Internet of Things data stable transmission optimization method, system, device and medium

Abstract

The invention discloses an internet of things data stable transmission optimization method, system and device and a medium, and relates to the field of internet of things, and the method comprises the steps: generating a plurality of test scenes through a multi-network simulation model; inputting the plurality of test scenes into the target Internet of Things to obtain Internet of Things test data; inputting the Internet of Things test data into the transmission performance prediction model to obtain performance prediction data; inputting the performance prediction data into a data mining and analysis model to obtain data mining information; generating a data stability analysis report according to the data mining information; and determining an abnormal point according to the data stability analysis report to form a detection and maintenance scheme. The stability, the reliability and the operation and maintenance efficiency of the Internet of Things system are effectively improved, and the method has wide application prospects and remarkable engineering practice values.

Description

Technical Field

[0001] This specification relates to the field of the Internet of Things. More specifically, the present invention relates to a method, system, device, and medium for optimizing the stable transmission of Internet of Things data. Background Art

[0002] With the rapid development of Internet of Things (IoT) technology, a large number of heterogeneous terminal devices conduct data interaction through radio access networks, core networks, and the Internet. The problem of data transmission stability in IoT systems has increasingly become a key factor affecting the reliability, real-time performance, and availability of the systems.

[0003] In the prior art, static scenario simulation or limited scenario testing is usually adopted to evaluate the data transmission performance of the Internet of Things. However, due to the single test environment and the lack of simulation of dynamic and variable network conditions, it is difficult to truly reflect the actual transmission behavior of IoT systems in complex environments, resulting in insufficient performance evaluation coverage, lagging anomaly detection, and untimely system maintenance response.

[0004] In addition, traditional methods mostly rely on rule setting or static thresholds for anomaly judgment, and cannot effectively adapt to the dynamic changes of network states and the interactive influence of multiple factors, lacking intelligent and systematic stability analysis and fault tracing capabilities.

[0005] Therefore, there is an urgent need for a method that can comprehensively evaluate the end-to-end transmission performance and optimize the stability of IoT systems based on intelligent learning and in-depth analysis mechanisms in a variable network environment. Summary of the Invention

[0006] A series of simplified concepts are introduced in the Summary of the Invention section, which will be further elaborated in the Detailed Implementation section. The Summary of the Invention section of the present invention does not mean to attempt to define the key features and essential technical features of the claimed technical solution, nor does it mean to attempt to determine the protection scope of the claimed technical solution.

[0007] In a first aspect, the present invention proposes a method for optimizing the stable transmission of Internet of Things data. The method includes: Generating multiple test scenarios through a multi-network simulation model, where the multiple test scenarios include radio access network test scenarios, core network test scenarios, and Internet test scenarios; Inputting the multiple test scenarios into a target Internet of Things to obtain Internet of Things test data; Inputting the Internet of Things test data into a transmission performance prediction model to obtain performance prediction data, where the performance prediction data includes transmission delay, packet loss rate, and throughput; Inputting the performance prediction data into a data mining and analysis model to obtain data mining information; Generate a data stability analysis report based on the above data mining information; Determine the abnormal points according to the above data stability analysis report to form a detection and maintenance plan.

[0008] In a feasible implementation manner, the above generates multiple test scenarios through a multi-network simulation model, including: Record the current network state, where the above network state includes node distribution information, link connection information, and link parameter information; Based on the above current network state, select a target action from the action space, where the above target action includes adding nodes, deleting nodes, modifying link delays, adjusting link bandwidths, introducing link failures, and changing node mobility models; After executing the above target action, update the network state based on the action impact, and evaluate the complexity, authenticity, and testability of the generated scenario according to the preset above reward function; Use a reinforcement learning algorithm to train the policy so that the optimal action can be selected under different network states to generate test scenarios that meet the target complexity requirements; Output the above multiple finally generated test scenarios when the termination condition is met, where the above termination condition includes reaching the preset number of nodes, number of links, complexity level, and maximum number of generated steps.

[0009] In a feasible implementation manner, the above selecting a target action from the action space includes: Generate a random number r between 0 and 1 at each decision moment; When the random number r is greater than the preset exploration rate ε, select the action with the maximum action value under the current network state; When the random number r is less than or equal to the exploration rate ε, randomly select an action from the action space, where the above exploration rate ε is used to control the balance between exploration and exploitation and decays dynamically according to the training process.

[0010] In a feasible implementation manner, the above using a reinforcement learning algorithm to train the policy so that the optimal action can be selected under different network states to generate test scenarios that meet the target complexity requirements includes: Perceive the current network state within each decision cycle, where the above network state includes the number of nodes, link connection conditions, and link parameter information; Based on the above current network state, select a target action according to the preset action selection strategy that emphasizes both exploration and exploitation; Based on the action execution result, evaluate the immediate reward according to the complexity, authenticity, and testability of the above scenario; The policy parameters are updated through a reinforcement learning algorithm, where the reinforcement learning algorithm includes one or more of Q-Learning, Deep Q-Network, and Proximal Policy Optimization.

[0011] In a feasible implementation, the transmission performance prediction model includes a multi-modal feature encoding layer, a feature fusion layer, a joint modeling layer, and a performance metric prediction layer; The above-mentioned multi-modal feature encoding layer is used to perform feature encoding on the network test data of the Internet of Things system. The multi-modal feature encoding layer includes: A network topology encoding module for extracting topological structure features based on node distribution information and link connection information; A link dynamic feature encoding module for extracting temporal dynamic features based on the delay, packet loss rate, and bandwidth information of the link changing over time; A traffic statistics feature encoding module for extracting static features based on the traffic statistics characteristics of nodes and links; The above-mentioned feature fusion layer is used to perform weighted fusion on the above-mentioned multi-modal features based on the self-attention mechanism to generate a unified comprehensive feature representation; The above-mentioned joint modeling layer is used to perform spatial relationship modeling using a graph convolutional network and time series modeling using a bidirectional long short-term memory network based on the above-mentioned comprehensive feature representation; The above-mentioned performance metric prediction head is used to predict the transmission delay, packet loss rate, and throughput performance metrics respectively according to the joint modeling results.

