Intelligent early warning method and device for low-code development and medium

Abstract

The invention discloses an intelligent early warning method and device for low-code development and a medium, and relates to the technical field of software development. The method comprises the following steps: acquiring historical monitoring data through a multi-source data interface, and pre-training a time sequence prediction model loaded in an artificial intelligence platform based on the historical monitoring data; creating a data acquisition form in the low-code platform, and establishing an automatic data processing flow through a flow engine; when a form data submission event is detected, a model reasoning service calling instruction is triggered, so that current monitoring data and corresponding historical time sequence data are input into the time sequence prediction model, and a prediction value of a future time period is obtained; wherein the historical time sequence data is historical monitoring data within a preset duration before the current monitoring data is obtained; and comparing the predicted value with an actual monitoring value after a future time period, and if the deviation exceeds a preset threshold value, sending an alarm notification based on a preset alarm mechanism.

Description

Technical Field

[0001] This application relates to the field of software development technology, and particularly to an intelligent early warning method, device, and medium for low-code development. Background Art

[0002] With the acceleration of enterprise digital transformation, the demand for intelligent early warning systems is increasing day by day. The development of traditional early warning systems highly depends on professional programming skills and complex system integration, involving a large amount of code writing and debugging, resulting in a long development cycle and high maintenance costs. Especially in the data preprocessing, model training, and deployment links, the traditional development mode requires the collaboration of cross-domain technical teams, with low efficiency and difficulty in quickly responding to business changes. In addition, in the prior art, the dynamic adjustment of early warning rules and the adaptability to business scenarios are insufficient, further restricting the flexibility and practicality of the system.

[0003] In recent years, the emergence of low-code platforms has simplified the application development process, but its application in the field of intelligent early warning still has significant limitations. Existing low-code platforms lack the ability to efficiently integrate with AI models and are difficult to achieve end-to-end automation from data collection, model inference to alarm triggering. For example, most platforms only support simple rule configuration and cannot perform dynamic prediction analysis in combination with real-time data, resulting in insufficient early warning accuracy and timeliness. At the same time, existing platforms have limited support for model optimization technologies (such as quantization, pruning), and it is difficult to balance model inference efficiency and resource consumption, restricting their application in complex business scenarios.

[0004] Furthermore, there are defects in the coordination between the message notification mechanism and the process engine in existing early warning systems. In traditional solutions, the alarm triggering and notification logic usually need to be implemented manually through coding, lacking visual configuration capabilities, resulting in cumbersome alarm rule updates and easy errors.

[0005] Therefore, how to achieve the efficient integration of low-code platforms and AI models and build a full-process automated system from data collection, model inference to alarm triggering has become a technical problem urgently to be solved by those skilled in the art. Summary of the Invention

[0006] Embodiments of this application provide an intelligent early warning method, device, and medium for low-code development to solve the following technical problems: how to achieve the efficient integration of low-code platforms and AI models and build a full-process automated system from data collection, model inference to alarm triggering.

[0007] In a first aspect, an embodiment of the present application provides an intelligent early warning method for low-code development. The method includes: obtaining historical monitoring data through a multi-source data interface, and pre-training a time series prediction model loaded in an artificial intelligence platform based on the historical monitoring data; creating a data collection form in a low-code platform, and establishing an automated data processing process through a process engine; when a form data submission event is detected, triggering a model inference service call instruction to input the current monitoring data and the corresponding historical time series data into the time series prediction model to obtain predicted values for a future period; wherein the historical time series data is the historical monitoring data within a preset duration before obtaining the current monitoring data; comparing the predicted values with the actual monitoring values after the future period, and if the deviation exceeds a preset threshold, sending an alarm notification based on a preset alarm mechanism.

[0008] In an embodiment of the present application, before pre-training the time series prediction model loaded in the artificial intelligence platform based on the historical monitoring data, the method further includes: identifying missing values and outliers in the historical monitoring data; filling the missing values using linear interpolation and removing the outliers; dividing the processed historical monitoring data into a training set and a validation set.

[0009] In an embodiment of the present application, pre-training the time series prediction model loaded in the artificial intelligence platform based on the historical monitoring data specifically includes: constructing a model input sequence using a sliding time window method based on the training set to convert the historical monitoring data within each time window into a multi-dimensional feature vector; synchronously implementing a model compression operation during model training; wherein the model compression operation includes performing fixed-point quantization on neural network weights, pruning redundant neuron connections, and transferring model parameters through knowledge distillation technology; after the model training is completed, correcting the initially trained model based on the validation set to obtain a converged time series prediction model.

[0010] In an embodiment of the present application, correcting the initially trained model based on the validation set specifically includes: calculating the accuracy, recall, and F1-score metrics of the time series prediction model based on the validation set data; when any of the metrics is lower than the corresponding preset metric threshold, adjusting the model hyperparameters and re-executing the pre-training process until a converged time series prediction model is obtained.

