unit together constitute the abnormal database. At the same time, the abnormal database will regularly update the composite data unit and the single data unit from the Internet of Things. The data generated by the working process of the working layer will be collected by the data collection layer, which will summarize the collected data and transmit it to the data processing layer. The data processing layer will complete feature extraction after data preprocessing and data trimming, and then transmit it to the processing terminal; The processing terminal first matches the data group to be checked with the composite data unit in the abnormal database. First, the composite data unit is searched in the abnormal database using the basic unit, and the composite data unit with multiple basic units is searched out and matched one by one. The individual data unit in the composite data unit is recorded as m, and the matched unit in the data group to be checked is recorded as n. Then the warning index N of this group of data matching is: N = (m + n) / (mn); After that, the data group to be tested and the single data unit in the abnormal database are matched. First, the units to be tested that are successfully matched in the final N value calculation are removed, and then the matching is performed again. Every time a data group to be tested is successfully matched, the N value is increased by 1. When N is greater than or equal to 5, the processing terminal sends a data warning to the warning layer, and the warning layer sends a warning feedback to the working layer. The technical personnel at the working layer will handle it according to the warning feedback.

2. According to claim 1, a data recognition and early warning system based on intelligent AI is characterized in that: The number of individual data units in the composite data unit does not exceed 5. When AI cannot reduce the number of individual data units, manual intervention is performed to delete the individual data units.

3. According to claim 1, a data recognition and early warning system based on intelligent AI is characterized in that: The warning index N is rounded up in each calculation step.

4. The data recognition and early warning system based on intelligent AI according to claim 1 is characterized in that: When the data group to be checked has not completed matching with the composite data unit in the abnormal database, N is recorded as 0 in the composite data unit matching stage.

5. The data recognition and early warning system based on intelligent AI according to claim 1 is characterized in that: The abnormal database signal is connected to the professional terminology database. When the abnormal database receives the alarm data triggered by the early warning layer, it will count the alarm data and feed back the high-frequency data to the professional terminology database. When the high-frequency data conforms to the professional terminology in the current field and is used normally, the current high-frequency data will be desensitized.

6. The data recognition and early warning system based on intelligent AI according to claim 5 is characterized in that: In a composite data unit, after an individual data unit is desensitized, it is still a search object during the data retrieval process, but during the data matching process, the individual data unit is no longer matched.

7. The data recognition and early warning system based on intelligent AI according to claim 5, characterized in that: The single data unit adds domain risk hot words according to its own domain.

8. The data recognition and early warning system based on intelligent AI according to claim 1 is characterized by: The working layer comprises an early warning device (1), wherein the early warning device (1) comprises a communication body (101) and an external antenna (102), and the external antenna (102) is fixedly connected to the communication body (101); The outer side of the external antenna (102) is sleeved with a signal shielding cap (2) and a support unit (3); the support unit (3) is fixedly connected to the communication body (101); a compression spring is connected between the signal shielding cap (2) and the support unit (3); one end of the compression spring is fixedly connected to the signal shielding cap (2); the other end of the signal shielding cap (2) is in contact with the support unit (3); the support unit (3) comprises a fixing ring (301); the fixing ring (301) is fixedly connected to the communication body (101); the fixing ring (301) is away from the communication body (101). One end of the communication body (101) is fixedly connected to a plurality of connection piles (302), and a signal shielding cap (2) is sleeved on the outside of the plurality of connection piles (302); one end of the connection pile (302) away from the fixing ring (301) is fixedly connected to a limit block (303); a pressure sensor (5) and an electromagnetic ring (4) whose positions match the external antenna (102) are fixedly connected to the upper end of the communication body (101) in sequence from bottom to top; the sum of the heights of the electromagnetic ring (4) and the pressure sensor (5) is equal to the height of the fixing ring (301); and the pressure sensor (5) is connected to the communication body (101) by wire.

9. The data recognition and early warning system based on intelligent AI according to claim 8, characterized in that: One end of the connection pile (302) close to the signal shielding cap (2) is fixedly connected to a hysteresis unit (6), the hysteresis unit (6) is filled with a buffer solution, a plurality of elastic balls (7) are suspended in the buffer solution, and the volume of the elastic balls (7) accounts for 80% of the volume of the inner cavity of the hysteresis unit (6).

Images