[0012] In a feasible implementation, the training process of the above-mentioned data mining and analysis model specifically includes: Perform data cleaning and standardization processing on the standard performance prediction data, and extract feature information, where the above-mentioned feature information includes the current value of the performance metric, the change rate, the lag feature, and the feature information of the local fluctuation characteristics; Based on the above-mentioned feature information, use a dynamic Bayesian network for causal structure learning to construct a dynamic causal relationship graph of the performance metrics evolving over time; Based on the above-mentioned dynamic causal relationship graph, identify abnormal nodes and abnormal metric changes, and trace the abnormal propagation path to infer the abnormal cause chain; Introduce a contrast learning mechanism and use real abnormal links and random links for training and optimization; Adopt an incremental update strategy to update the dynamic causal relationship graph and the anomaly detection model in real time based on the newly added performance prediction data to obtain the above-mentioned data mining and analysis model.

[0013] In a feasible implementation, the above-mentioned based on the above-mentioned feature information, using a dynamic Bayesian network for causal structure learning to construct a dynamic causal relationship graph of the performance metrics evolving over time, includes: Divide the above performance characteristic information into multiple time slices in chronological order, where each of the above time slices corresponds to a time step; Within each of the above time slices, determine the synchronous dependence relationship between different performance metrics at the same time step based on conditional independence testing, and generate a synchronous dependence structure; Between adjacent above time slices, based on the characteristics of the change of performance metrics over time, learn the time transfer dependence relationship between time steps, and generate a transfer structure across time steps; Combine the above synchronous dependence structure and the above transfer structure across time steps, and based on the Bayesian estimation method, learn the conditional probability distribution parameters between the nodes in the dynamic causal relationship graph to form a dynamic causal relationship graph describing the time series evolution characteristics of performance metrics.

[0014] In a second aspect, the present invention proposes an optimized system for stable transmission of Internet of Things data, including: A first generation unit for generating multiple test scenarios through a multi-network simulation model, where the multiple test scenarios include a radio access network test scenario, a core network test scenario, and an Internet test scenario; A first acquisition unit for inputting the multiple test scenarios into a target Internet of Things to obtain Internet of Things test data; A second acquisition unit for inputting the Internet of Things test data into a transmission performance prediction model to obtain performance prediction data, where the performance prediction data includes transmission delay, packet loss rate, and throughput; A third acquisition unit for inputting the performance prediction data into a data mining and analysis model to obtain data mining information; A second generation unit for generating a data stability analysis report according to the data mining information; A determination unit for determining abnormal points according to the data stability analysis report to form a detection and maintenance plan.

[0015] In a third aspect, the present invention proposes an electronic device, including: a memory and a processor, characterized in that the processor is used to implement the steps of the optimized method for stable transmission of Internet of Things data as described in any item of the first aspect when executing the computer program stored in the memory.

[0016] In a fourth aspect, the present invention proposes a computer-readable storage medium, on which a computer program is stored, characterized in that the computer program implements the steps of the optimized method for stable transmission of Internet of Things data as described in any item of the first aspect when executed by a processor.

[0017] In summary, the optimized method for stable transmission of Internet of Things data provided by the present invention can systematically construct test scenarios with multiple levels, multiple variations, and multiple instability factors in the radio access network, core network, and Internet by introducing a multi-network simulation model. Compared with traditional static single-scenario testing, it greatly improves the authenticity and diversity of the test environment, thus significantly enhancing the ability of the Internet of Things system to evaluate data transmission stability in different complex environments. In the performance prediction stage, the present invention jointly models through multi-modal feature encoding, graph neural network, and time-series neural network, comprehensively considering multi-dimensional performance indicators such as transmission delay, packet loss rate, and throughput, realizing the comprehensive prediction of end-to-end transmission performance, overcoming the problem of information loss caused by single-index local evaluation in the prior art, and effectively improving the accuracy and comprehensiveness of the overall system performance perception. Further, in the process of data mining and analysis, the present invention uses a dynamic Bayesian network to model the causal relationship evolving over time between performance indicators, and combines a contrastive learning mechanism to optimize the inference of the abnormal cause chain, which can accurately identify abnormal nodes and their propagation links, infer the cause of the abnormality, greatly improving the intelligence, interpretability, and localization accuracy of anomaly detection, and solving the problems of lagging anomaly detection and inability to trace the cause in the existing methods. At the same time, the present invention introduces an incremental update strategy to dynamically correct the causal relationship model and anomaly detection model based on newly added performance data, ensuring that the system can adapt to network state changes and performance evolution trends in real time, and still maintaining high-efficiency and accurate data mining and stability analysis capabilities in the face of dynamic changes in network structure and frequent access and exit of devices. Finally, the present invention generates a data stability analysis report by mining information, determines the abnormal points based on this, forms a detection and maintenance plan, and constructs an integrated closed-loop optimization mechanism from performance evaluation, anomaly detection to maintenance decision-making, effectively improving the stability, reliability, and operation and maintenance efficiency of the Internet of Things system, and having broad application prospects and significant engineering practice value.

[0018] The optimized method for stable transmission of Internet of Things data proposed by the present invention. Other advantages, objectives, and features of the present invention will be partially reflected by the following description, and partially will be understood by those skilled in the art through the research and practice of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] By reading the detailed description of the preferred embodiments below, various other advantages and benefits will become clear to those of ordinary skill in the art. The drawings are only for the purpose of showing the preferred embodiments and are not considered to limit this specification. Moreover, throughout the drawings, the same reference numerals are used to represent the same components. In the drawings: Figure 1 It is a schematic flowchart of an optimized method for stable transmission of Internet of Things data provided by an embodiment of the present invention; Figure 2Schematic diagram of an optimized system structure for stable transmission of Internet of Things data provided by an embodiment of the present invention; Figure 3 Schematic diagram of the structure of an electronic device provided by an embodiment of the present invention. Detailed implementation manners

[0020] The terms "first", "second", "third", "fourth", etc. (if any) in the description and claims of the present invention and the above-mentioned drawings are used to distinguish similar objects and do not necessarily describe a specific order or sequence. It should be understood that such data can be interchanged under appropriate circumstances so that the embodiments described herein can be implemented in an order different from that shown or described herein. In addition, the terms "comprising" and "having" and any variations thereof are intended to cover non-exclusive inclusion. For example, a process, method, system, product or device comprising a series of steps or units does not necessarily have to be limited to those steps or units clearly listed, but may include other steps or units not clearly listed or inherent to these processes, methods, products or devices. The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments.