[0011] In an embodiment of the present application, triggering a model inference service call instruction specifically includes: parsing the measurement point identifier and data acquisition timestamp in the form data; associating a preset geographical location database according to the measurement point identifier to obtain the storage path of the historical monitoring data corresponding to this measurement point; based on the data acquisition timestamp and a preset duration, extracting the historical time series data from the storage path, arranging it in the order of time windows, and inputting it into the time series prediction model.

[0012] In an embodiment of the present application, sending an alarm notification based on a preset alarm mechanism specifically includes: generating alarm message content including the current monitoring timestamp, monitoring point identifier, predicted value deviation amount, and monitoring change trend chart; by parallelly calling the SMTP protocol interface, SMS gateway API, and WeChat enterprise number service interface, converting the alarm message content into email body, SMS text, and enterprise application message formats respectively; according to the preset interval range where the deviation amount is located, dynamically selecting the message recipient list and performing a multi-level alarm push operation.

[0013] In an embodiment of the present application, after sending an alarm notification based on a preset alarm mechanism, the method further includes: storing the alarm message content, sending timestamp, recipient identifier, and sending status code into the message log database; in response to a query request submitted by the user through the low-code platform, retrieving historical alarm records from the message log database, and generating a visual analysis chart including time distribution and alarm type statistics based on the records.

[0014] In an embodiment of the present application, after sending an alarm notification based on a preset alarm mechanism, the method further includes: counting the number of times the alarm notification is triggered within a preset time period; when the number of times exceeds a preset threshold, adding the actual monitoring value after the future time period to the historical monitoring dataset; based on the updated historical monitoring dataset, reconstructing the model input sequence using the sliding time window method, and performing a model compression operation on the time series prediction model to generate an optimized time series prediction model; deploying the optimized time series prediction model to the artificial intelligence platform through the model hot reload interface to replace the original time series prediction model.

[0015] In a second aspect, an embodiment of the present application further provides an intelligent early warning device for low-code development, the device includes: at least one processor; and a memory communicatively connected to at least one processor; wherein, the memory stores instructions executable by at least one processor, and the instructions are executed by at least one processor so that at least one processor can execute an intelligent early warning method for low-code development as described in any one of the above.

[0016] In a third aspect, an embodiment of the present application further provides a non-volatile computer storage medium for intelligent early warning in low-code development, storing computer-executable instructions, which, when executed, implement an intelligent early warning method for low-code development as described in any one of the above.

[0017] The intelligent early warning method, device, and medium for low-code development provided by the embodiments of the present application have the following beneficial effects: By deeply integrating the artificial intelligence platform, low-code platform, and process engine, the full-process automation and intelligence of the early warning system development are realized. Non-professional users can quickly configure business forms, early warning rules, and multi-level notification strategies through a visual interface, and automatically call the AI model to complete data reasoning and alarm decision-making, significantly reducing the technical threshold and development cost; Based on the automated scheduling ability of the process engine, it can seamlessly connect the data collection, model reasoning, alarm triggering, and message pushing links, significantly improving the timeliness of early warning response. At the same time, combined with optimization technologies such as model quantization and pruning, resource consumption is reduced while ensuring prediction accuracy, which is suitable for the efficient deployment and dynamic expansion of complex business scenarios, providing reliable support for enterprises to build flexible and accurate intelligent early warning systems. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] The drawings described herein are used to provide a further understanding of the present application, and constitute a part of the present application. The schematic embodiments of the present application and their descriptions are used to explain the present application, and do not constitute an improper limitation to the present application. In the drawings: Figure 1 It is a flowchart of an intelligent early warning method for low-code development provided by an embodiment of the present application; Figure 2 It is a schematic internal structure diagram of an intelligent early warning device for low-code development provided by an embodiment of the present application. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0019] To make the purpose, technical solutions, and advantages of the present application clearer, the technical solutions of the present application will be clearly and completely described below in conjunction with the specific embodiments of the present application and the corresponding drawings. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present application.

[0020] The embodiments of the present application provide an intelligent early warning method, device, and medium for low-code development to solve the following technical problems: How to achieve the efficient integration of the low-code platform and the AI model and build a full-process automated system from data collection, model reasoning to alarm triggering.

[0021] The technical solutions proposed in the embodiments of the present application will be described in detail below with reference to the accompanying drawings.

[0022] Figure 1 It is a flowchart of an intelligent early warning method for low-code development provided by an embodiment of the present application. As Figure 1 shown, an intelligent early warning method for low-code development provided by an embodiment of the present application specifically includes the following steps: Step 101: Obtain historical monitoring data through a multi-source data interface, and pre-train the time series prediction model loaded in the artificial intelligence platform based on the historical monitoring data.