[0021] Please refer to Figure 1 , which is a schematic flowchart of an optimized method for stable transmission of Internet of Things data provided by an embodiment of the present invention, and specifically may include: S110. Generate a variety of test scenarios through a multi-network simulation model, where the above-mentioned variety of test scenarios include a radio access network test scenario, a core network test scenario, and an Internet test scenario; Exemplarily, first, use a multi-network simulation model to generate diverse test scenarios. The above-mentioned variety of test scenarios cover a radio access network test scenario, a core network test scenario, and an Internet test scenario. By simulating different network levels and transmission environments, it is possible to effectively construct various typical application scenarios such as an access end, a relay end, and public network transmission, providing rich test conditions for the subsequent stability evaluation of the Internet of Things system.

[0022] S120. Input the above-mentioned variety of test scenarios into the target Internet of Things to obtain Internet of Things test data; Exemplarily, after generating the test scenarios, input the variety of test scenarios into the target Internet of Things system to be tested for simulation or actual deployment. By running Internet of Things terminal devices under different network conditions, collect and record end-to-end data transmission information to form Internet of Things test data, providing basic data support for performance prediction and anomaly analysis.

[0023] S130. Input the above Internet of Things test data into the transmission performance prediction model to obtain performance prediction data, where the performance prediction data includes transmission delay, packet loss rate, and throughput; Exemplarily, then, input the obtained Internet of Things test data into the transmission performance prediction model. Through feature encoding, feature fusion, and joint modeling processes, the above transmission performance prediction model outputs multi-dimensional performance prediction data including transmission delay, packet loss rate, and throughput, so as to preliminarily judge the transmission quality performance of the Internet of Things system under different test scenarios.

[0024] S140. Input the above performance prediction data into the data mining and analysis model to obtain data mining information; Exemplarily, subsequently, input the performance prediction data into the data mining and analysis model. Through a series of data analysis methods such as causal structure learning, anomaly detection, and anomaly cause chain reasoning, identify potential abnormal patterns, performance degradation trends, and system stability risk points, and generate preliminary data mining information.

[0025] S150. Generate a data stability analysis report based on the above data mining information; Exemplarily, based on the obtained data mining information above, further generate a data stability analysis report for the Internet of Things system. This report includes content such as the overall system stability score, main abnormal node and link identifiers, anomaly cause inference results, risk level assessment, etc., providing a basis for subsequent maintenance decisions and optimization adjustments.

[0026] S160. Determine the abnormal points according to the above data stability analysis report to form a detection and maintenance plan.

[0027] Exemplarily, finally, according to the abnormal nodes, abnormal links, and corresponding abnormal reasons marked in the data stability analysis report, determine the abnormal points that need to be maintained, and formulate targeted detection and maintenance plans, including but not limited to operation strategies such as node replacement, link reconstruction, and parameter optimization, to improve the stability and reliability of the Internet of Things system in a complex and changing network environment.

[0028] In summary, the optimized method for stable transmission of Internet of Things data provided by the present invention can systematically construct test scenarios with multi-level, multi-variation, and multi-instability factors of radio access network, core network, and Internet by introducing a multi-network simulation model. Compared with traditional static single-scenario testing, the authenticity and diversity of the test environment are greatly improved, thereby significantly enhancing the data transmission stability evaluation ability of the Internet of Things system in different complex environments. In the performance prediction stage, the present invention jointly models through multi-modal feature encoding, graph neural network, and time-series neural network, comprehensively considers multi-dimensional performance indicators such as transmission delay, packet loss rate, and throughput, realizes the comprehensive prediction of end-to-end transmission performance, overcomes the problem of information loss caused by single-index local evaluation in the prior art, and effectively improves the accuracy and comprehensiveness of the overall system performance perception. Further, in the process of data mining and analysis, the present invention uses a dynamic Bayesian network to model the causal relationship evolving over time between performance indicators, combines a contrastive learning mechanism to optimize the inference of the abnormal cause chain, can accurately identify abnormal nodes and their propagation links, infer the cause of the abnormality, greatly improves the intelligence, interpretability, and localization accuracy of anomaly detection, and solves the problems of lagging anomaly detection and inability to trace the cause in the existing methods. At the same time, the present invention introduces an incremental update strategy, dynamically corrects the causal relationship model and anomaly detection model based on newly added performance data, ensures that the system can adapt to network state changes and performance evolution trends in real time, and can still maintain efficient and accurate data mining and stability analysis capabilities in the face of dynamic changes in network structure and frequent access and exit of devices. Finally, the present invention generates a data stability analysis report by mining information, determines abnormal points based on this, forms a detection and maintenance plan, and constructs an integrated closed-loop optimization mechanism from performance evaluation, anomaly detection to maintenance decision-making, effectively improving the stability, reliability, and operation and maintenance efficiency of the Internet of Things system, and having broad application prospects and significant engineering practice value.

[0029] In a feasible implementation manner, the generation of multiple test scenarios through the multi-network simulation model includes: Record the current network state, where the above network state includes node distribution information, link connection information, and link parameter information; Based on the above current network state, select a target action from the action space, where the above target action includes adding nodes, deleting nodes, modifying link delay, adjusting link bandwidth, introducing link faults, and changing the node mobility model; After executing the above target action, update the network state based on the action impact, and evaluate the complexity, authenticity, and testability of the generated scenario according to the preset above reward function; Use a reinforcement learning algorithm to train the strategy so that the optimal action can be selected in different network states to generate a test scenario that meets the target complexity requirements; Output the finally generated multiple test scenarios when the termination conditions are met, where the termination conditions include reaching the preset number of nodes, number of links, complexity level, and maximum number of generation steps.

[0030] Exemplarily, first, the system records the network status information of the current Internet of Things system in real time or periodically. The above network status at least includes: node distribution information (such as the number of nodes, node position coordinates, node connection density); link connection information (such as link topology relationship, link existence mark); link parameter information (such as link bandwidth, link delay, link packet loss rate, link stability index). By recording the current network status, an initial environment model of the Internet of Things system is established to provide basic data support for subsequent action decision-making and test scenario evolution.

[0031] Based on the above-recorded current network status, the system selects a target action from a preset action space (Action Space). The action space includes but is not limited to: adding a node (Add Node); deleting a node (Remove Node); modifying the link delay (Modify Link Delay); adjusting the link bandwidth (Adjust LinkBandwidth); introducing a link failure (Inject Link Failure); changing the node mobility model (ChangeMobility Model).