[0023] In an embodiment of the present application, before pre-training the time series prediction model loaded in the artificial intelligence platform based on the historical monitoring data, the method further includes: identifying missing values and outliers in the historical monitoring data; filling the missing values using linear interpolation method, and performing elimination processing on the outliers; dividing the processed historical monitoring data into a training set and a validation set.

[0024] In an embodiment of the present application, pre-training the time series prediction model loaded in the artificial intelligence platform based on the historical monitoring data specifically includes: constructing a model input sequence using the sliding time window method based on the training set to convert the historical monitoring data within each time window into a multi-dimensional feature vector; synchronously implementing model compression operations during model training; where the model compression operations include performing fixed-point quantization processing on the neural network weights, pruning redundant neuron connections, and migrating model parameters through knowledge distillation technology; after the model training is completed, correcting the initially trained model based on the validation set to obtain a converged time series prediction model.

[0025] In an embodiment of the present application, correcting the initially trained model based on the validation set specifically includes: calculating the accuracy rate, recall rate, and F1 score metrics of the time series prediction model based on the validation set data; when any of the metrics is lower than the corresponding preset metric threshold, adjusting the model hyperparameters and re-executing the pre-training process until a converged time series prediction model is obtained.

[0026] It can be understood that the "multi-source data interface" refers to a heterogeneous data access channel that supports sensors, databases, and manual input. Exemplarily, in a water level monitoring scenario, sensor data is transmitted to the system in real time through the MQTT protocol, historical water level records in the database are imported in batches through the JDBC interface, and manual inspection data is manually filled in through the Web form provided by the low-code platform. The multi-source data interface standardizes and encapsulates the data in a unified JSON or CSV format to ensure that data from different sources can be compatible with subsequent processing modules.

[0027] It should be noted that the data preprocessing operation is synchronously executed at this stage. Specifically, the system first identifies missing values and outliers in the historical monitoring data. Exemplarily, when the water level sensor causes data loss in a certain period due to power failure, the system uses linear interpolation to fill the gap: according to the data of adjacent timestamps before and after the missing point, smooth transition filling values are generated according to the time interval ratio. For outliers (such as sudden increase or decrease data points caused by sensor failures), the system detects them through the calculation of the standard deviation of the sliding window. If the deviation of a data point from the mean value within the window exceeds 3 times the standard deviation, it is determined as an outlier and excluded. After cleaning, the data is divided into a training set and a validation set, where the training set is used for model parameter learning, and the validation set is used to evaluate the generalization ability of the model to avoid overfitting problems.

[0028] In this embodiment, the time series prediction model selects the LSTM neural network, and its input sequence is constructed by the sliding time window method. Exemplarily, taking water level prediction as an example, each time window contains the water level data and associated environmental parameters (such as daily rainfall, temperature) for 7 consecutive days, and the output is the predicted water level value for the 8th day. During the model training process, the model compression technology in claim 3 is synchronously implemented: 1. Fixed-point quantization: Compress the 32-bit floating-point weight parameters in the LSTM network into 8-bit integer representation, reducing the storage space of the model and the consumption of computing resources; 2. Neuron pruning: Based on the evaluation of weight importance, remove redundant neuron connections with absolute values lower than the threshold to simplify the network structure; 3. Knowledge distillation: Transfer the prediction knowledge of the complex teacher model to the lightweight student model, and improve the accuracy of the small model through soft label training.

[0029] It should be noted that after the model training is completed, the accuracy rate, recall rate, and F1 score indicators are calculated based on the validation set data. Exemplarily, if the F1 score of the validation set is lower than the preset threshold (such as 0.85), it is determined that the model has not converged, and the number of hidden layer neurons or the learning rate hyperparameters of the LSTM need to be adjusted and retrained until the indicators reach the standard. This process ensures that the model has reliable prediction ability in the actual business scenario.

[0030] Through the above implementation methods, the entire process of data acquisition, cleaning, model training optimization, and verification is fully covered, laying a foundation for the subsequent model inference and alarm triggering steps. The preprocessing and model compression technologies significantly improve the data quality and inference efficiency, enabling the time series prediction model to meet the requirements of lightweight and real-time of the low-code platform. Exemplarily, in a certain water conservancy management scenario, the inference time of the pruned and quantized LSTM model is reduced from the original 320ms to 95ms, meeting the response requirements of the low-code platform for real-time early warning.

[0031] Step 102: Create a data collection form in the low-code platform and establish an automated data processing process through the process engine.

[0032] It can be understood that the "data collection form" is a data entry interface custom-built by the user through the visual designer of the low-code platform. Exemplarily, in the water level monitoring scenario, the form includes the following core fields: 1. Measurement point identifier: Used to uniquely identify the monitoring point (such as "Reservoir A-002"), implemented through a dropdown menu or a map point selection component; 2. Data collection timestamp: Automatically records the submission time and supports manual correction to be compatible with the offline filling scenario; 3. Current water level value: Supports manual input or automatic synchronization of real-time data through the sensor interface; 4. Associated environmental parameters: Such as daily rainfall, temperature, etc., dynamically loaded through an external database interface or filled in manually.

[0033] Specifically, the measurement point identifier is associated with a preset geographical location database. For example, when the user selects "Reservoir A-002", the system automatically associates metadata such as the longitude and latitude coordinates, historical data storage path, and safety water level threshold of this measurement point, providing context information for subsequent model inference.

[0034] In this embodiment, the configuration of the process engine is completed through a graphical interface. The user drags and drops nodes (such as "data submission", "model call", "alarm judgment") and sets logical conditions to build an end-to-end automated process. Exemplarily, the process design includes the following steps: Data submission event trigger: When the user submits the form, the process engine automatically captures the event and extracts the measurement point identifier, timestamp, and monitoring data in the form; Model inference service call: According to the model type associated with the measurement point identifier (such as the water level prediction model), the process engine sends an inference request to the artificial intelligence platform; Historical data dynamic loading: Based on the logic described in claim 5, the process engine extracts historical time series data from the specified storage path according to the timestamp and the preset duration (such as "the past 7 days"), sorts it by time window, and inputs it into the model; Rule judgment and branch processing: Set conditional branches (such as "deviation ≤ 5%" for log recording, "deviation > 5%" for triggering an alarm) to implement different business processes.

[0035] Specifically, the asynchronous task scheduling mechanism of the process engine ensures stability in high-concurrency scenarios. For example, when multiple users submit data simultaneously, the process engine sorts tasks by priority through a message queue (such as RabbitMQ) to avoid blocking caused by resource competition. In addition, the process supports the configuration of a "retry policy". If the model call fails due to network fluctuations, the engine automatically reinitiates the request after a set delay until it succeeds or reaches the maximum number of retries.

[0036] Through the above implementation methods, users can complete the design of the data collection interface and the orchestration of business processes without writing code. At the same time, through the dynamic association of measurement point identifiers, accurate data context is provided for subsequent model inference; the deep collaboration between the low-code platform and the process engine is also achieved. Exemplarily, a certain environmental protection department completed the construction of a water quality monitoring and early warning system within 1 hour by dragging and dropping form fields (such as pH value, dissolved oxygen) and configuring process rules, while traditional coding development would take several weeks. This process fully demonstrates the core advantages of the low-code platform in improving business agility.

[0037] Step 103: When a form data submission event is detected, trigger a model inference service call instruction to input the current monitoring data and the corresponding historical time series data into the time series prediction model to obtain predicted values for future time periods.

[0038] In this embodiment, the historical time series data is the historical monitoring data within a preset time period before obtaining the current monitoring data.

[0039] In an embodiment of the present application, triggering the model inference service call instruction specifically includes: parsing the measurement point identifier and data collection timestamp in the form data; associating a preset geographical location database according to the measurement point identifier to obtain the storage path of the historical monitoring data corresponding to this measurement point; based on the data collection timestamp and the preset time period, extract the historical time series data from the storage path, arrange it in the order of time windows, and then input it into the time series prediction model.

[0040] It can be understood that the "form data submission event" refers to the event signal captured by the process engine after the user completes data filling in the low-code platform and triggers the submission operation. Exemplarily, when the reservoir administrator enters the daily water level value through the form and clicks the submit button, the process engine immediately starts the preset automated processing link.

[0041] It should be noted that when a form data submission event is detected, the process engine first parses the measurement point identifier (such as "Reservoir A - 002") and the data collection timestamp (such as "2024 - 07 - 01 14:00") in the form data. The measurement point identifier is used to associate with a preset geographical location database to obtain the metadata configuration of the monitoring point, including the historical data storage path, model type (such as the water level prediction LSTM model), and associated parameters (such as the rainfall threshold). For example, the historical data corresponding to the identifier "Reservoir A - 002" is stored in the / reservoir / A / 002 path of the time series database, and the data in this path is arranged in ascending order of the timestamp.

[0042] In this embodiment, the extraction of historical time series data is based on the time window rule. Specifically, the system filters out continuous historical monitoring records from the storage path according to the data collection timestamp and a preset duration (such as "the past 7 days"). Exemplarily, if the collection time of the currently submitted data is "2024 - 07 - 01", then the water level data and associated environmental parameters (such as daily rainfall) during the period from "2024 - 06 - 24" to "2024 - 06 - 30" are extracted and concatenated into a multi - dimensional feature vector in chronological order to form the model input sequence. This process ensures that the model can make predictions based on a complete time context.

[0043] Furthermore, the model inference service call is implemented through a standardized API interface. It can be understood that the process engine encapsulates the current monitoring data (such as the daily water level value) and the historical time series data into a structured request body and sends it to the inference service endpoint of the artificial intelligence platform through the HTTP protocol. Exemplarily, the request body contains the following fields: sensor_id: the measurement point identifier, used to match the pre - deployed model version; current_data: the currently entered water level value and environmental parameters; history_sequence: the historical time series data arranged in a time window.