[0032] After executing the selected target action, the system immediately updates the network status based on the action impact. The update process can define a corresponding state transition function (StateTransitionFunction) according to the action type. The network status update can be expressed as:

[0033] where, represents the network status at time step t; represents the target action selected at time step t; represents the new network status reached after the action ; f(·) is the state transition function, which is specifically determined by the action type and the network response mechanism.

[0034] The reward function R( , ) can be constructed in the following form:

[0035] where, is the reward value after executing the action at time step t; is the complexity score of the generated scenario; is the authenticity score of the generated scenario; Test is the testability score of the generated scenario; are the historical mean and standard deviation of the complexity score; are the historical mean and standard deviation of the authenticity score; are the historical mean and standard deviation of the testability score; is the S-shaped normalization function (such as the Sigmoid function), defined as: is used to standardize the input features to the interval, improving the comparability of different score items; are the weight coefficients corresponding to each score item, , and ; is the penalty coefficient; is the indicator function, which outputs 1 when the generated network state violates the constraint conditions (such as node isolation, link disconnection, topology break), and outputs 0 otherwise, playing a role in punishing illegal states.

[0036] The reward function adopts standardization processing to prevent different indicators from affecting the balance of reward calculation due to different numerical scales; uses Sigmoid non-linear normalization to make the reward change smoothly and avoid oscillations in the training process; introduces a constraint penalty term to force the network to avoid unavailable or invalid test scenarios during the evolution process; On this basis, a reinforcement learning algorithm (such as Q-Learning, DQN, PPO, etc.) is used to train the action selection strategy, enabling the system to adaptively select the optimal action in different network states and guiding the scenario to evolve towards the target complexity, authenticity, and testability requirements. The goal of strategy training is to maximize the cumulative reward:

[0037] where, π* is the optimal strategy; π is the action selection strategy; is the discount factor , controlling the importance of future rewards; T is the maximum number of steps in the generation process.

[0038] Through continuous training, the reinforcement learning agent can optimize the action decision path and generate high-quality test scenarios that meet the target complexity requirements.

[0039] During the test scenario generation process, it is detected in real time whether the preset termination conditions are met. The above termination conditions include but are not limited to: the number of nodes reaches the target range; the number of links meets the preset density; the scenario complexity level reaches the set threshold; the maximum number of generation steps limit is reached.

[0040] When any termination condition is met, stop the scenario evolution and output the finally generated multiple test scenarios for subsequent data transmission testing and performance evaluation of the Internet of Things system.

[0041] Through the above specific implementation manners, the present invention can flexibly select scenario evolution actions based on the current network state, and dynamically generate multiple test scenarios that meet the requirements of target complexity, authenticity, and testability under the guidance of the reinforcement learning algorithm. Compared with the traditional static configuration or rule-driven test scenario generation methods, the present invention not only greatly improves the richness and adaptability of the test environment, but also significantly enhances the intelligence level of the test process, effectively improving the comprehensiveness and accuracy of the data transmission performance evaluation of the Internet of Things system under different network conditions. During the action selection process, the exploration and optimality can be balanced based on the reinforcement learning strategy.

[0042] In a feasible implementation manner, the above selection of the target action from the action space includes: At each decision-making moment, generate a random number r between 0 and 1; When the random number r is greater than the preset exploration rate ε, select the action with the maximum action value in the current network state; When the random number r is less than or equal to the exploration rate ε, randomly select an action from the action space, where the above exploration rate ε is used to control the balance between exploration and exploitation and dynamically decays according to the training process.

[0043] Exemplarily, at each decision-making moment t, the system generates a random number r uniformly distributed between 0 and 1 to determine the current action selection method. The generation of this random number can be implemented based on a standard pseudo-random number generator to ensure a certain degree of uncertainty is introduced in the decision-making process to promote exploration behavior.

[0044] Compare the generated random number r with the current preset exploration rate ε and execute the following two different action selection strategies: (1) When the random number r is greater than the exploration rate ε, the system selects the action A* with the maximum action value (Q value) in the current network state, that is:

[0045] where, represents the expected cumulative return of executing action A in the current state ; A is the current action space. At this time, the system tends to select the currently considered optimal action to maximize the immediate benefit and future return.

[0046] (2) When the random number r is less than or equal to the exploration rate ε, the system randomly selects an action A' from the action space A, that is:

[0047] Through random action selection, the agent is guided to explore new state transition paths, discover potentially better action sequences, and avoid getting stuck in local optima.

[0048] To achieve the natural evolution of the training process from exploration to exploitation, the system sets the exploration rate ε to decay dynamically with the training progress. The value of ε can be updated using the exponential decay formula:

[0049] where is the initial exploration rate, usually set to a relatively high value (such as 0.9); decay is the decay factor (such as 0.99), which controls the rate of decrease of the exploration rate at each step; step is the current number of training steps; is the lower limit of the minimum exploration rate, ensuring that there is still a small amount of exploration even in the later stage.

[0050] Through the above dynamic adjustment mechanism, extensive exploration is encouraged in the early stage of training, while in the later stage of training, it gradually focuses on the optimal action path, improving the stability and convergence of the strategy.

[0051] Through the action selection strategy adopted in this embodiment, during the generation process of the Internet of Things test scenario, it can effectively balance the contradiction between exploration and exploitation. It not only ensures a full search of the action space, avoids getting stuck in local optima in the early stage, but also can concentrate on selecting high-value actions in the later stage of training, accelerating the convergence of the strategy, and significantly improving the efficiency and quality of the test scenario evolution. At the same time, the dynamic decay mechanism of the exploration rate further enhances the self-adaptability of the system, flexibly adjusts the action selection tendency at different training stages, improves the intelligence and robustness of the test scenario generation process, and thus provides a solid support for optimizing the data stability of the Internet of Things system in a complex and changeable network environment.

[0052] In a feasible embodiment, training the strategy using the reinforcement learning algorithm enables the selection of optimal actions in different network states to generate test scenarios that meet the requirements of the target complexity, including: Perceive the current network state within each decision cycle, where the above network state includes the number of nodes, link connection conditions, and link parameter information; Based on the above current network state, select the target action according to the preset action selection strategy that emphasizes both exploration and exploitation; Based on the action execution result, evaluate the immediate reward according to the complexity, authenticity, and testability of the above scenario; Update the policy parameters through the reinforcement learning algorithm, where the above reinforcement learning algorithm includes one or more of Q-Learning, deep Q-network, and proximal policy optimization.

[0053] Exemplarily, within each decision cycle, the system perceives and extracts the network state information of the current Internet of Things environment in real time. The network state includes but is not limited to: the number of nodes (such as the total number of online nodes, node density distribution); the link connection situation (such as the link topology relationship between nodes, link connectivity rate); link parameter information (such as link bandwidth, link delay, link stability index, link packet loss rate). By perceiving the above multi-dimensional network state characteristics, it provides a comprehensive and accurate environmental input for action selection, ensuring that the action decision can make a reasonable response according to the actual situation of the current network.

[0054] After obtaining the current network state, the system selects the target action according to the preset Exploration-Exploitation Balance strategy, specifically including: adopting the ε-greedy strategy, or other action selection mechanisms that combine action value (such as Q-value) with random exploration; in the exploitation phase, preferentially select the action with the highest current action value (Q-value) to maximize the expected return; in the exploration phase, randomly select actions to guide the strategy to discover potentially better action sequences; the exploration rate ε decays dynamically according to the number of training steps, mainly for exploration in the initial stage and mainly for exploitation in the later stage. Through the above action selection strategy, the system can effectively balance the exploration of new paths and the utilization of optimal actions during the training process.

[0055] After executing the target action, the system comprehensively evaluates the changes in the scenario in three dimensions: Complexity, Realism, and Testability according to the updated network state, and calculates the immediate reward value.

[0056] The reward function can be exemplified as follows:

[0057] Among them, are the complexity, realism, and testability score functions respectively; are the weighting coefficients of each score item, , and ; is the illegal state penalty coefficient; is the indicator function, which outputs a penalty when the network state violates constraints such as structural connectivity or node validity.

[0058] Complexity score function is used to measure the complexity of the current network scenario, reflecting whether the network structure is complex and diverse enough. It can be constructed based on indicators such as the number of nodes, the number of links, the average number of layers, and the topology value of the network, such as:

[0059] where Nnodes is the number of current network nodes; Nlinks is the number of current network links; are the preset maximum number of nodes and maximum number of links (normalization benchmark); Entropy(G) is the network topology value, which is used to measure the network structure complexity (such as based on the degree distribution value); : the weighting coefficient of each index, , and .

[0060] Authenticity score function measures the consistency between the current generated scenario and the characteristics of the real Internet of Things deployment, and reflects the authenticity of the generated scenario. It can be based on the closeness of link parameters (delay, cloud packet rate, bandwidth, etc.) to the real statistical data, such as:

[0061] where MSE(Delay) is the mean square error between the current link delay and the real Internet of Things link delay distribution; MSE(Loss) is the mean square error of the packet loss rate; MSE(Bandwidth) is the mean square error of the bandwidth utilization rate; are the weighting coefficients of each feature item, , and .

[0062] Testability score function is used to measure the support of the current network scenario for the test task, and reflects its testability and analysis value. It can be calculated based on indicators such as connectivity, coverage rate, and path diversity, such as:

[0063] where ConnRatio(G) is the network connectivity rate (number of connected components / number of nodes), and the higher the connectivity rate, the better; PathDiversity(G) is the path diversity index between node pairs (such as the average number of reachable paths between node pairs); IsolatedNodes(G) is the number of isolated nodes; : weighting coefficient, , and . A high connectivity rate and good path diversity indicate that the network is more conducive to multi-path testing; more isolated nodes reduce the testability, so it is used as a negative penalty term.

[0064] The policy parameters are updated through a reinforcement learning algorithm, where the above reinforcement learning algorithm includes one or more of Q-Learning, deep Q network, and proximal policy optimization.

[0065] Through this embodiment, the present invention can, on the basis of dynamically perceiving the network state, adopt an action selection strategy that equally emphasizes exploration and exploitation, combine a multi-dimensional reward evaluation mechanism of complexity, authenticity, and testability, and continuously optimize the action decision-making process using a reinforcement learning algorithm, so as to intelligently generate high-quality and diverse test scenarios in different network environments. Compared with traditional rule-setting or static strategy methods, the present invention significantly improves the intelligence, self-adaptability, and optimization of test scenario evolution, effectively enhances the accuracy and coverage breadth of the transmission performance evaluation of the Internet of Things system, and has extremely high practical application value and promotion prospects.

[0066] In a feasible embodiment, the transmission performance prediction model includes a multi-modal feature encoding layer, a feature fusion layer, a joint modeling layer, and a performance metric prediction layer; The above-mentioned multi-modal feature encoding layer is used to perform feature encoding on the network test data of the Internet of Things system. The above-mentioned multi-modal feature encoding layer includes: A network topology encoding module for extracting topological structure features based on node distribution information and link connection information; A link dynamic feature encoding module for extracting temporal dynamic features based on the delay, packet loss rate, and bandwidth information of the link varying with time; A traffic statistics feature encoding module for extracting static features based on the traffic statistics characteristics of nodes and links; The above-mentioned feature fusion layer is used to perform weighted fusion on the above-mentioned multi-modal features based on the self-attention mechanism to generate a unified comprehensive feature representation; The above-mentioned joint modeling layer is used to perform spatial relationship modeling using a graph convolutional network and time series modeling using a bidirectional long short-term memory network on the basis of the above-mentioned comprehensive feature representation; The above-mentioned performance metric prediction head is used to predict transmission delay, packet loss rate, and throughput performance metrics respectively according to the joint modeling results.

[0067] Exemplarily, first, the system extracts and encodes the collected network test data through the multi-modal feature encoding layer. The above-mentioned multi-modal feature encoding layer includes the following sub-modules: (1) Network topology encoding module: used to extract topological structure features based on node distribution information and link connection information. Using features such as node position relationship, link connectivity matrix, node degree, and connectivity rate, And encoding the network structure through graph convolution (GCN) or node embedding methods (such as GraphSAGE) to obtain vector representations of each node and the overall topology.

[0068] (2) Link Dynamic Feature Encoding Module: It is used to extract temporal dynamic features based on the performance metrics of the link that change over time (such as link delay, packet loss rate, bandwidth change). A bidirectional long short-term memory network (Bi-LSTM) is adopted to capture the temporal evolution pattern and periodic change law of the link state, and obtain the representation of link-level dynamic behavior features.