[0044] It should be noted that after receiving the request, the model inference service automatically loads the time series prediction model (such as the LSTM model in ONNX format) associated with the measurement point identifier, performs forward calculation, and returns the prediction result. For example, for the water level prediction scenario, the predicted value output by the model is the estimated water level value 24 hours later, and its format is a JSON object: {"predicted_level": 3.2, "confidence": 0.92}, where confidence represents the confidence level of the model in the prediction result for reference in subsequent alarm logic.

[0045] Specifically, the model inference process also includes an exception handling mechanism. Exemplarily, if there is missing historical time-series data (e.g., due to storage failures, data for some time periods is unavailable), the model service will automatically enable an interpolation algorithm to generate substitute data, or return an error code to require the process engine to re-trigger the data completion process, thereby ensuring the robustness of the prediction process.

[0046] Through the above implementation, seamless connection from data submission to model inference is achieved. For example, in the scenario of urban waterlogging early warning in a certain city, when a monitoring point submits instantaneous rainfall data, the process engine automatically calls the model to predict the water level change in the next hour, and combines with Geographic Information System (GIS) data to evaluate the waterlogging risk, providing a decision-making basis for emergency response. This process fully reflects the technical value of data association and model invocation, ensuring the timeliness and accuracy of the prediction results.

[0047] Step 104: Compare the predicted value with the actual monitored value after a future time period. If the deviation exceeds a preset threshold, send an alarm notification based on a preset alarm mechanism.

[0048] In an embodiment of the present application, sending an alarm notification based on a preset alarm mechanism specifically includes: generating alarm message content including the current monitoring timestamp, monitoring point identifier, predicted value deviation amount, and monitoring change trend chart; by parallelly invoking the SMTP protocol interface, SMS gateway API, and WeChat enterprise account service interface, converting the alarm message content into email body, SMS text, and enterprise application message formats respectively; dynamically selecting a list of message recipients according to the preset interval range where the deviation amount is located and performing a multi-level alarm push operation.

[0049] In an embodiment of the present application, after sending an alarm notification based on a preset alarm mechanism, the method further includes: storing the alarm message content, sending timestamp, recipient identifier, and sending status code to a message log database; in response to a query request submitted by a user through a low-code platform, retrieving historical alarm records from the message log database, and generating a visual analysis chart including time distribution and alarm type statistics based on the records.

[0050] It can be understood that the acquisition method of the "actual monitored value" is dynamically adapted according to the business scenario. Exemplarily, in an automated monitoring scenario, the system polls the latest data at regular intervals through a sensor interface; while in a semi-manual scenario (such as water level monitoring in remote areas), the actual value can be supplemented by on-site personnel through a secondary confirmation form designed by the low-code platform. The comparison time window between the actual value and the predicted value is defined by business rules. For example, if a water level prediction model outputs predicted values for the next 24 hours, the system automatically pulls the actual data corresponding to the time stamp 24 hours later for deviation calculation.

[0051] It should be noted that the deviation calculation uses the relative error formula: Deviation = |Predicted Value - Actual Value| / Actual Value × 100%. The preset threshold is dynamically configured according to the risk level of the monitoring point. Exemplarily, for high-risk reservoirs, the threshold is set to a lower value to achieve sensitive early warning; for low-risk river channels, the threshold can be appropriately relaxed to reduce false alarms. The alarm rule supports a multi-level trigger mechanism: when the deviation is within the "prompt level" range, only the log is recorded and the operation and maintenance personnel are notified; when it reaches the "emergency level", the emergency system (such as the flood discharge gate control interface) is linked and multiple levels of responsible persons are notified.

[0052] In this embodiment, the message template is dynamically configured through a low-code platform, supporting the embedding of variable fields (such as {Prediction Deviation}, {Monitoring Point Location}) and visual charts. Exemplarily, the water level early warning email template contains the following content: Monitoring point identifier and location: A screenshot of the map generated by associating with the Geographic Information System (GIS); Comparison curve graph of predicted value and actual value: Generating trend visualization based on historical data and real-time data; Deviation analysis suggestion: Such as "The current deviation has reached 15%, it is recommended to start a patrol and check the status of the sensor".

[0053] Furthermore, the message notification system is called through the process engine to send in parallel via multiple-channel interfaces. It can be understood that the SMTP protocol interface, SMS gateway, and WeChat enterprise number service are only exemplary implementations, and other communication protocols (such as DingTalk robot, voice call) can be extended during actual deployment. Exemplarily, when an emergency alarm is triggered, the system simultaneously performs the following operations: Email notification: Sending the alarm content in HTML format to the operation and maintenance team's email; SMS push: Extracting key information (such as "The water level deviation of Reservoir A-002 exceeds the limit") and sending it to the responsible person's mobile phone via the SMS gateway; WeChat enterprise number message: Pushing an interactive card message, supporting one-click jump to the low-code platform to view details.