[0069] (3) Traffic Statistics Feature Encoding Module: It is used to extract static features based on the traffic statistics characteristics of nodes and links (such as average traffic, peak traffic, traffic standard deviation). A multi-layer perceptron (MLP) is used to map and enhance the extracted statistics, and form a static feature vector describing traffic behavior.

[0070] Through the above three types of feature encoding from different sources, the system can comprehensively represent the network structure, link dynamic changes, and traffic characteristics, laying a foundation for subsequent feature fusion and prediction modeling.

[0071] Feature Fusion Layer: After completing multi-modal feature extraction, the system performs weighted fusion on the above multi-modal features through the feature fusion layer to generate a unified comprehensive feature representation. In the present invention, the self-attention mechanism (Self-Attention) is adopted. According to the importance of each modal feature to the target task (transmission performance prediction), different feature weights are dynamically allocated to achieve information interaction and fusion between features, and avoid information redundancy or conflict caused by simple concatenation. The fused comprehensive features not only contain network structure information, but also fuse dynamic link behaviors and traffic statistical patterns, and have stronger discriminability and adaptability.

[0072] Joint Modeling Layer: After feature fusion is completed, the system enters the joint modeling stage to further explore the deep-seated correlation relationships between features. The joint modeling layer includes: (1) Spatial Relationship Modeling: Based on the fused feature representation, a graph convolutional network (GCN) is used to model the spatial dependence relationships of nodes and links, and capture local neighborhood structure features and long-range dependence characteristics.

[0073] (2) Time Series Modeling: On the basis of spatial modeling, a bidirectional long short-term memory network (Bi-LSTM) or a temporal convolutional network (TCN) is further used to model the dynamic changes in the time dimension, and capture the trend changes and fluctuation patterns of performance metrics over time.

[0074] Through spatial-temporal joint modeling, the system can fully understand the complex and dynamically changing network characteristics in the Internet of Things environment.

[0075] Based on the comprehensive feature representation output by the performance metric prediction head in the joint modeling layer, an independent performance metric prediction sub-module (Multi-Head Prediction) is set up to predict the following performance metrics respectively: transmission delay (Delay); packet loss rate (Packet Loss Rate); throughput (Throughput). Each prediction head can use an independent multi-layer perceptron (MLP) for regression prediction to achieve specialized optimization and customized modeling for different performance metrics, improving prediction accuracy and robustness.

[0076] Through the transmission performance prediction model designed in this embodiment, the present invention realizes unified modeling based on multi-modal features, fully integrating network structure features, link dynamic behavior features, and traffic statistical features. And through spatio-temporal joint modeling, the correlation between data is deeply mined, enabling the prediction system to accurately and efficiently predict key performance metrics such as transmission delay, packet loss rate, and throughput in a complex and changing Internet of Things environment. Compared with traditional single-feature input or single-layer modeling methods, the present invention significantly improves the accuracy, stability, and adaptability of performance prediction, providing strong support for data stability analysis and optimization of Internet of Things systems, and having important engineering application value and promotion prospects.

[0077] In a feasible embodiment, the training process of the above data mining and analysis model specifically includes: Perform data cleaning and standardization processing on the standard performance prediction data, and extract feature information. Among them, the above feature information includes feature information such as the current value of the performance metric, the change rate, the lag feature, and the local fluctuation characteristic. Based on the above feature information, use a dynamic Bayesian network for causal structure learning to construct a dynamic causal relationship graph of the evolution of performance metrics over time. Based on the above dynamic causal relationship graph, identify abnormal nodes and abnormal metric changes, and trace the abnormal propagation path to infer the abnormal cause chain. Introduce a contrastive learning mechanism and use real abnormal links and random links for training and optimization. Adopt an incremental update strategy to update the dynamic causal relationship graph and the anomaly detection model in real time based on the newly added performance prediction data to obtain the above data mining and analysis model.

[0078] Exemplarily, first, the system performs data preprocessing operations on the collected standard performance prediction data. It mainly includes: Data cleaning: eliminating invalid, abnormal, and severely missing data records to ensure the quality of input data; Data standardization: normalizing or standardizing different performance metrics (such as latency, packet loss rate, throughput) to eliminate the influence between different dimensions and improve training stability; Feature extraction: Based on the cleaned and standardized data, extract the following four types of feature information: The current value of performance metrics (such as the latency value, packet loss rate value, throughput value at the current moment); Rate-of-change features (such as latency change rate, packet loss rate change rate, calculating the first-order difference); Lag features (such as the historical performance metric values of the previous n time steps to capture temporal dependencies); Local fluctuation characteristics (such as moving window mean, moving window standard deviation to measure the degree of metric fluctuation).

[0079] Through the above processing, a set of feature vectors with a unified format and consistent dimensions is formed, laying a foundation for causal structure modeling and anomaly detection.

[0080] After feature extraction is completed, the system conducts causal structure learning based on the Dynamic Bayesian Network (DBN). This causal relationship graph not only reveals the static associations between performance metrics, but also can depict the causal propagation characteristics that change over time.

[0081] After completing the causal relationship modeling, the system conducts anomaly detection and anomaly cause analysis based on the dynamic causal relationship graph, specifically including: Detecting abnormal fluctuations in performance metrics in the causal graph nodes (such as exceeding the prediction interval, abnormal amplification of the rate of change); Marking abnormal nodes and tracing their possible cause nodes and abnormal propagation paths along the causal chain; Forming a root cause chain to clarify the order of anomaly occurrence and the scope of influence. Through causal chain inference, it is possible to distinguish direct abnormal nodes from indirectly affected nodes, improving the accuracy of fault location and anomaly explanation.

[0082] To further improve the accuracy and robustness of root cause chain inference, the present invention introduces a contrastive learning mechanism, specifically including: Using the real abnormal links manually labeled or inferred in historical data as positive samples; Randomly generating pseudo-links that do not conform to causal logic as negative samples; Through training, minimizing the difference degree between the positive sample links and the negative sample links in the feature space; Optimizing the abnormal chain inference model to make it have better discrimination ability in complex scenarios. By introducing contrastive learning, not only the accuracy of abnormal chain inference is improved, but also the adaptability of the model to changes in abnormal types and network structures is enhanced.

[0083] Considering that the environment of the Internet of Things system changes frequently and dynamically, the present invention adopts an incremental learning strategy, which specifically includes: real-time inputting newly arrived performance prediction data into the training process; dynamically correcting the causal relationship graph structure and conditional probability parameters to adapt to the evolution trend of indicators; updating the anomaly detection model to maintain the detection sensitivity to new anomaly patterns; and avoiding the model aging and failure caused by traditional static training methods. Through the incremental update strategy, the system can learn online for a long time and continuously improve the timeliness and accuracy of anomaly detection and cause inference.

[0084] Through this embodiment, in the process of data stability analysis of the Internet of Things system, the present invention not only realizes the dynamic causal relationship modeling based on multi-feature information, but also combines anomaly detection with cause chain tracing, further introduces a contrast learning mechanism to optimize the inference accuracy, and maintains the long-term adaptability and effectiveness of the model through an incremental learning mechanism. Compared with traditional methods based on static features or rule settings, the present invention significantly improves the intelligence, interpretability and real-time performance of anomaly detection, effectively enhances the stability guarantee ability of the Internet of Things system in a complex dynamic environment, and has extremely high practical engineering application value and promotion potential.

[0085] In a feasible embodiment, based on the above feature information, a dynamic Bayesian network is used for causal structure learning to construct a dynamic causal relationship graph of the evolution of performance indicators over time, including: Dividing the above performance feature information into multiple time slices in chronological order, where each of the above time slices corresponds to a time step; Within each of the above time slices, based on conditional independence tests, determining the synchronous dependence relationship between different performance indicators at the same time step to generate a synchronous dependence structure; Between adjacent above time slices, based on the characteristics of the change of performance indicators over time, learning the time transfer dependence relationship between time steps to generate a transfer structure across time steps; Combining the above synchronous dependence structure and the above transfer structure across time steps, and based on the Bayesian estimation method, learning the conditional probability distribution parameters between the nodes in the dynamic causal relationship graph to form a dynamic causal relationship graph describing the temporal evolution characteristics of performance indicators.

[0086] Exemplarily, first, the system divides the extracted performance feature information according to the time sequence to form a series of continuous time slices, and each time slice corresponds to the performance state of the IoT system at a certain fixed time step. The length of the time slice can be set to seconds, minutes, or hours according to the application requirements; each slice contains the performance metric feature vectors of all nodes or links at this time step, such as latency, packet loss rate, throughput, and their derived features (change rate, volatility, etc.). Through the division of time slices, complex continuous time series data can be structured, facilitating subsequent independent modeling of synchronous dependencies and time transfer dependencies.

[0087] Within each time slice, the system learns the synchronous dependency relationships (Intra-Slice Dependencies) between different performance metrics at the same time step based on the conditional independence test method. The specific steps include: selecting any two performance metrics (such as latency and packet loss rate); testing whether they are conditionally independent at the current time step, with other performance metrics as the conditional set; if there is still a significant correlation between the two metrics after controlling other metrics, a synchronous causal edge is established between them; repeating the above process for all metric pairs, and finally generating the synchronous dependency structure at the current time step.

[0088] The conditional independence test can adopt standard statistical methods, such as Pearson chi-square test, mutual information test, or Z test based on Gaussian hypothesis, and can be flexibly selected according to the data distribution characteristics. Through synchronous dependency learning, the direct interaction relationships between performance metrics within a single time step can be revealed, providing a basis for local anomaly detection and chain inference.

[0089] After the synchronous dependency learning is completed, the system further learns the time transfer dependency relationships (Inter-Slice Dependencies) of performance metrics between adjacent time slices. The specific methods include: taking values of any performance metric A at time step t and time step t + 1 respectively to establish a transfer test; analyzing the influence relationship of the state of A at time step t on other performance metrics (such as B) at time step t + 1; if A at time t has a significant predictive power or causal influence on B at time t + 1, a cross-time step dependency edge is established between them; repeating the above operations to cover all metric pairs and forming a time transfer dependency structure.

[0090] By learning the time transfer dependencies, the system can capture the potential causal propagation chain of performance metrics evolving over time, supporting the modeling and prediction of abnormal trends and metric contagion effects.

[0091] After the synchronous dependence structure and the time-transfer dependence structure are respectively learned, the system combines the two to form a complete Dynamic Causal Graph. Synchronous dependence edges describe the immediate causal relationships within the same time step, and time-transfer edges describe the evolutionary causal relationships between adjacent time steps, thus forming a Directed Acyclic Graph (Dynamic Bayesian Network, DBN) model as a whole.

[0092] Subsequently, based on the existing data samples, the Bayesian estimation method is used to learn the conditional probability distribution (Conditional Probability Distribution, CPD) of each causal edge in the dynamic causal graph. Specifically, for each node, learn its conditional probability distribution given the states of its parent nodes; if the performance metric is a continuous value, a Gaussian distribution can be used to fit the conditional probability; if it is a discretized metric, a multinomial distribution can be used for modeling. After completing the parameter learning through Bayesian estimation, the dynamic causal graph can not only reflect the structural dependencies between the metrics, but also provide a quantitative estimate of the causal strength, providing a solid support for subsequent anomaly detection, causal chain inference, and system behavior prediction.

[0093] Through this embodiment, the present invention can dynamically construct a causal relationship graph of the performance metrics in the IoT system evolving over time based on the performance characteristic information, not only realizing the efficient modeling of synchronous causal relationships within the same time step, but also capturing the evolutionary causal paths across time steps, and accurately modeling the conditional probability relationships between the metrics in combination with the Bayesian estimation method, significantly improving the accuracy and interpretability of system anomaly pattern recognition, anomaly tracing, and behavior prediction. Compared with traditional static analysis or simple time series analysis methods, the present invention can more intelligently and adaptively cope with the complex and ever-changing performance evolution process in the IoT environment. It has broad application value and engineering practice significance.

[0094] As Figure 2 shown, the present invention also provides an IoT data stable transmission optimization system, including: A first generation unit 101, configured to generate multiple test scenarios through a multi-network simulation model, where the multiple test scenarios include a radio access network test scenario, a core network test scenario, and an Internet test scenario; A first acquisition unit 102, configured to input the multiple test scenarios into a target IoT to obtain IoT test data; A second acquisition unit 103, configured to input the IoT test data into a transmission performance prediction model to obtain performance prediction data, where the performance prediction data includes transmission delay, packet loss rate, and throughput; A third acquisition unit 104, configured to input the above performance prediction data into a data mining and analysis model to obtain data mining information; A second generation unit 105, configured to generate a data stability analysis report according to the above data mining information; A determination unit 106, configured to determine abnormal points according to the above data stability analysis report to form a detection and maintenance plan.