[0054] It should be noted that the sending status (success / failure), timestamp, recipient, and original content of all alarm messages are recorded in the message log database. Exemplarily, the log table structure contains fields such as alert_id (unique alarm identifier), send_time (sending time), receiver (recipient's email or mobile phone number), status_code (HTTP response status code), etc., for subsequent traceability and analysis.

[0055] In addition, users can retrieve historical records based on the time range, monitoring point, or alarm level through the custom query interface of the low-code platform. The system automatically converts the query results into visual charts, such as: Time distribution heat map: Showing the alarm frequency in different time periods to help identify periodic risks; Alarm type statistical chart: Classifying and counting by deviation level to assist in optimizing the model or adjusting the threshold.

[0056] Through the above embodiments, a closed-loop management from deviation judgment to alarm notification is achieved. For example, in the gas leakage monitoring scenario of a chemical industrial park, when the deviation between the predicted concentration and the actual value exceeds the limit, the system not only pushes an alarm to the on-duty personnel, but also automatically triggers the start command of the exhaust equipment, and records the whole process of the event in the log database, forming an auditable safety management chain.

[0057] In one embodiment of the present application, after sending an alarm notification based on a preset alarm mechanism, the method further includes: counting the number of trigger times of the alarm notification within a preset time period; when the number of trigger times exceeds a preset number threshold, adding the actual monitoring value after the future time period to the historical monitoring data set; based on the updated historical monitoring data set, using a sliding time window method to reconstruct the model input sequence, and performing a model compression operation on the time series prediction model to generate an optimized time series prediction model; deploying the optimized time series prediction model to the artificial intelligence platform through a model hot-loading interface to replace the original time series prediction model.

[0058] It can be understood that the "number of trigger times of the alarm notification within a preset time period" is an important indicator for measuring the degradation of the model prediction ability. Exemplarily, in the water level monitoring scenario, if a certain reservoir frequently triggers alarms for several consecutive days (such as exceeding the preset number within a single week), it indicates that the current model may not be able to accurately capture the new rules of water level changes (such as sudden heavy rainfall patterns or upstream flood discharge impacts). At this time, the system automatically starts the model optimization process to ensure that the prediction ability is synchronized with the real-time data trend.

[0059] It should be noted that when the system detects that the number of alarm trigger times exceeds the threshold, the process engine automatically performs the following operations: Data closed-loop feedback: Append the actual monitoring value after the future time period (such as the real water level value of the current day) to the historical data set to form an enhanced data set containing the latest business dynamics; Reconstruction of the model input sequence: Use the sliding time window method to re-partition the training set and the validation set for the updated data set. Exemplarily, if the original data set contains data for the past three years and new high-frequency alarm period data for one month is added, the system reconstructs the window sequence in chronological order to ensure that the model learns the latest change patterns; Lightweight retraining: Based on model compression techniques (quantization, pruning, distillation), iteratively optimize the time series prediction model. For example, the pruning operation removes neuron connections with low weight contribution in the new data, and the quantization operation reduces the model calculation precision from FP32 to INT8 to meet the deployment requirements of edge devices.

[0060] Furthermore, the optimized model seamlessly replaces the original model through a hot reload interface. It can be understood that the hot reload process adopts a gray release strategy to avoid service interruption. Exemplarily, the process engine routes a portion of the inference requests (e.g., 10%) to the new model for A / B testing. If its prediction deviation rate is lower than that of the old model, the traffic ratio is gradually increased until a complete switchover. During this process, the old and new model versions coexist on the artificial intelligence platform and are isolated and managed through version tags to ensure the operability of the rollback mechanism.

[0061] Specifically, the implementation of model hot reload relies on containerization technologies (such as Docker) and microservices architecture. Exemplarily, when the new model passes the verification, the system encapsulates it as an independent service instance and dynamically registers it to the inference service cluster through the API gateway. The process engine automatically updates the call endpoint to the new model version, while historical requests can still be processed by the old model until the session ends. This design significantly reduces the impact of model updates on business continuity.

[0062] Through the above implementation methods, data-driven model self-optimization is achieved. For example, in a river water level monitoring system, due to sudden changes in rainfall patterns during the rainy season, frequent alarms occurred. After the system automatically collected new data and retrained the model, the prediction deviation rate decreased from the original 12% to 5%, effectively reducing the number of false alarms. This process requires no manual intervention, fully demonstrating the intelligent advantages of the collaboration between the low-code platform and the AI model.

[0063] The above is the method embodiment proposed in this application. Based on the same inventive concept, the embodiments of this application also provide an intelligent warning device for low-code development, the structure of which is as Figure 2 shown.

[0064] Figure 2 This is a schematic diagram of the internal structure of an intelligent warning device for low-code development provided by the embodiment of this application. As Figure 2 shown, the device includes: At least one processor 201; And a memory 202 communicatively connected to the at least one processor; Wherein, the memory 202 stores instructions executable by the at least one processor. The instructions are executed by the at least one processor 201, enabling the at least one processor 201 to: Obtain historical monitoring data through a multi-source data interface and pre-train a time series prediction model loaded in the artificial intelligence platform based on the historical monitoring data; Create a data collection form in the low-code platform and establish an automated data processing process through the process engine; When a form data submission event is detected, a model inference service call instruction is triggered to input the current monitoring data and the corresponding historical time-series data into the time-series prediction model to obtain predicted values for a future time period; wherein, the historical time-series data is the historical monitoring data within a preset time period before obtaining the current monitoring data; Compare the predicted value with the actual monitoring value after the future time period. If the deviation exceeds a preset threshold, send an alarm notification based on a preset alarm mechanism.

[0065] Some embodiments of the present application provide corresponding to Figure 1 A non-volatile computer storage medium for intelligent early warning for low-code development stores computer-executable instructions, and the computer-executable instructions are set as: Obtain historical monitoring data through a multi-source data interface, and pre-train the time-series prediction model loaded in the artificial intelligence platform based on the historical monitoring data; Create a data collection form in the low-code platform and establish an automated data processing process through a process engine; When a form data submission event is detected, a model inference service call instruction is triggered to input the current monitoring data and the corresponding historical time-series data into the time-series prediction model to obtain predicted values for a future time period; wherein, the historical time-series data is the historical monitoring data within a preset time period before obtaining the current monitoring data; Compare the predicted value with the actual monitoring value after the future time period. If the deviation exceeds a preset threshold, send an alarm notification based on a preset alarm mechanism.

[0066] Each embodiment in the present application is described in a progressive manner. The same or similar parts among the embodiments can be referred to each other, and the key points of each embodiment are the differences from other embodiments. In particular, for the embodiments of the Internet of Things devices and media, since they are basically similar to the method embodiments, the description is relatively simple, and the relevant parts can refer to the partial description of the method embodiments.

[0067] The systems and media provided by the embodiments of the present application correspond one-to-one with the methods. Therefore, the systems and media also have beneficial technical effects similar to those of the corresponding methods. Since the beneficial technical effects of the methods have been described in detail above, the beneficial technical effects of the systems and media will not be repeated here.

[0068] Those skilled in the art should understand that the embodiments of the present application can be provided as a method, a system, or a computer program product. Therefore, the present application can take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware aspects. Moreover, the present application can take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) that contain computer-usable program code.

[0069] The present application is described with reference to the flowcharts and / or block diagrams of methods, apparatuses (systems), and computer program products according to the embodiments of the present application. It should be understood that each flow and / or block in the flowchart and / or block diagram can be implemented by computer program instructions, and the combination of the flows and / or blocks in the flowchart and / or block diagram can also be implemented by computer program instructions. These computer program instructions can be provided to the processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing devices to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing devices generate a device for implementing the functions specified in Figure 1 one or more flows and / or blocks Figure 1 one or more blocks.

[0070] These computer program instructions can also be stored in a computer-readable memory that can direct a computer or other programmable data processing device to work in a specific manner, so that the instructions stored in the computer-readable memory generate a manufactured article including an instruction device, and the instruction device implements the functions specified in Figure 1 one or more flows and / or blocks Figure 1 one or more blocks.

[0071] These computer program instructions can also be loaded onto a computer or other programmable data processing device, so that a series of operation steps are executed on the computer or other programmable device to generate a computer-implemented process. Thus, the instructions executed on the computer or other programmable device provide steps for implementing the functions specified in Figure 1 one or more flows and / or blocks Figure 1 one or more blocks.

[0072] In a typical configuration, a computing device includes one or more processors (CPUs), an input / output interface, a network interface, and a memory.

[0073] The memory may include non-permanent memory in the computer-readable medium, random access memory (RAM), and / or non-volatile memory in the form of, for example, read-only memory (ROM) or flash memory (flashRAM). The memory is an example of a computer-readable medium.

[0074] A computer-readable medium includes both permanent and non-permanent, removable and non-removable media that can implement information storage by any method or technology. The information can be computer-readable instructions, data structures, program modules, or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, compact disc read-only memory (CD-ROM), digital versatile disc (DVD) or other optical storage, magnetic cassettes, magnetic tape magnetic disk storage or other magnetic storage devices, or any other non-transitory medium that can be used to store information that can be accessed by a computing device. As defined herein, a computer-readable medium does not include transitory computer-readable media such as modulated data signals and carrier waves.

[0075] It should also be noted that the term "comprising", "including" or any other variant thereof is intended to cover non-exclusive inclusion, such that a process, method, article or apparatus comprising a series of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article or apparatus. Without further limitation, an element defined by the statement "comprising an..." does not exclude the presence of additional identical elements in the process, method, article or apparatus comprising the element.

[0076] The above description is only for the embodiments of the present application and is not intended to limit the present application. For those skilled in the art, various changes and modifications can be made to the present application. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application shall be included within the scope of the claims of the present application.

Claims

1. An intelligent early warning method for low-code development, characterized in that, The method includes: Obtaining historical monitoring data through a multi-source data interface, and pre-training a time series prediction model loaded in an artificial intelligence platform based on the historical monitoring data; Creating a data collection form in a low-code platform, and establishing an automated data processing process through a process engine; When a form data submission event is detected, triggering a model inference service call instruction to input the current monitoring data and the corresponding historical time series data into the time series prediction model to obtain predicted values for a future period; wherein, the historical time series data is the historical monitoring data within a preset time period before obtaining the current monitoring data; Comparing the predicted values with the actual monitoring values after the future period, and if the deviation exceeds a preset threshold, sending an alarm notification based on a preset alarm mechanism.

2. The intelligent early warning method for low-code development according to claim 1, wherein Before pre-training the time series prediction model loaded in the artificial intelligence platform based on the historical monitoring data, the method further includes: Identifying missing values and outliers in the historical monitoring data; Filling the missing values using linear interpolation method, and performing outlier removal processing on the outliers; Dividing the processed historical monitoring data into a training set and a validation set.

3. The intelligent early warning method for low-code development according to claim 2, wherein Pre-training the time series prediction model loaded in the artificial intelligence platform based on the historical monitoring data specifically includes: Based on the training set, constructing a model input sequence using the sliding time window method to convert the historical monitoring data within each time window into a multi-dimensional feature vector; Synchronously implementing model compression operations during model training; wherein, the model compression operations include performing fixed-point quantization processing on the neural network weights, pruning redundant neuron connections, and migrating model parameters through knowledge distillation technology; After the model training is completed, correcting the initially trained model based on the validation set to obtain a converged time series prediction model.

4. An intelligent early warning method for low-code development according to claim 1, characterized in that Correcting the initially trained model based on the validation set specifically includes: Calculating the accuracy rate, recall rate, and F1 score metrics of the time series prediction model based on the validation set data; When any of the metrics is lower than the corresponding preset metric threshold, adjusting the model hyperparameters and re-executing the pre-training process until a converged time series prediction model is obtained.

5. An intelligent early warning method for low-code development according to claim 1, characterized in that, Triggering the model inference service call instruction specifically includes: Parsing the measurement point identifier and data collection timestamp in the form data; Associating a preset geographical location database according to the measurement point identifier to obtain the storage path of the historical monitoring data corresponding to the measurement point; Based on the data collection timestamp and the preset time period, extracting the historical time series data from the storage path, arranging it in the order of time windows, and inputting it into the time series prediction model.

6. The intelligent early warning method for low-code development according to claim 1, characterized in that Sending an alarm notification based on a preset alarm mechanism specifically includes: Generating alarm message content including the current monitoring timestamp, monitoring point identifier, predicted value deviation amount, and monitoring change trend chart; Converting the alarm message content into email body, SMS text, and enterprise application message formats respectively by parallelly calling the SMTP protocol interface, SMS gateway API, and WeChat enterprise account service interface; According to the preset interval range where the deviation amount is located, dynamically selecting a list of message recipients and performing multi-level alarm push operations.

7. An intelligent early warning method for low-code development according to claim 6, characterized in that After sending an alarm notification based on a preset alarm mechanism, the method further includes: Storing the alarm message content, sending timestamp, recipient identifier, and sending status code in a message log database; In response to a query request submitted by a user through a low-code platform, retrieving historical alarm records from the message log database and generating a visual analysis chart including time distribution and alarm type statistics based on the records.

8. An intelligent early warning method for low-code development according to claim 1, characterized in that After sending an alarm notification based on a preset alarm mechanism, the method further includes: Counting the number of times the alarm notification is triggered within a preset time period; When the number of trigger times exceeds a preset threshold, adding the actual monitoring value after the future period to the historical monitoring data set; Based on the updated historical monitoring data set, reconstructing the model input sequence using a sliding time window method and performing a model compression operation on the time series prediction model to generate an optimized time series prediction model; Deploying the optimized time series prediction model to an artificial intelligence platform through a model hot reload interface to replace the original time series prediction model.

9. An intelligent early warning device for low-code development, characterized in that, The device includes: At least one processor; And a memory communicatively connected to the at least one processor; Wherein, the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to execute a method according to any one of claims 1-8.

10. A non - volatile computer storage medium for intelligent early warning in low - code development, storing computer - executable instructions, characterized in that, When the computer-executable instructions are executed, a method according to any one of claims 1-8 is implemented.

Images