[0095] The above system can also perform the following steps: In a feasible implementation manner, the above-mentioned generation of multiple test scenarios through a multi-network simulation model includes: Record the current network state, where the above network state includes node distribution information, link connection information, and link parameter information; Based on the above current network state, select a target action from the action space, where the above target action includes adding nodes, deleting nodes, modifying link delays, adjusting link bandwidths, introducing link failures, and changing node mobility models; After performing the above target action, update the network state based on the action impact, and evaluate the complexity, authenticity, and testability of the generated scenario according to the preset above reward function; Use a reinforcement learning algorithm to train a policy so that the optimal action can be selected under different network states to generate test scenarios that meet the target complexity requirements; Output the finally generated above multiple test scenarios when the termination condition is met, where the above termination condition includes reaching the preset number of nodes, number of links, complexity level, and maximum number of generated steps.

[0096] In a feasible implementation manner, the above-mentioned selection of a target action from the action space includes: Generate a random number r between 0 and 1 at each decision moment; When the random number r is greater than the preset exploration rate ε, select the action with the maximum action value under the current network state; When the random number r is less than or equal to the exploration rate ε, randomly select an action from the action space, where the above exploration rate ε is used to control the balance between exploration and exploitation and dynamically decays according to the training process.

[0097] In a feasible implementation manner, the above-mentioned use of a reinforcement learning algorithm to train a policy so that the optimal action can be selected under different network states to generate test scenarios that meet the target complexity requirements includes: Perceive the current network state within each decision cycle, where the above network state includes the number of nodes, link connection conditions, and link parameter information; Based on the above current network state, select a target action according to a preset action selection strategy that emphasizes both exploration and exploitation; Based on the action execution result, evaluate the immediate reward according to the complexity of the above scenario, the above authenticity, and the above testability; Update the policy parameters through a reinforcement learning algorithm, where the above reinforcement learning algorithm includes one or more of Q-Learning, Deep Q-Network, and Proximal Policy Optimization.

[0098] In a feasible implementation manner, the transmission performance prediction model includes a multi-modal feature encoding layer, a feature fusion layer, a joint modeling layer, and a performance metric prediction layer; The above multi-modal feature encoding layer is used to perform feature encoding on the network test data of the Internet of Things system. The above multi-modal feature encoding layer includes: A network topology encoding module for extracting topological structure features based on node distribution information and link connection information; A link dynamic feature encoding module for extracting temporal dynamic features based on the delay, packet loss rate, and bandwidth information of the link changing over time; A traffic statistics feature encoding module for extracting static features based on the traffic statistics characteristics of nodes and links; The above feature fusion layer is used to perform weighted fusion on the above multi-modal features based on the self-attention mechanism to generate a unified comprehensive feature representation; The above joint modeling layer is used to perform spatial relationship modeling using a graph convolutional network and time series modeling using a bidirectional long short-term memory network based on the above comprehensive feature representation; The above performance metric prediction head is used to predict transmission delay, packet loss rate, and throughput performance metrics respectively according to the joint modeling result.

[0099] In a feasible implementation manner, the training process of the above data mining and analysis model specifically includes: Perform data cleaning and standardization processing on the standard performance prediction data, and extract feature information. Among them, the above feature information includes feature information such as the current value of the performance metric, the change rate, the lag feature, and the local fluctuation characteristics; Based on the above feature information, use a dynamic Bayesian network for causal structure learning to construct a dynamic causal relationship graph of the evolution of performance metrics over time; Based on the above dynamic causal relationship graph, identify abnormal nodes and abnormal metric changes, and trace the abnormal propagation path to infer the abnormal cause chain; Introduce a contrast learning mechanism and use real abnormal links and random links for training and optimization; Adopt an incremental update strategy, and update the dynamic causal relationship graph and the anomaly detection model in real time based on the newly added performance prediction data to obtain the above data mining and analysis model.

[0100] In a feasible implementation manner, based on the above characteristic information, a dynamic Bayesian network is used for causal structure learning to construct a dynamic causal relationship graph of the evolution of performance indicators over time, including: The above performance characteristic information is divided into multiple time slices in chronological order, where each of the above time slices corresponds to a time step; Within each of the above time slices, based on conditional independence tests, the synchronous dependence relationships between different performance indicators at the same time step are determined to generate a synchronous dependence structure; Between adjacent above time slices, based on the characteristics of the change of performance indicators over time, the time transfer dependence relationships between time steps are learned to generate a transfer structure across time steps; The above synchronous dependence structure and the above transfer structure across time steps are combined, and based on the Bayesian estimation method, the conditional probability distribution parameters between the nodes in the dynamic causal relationship graph are learned to form a dynamic causal relationship graph describing the temporal evolution characteristics of performance indicators.

[0101] As Figure 3 shown, the present invention also provides an electronic device 30, including a memory 310, a processor 320, and a computer program 311 stored on the memory 310 and executable on the processor. When the processor 320 executes the computer program 311, any step of the above method for optimizing the stable transmission of Internet of Things data is implemented.

[0102] The present invention also provides a computer program product, which includes a computer program or computer executable instructions. The computer program or computer executable instructions are stored in a computer-readable storage medium. The processor of the electronic device reads the computer program or computer executable instructions from the computer-readable storage medium, and the processor executes the computer program or computer executable instructions, so that the electronic device executes any step of the above method for optimizing the stable transmission of Internet of Things data of the present invention.

[0103] The above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that: they can still modify the technical solutions recorded in the foregoing embodiments, or perform equivalent replacements on some of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. An optimization method for stable transmission of Internet of Things data, characterized in that, Including: Generating multiple test scenarios through a multi-network simulation model, where the multiple test scenarios include radio access network test scenarios, core network test scenarios, and Internet test scenarios; Inputting the multiple test scenarios into a target Internet of Things to obtain Internet of Things test data; Inputting the Internet of Things test data into a transmission performance prediction model to obtain performance prediction data, where the performance prediction data includes transmission delay, packet loss rate, and throughput; Inputting the performance prediction data into a data mining and analysis model to obtain data mining information; Generating a data stability analysis report according to the data mining information; Determining abnormal points according to the data stability analysis report to form a detection and maintenance plan.

2. The optimized method for stable transmission of Internet of Things data according to claim 1, wherein The generating multiple test scenarios through a multi-network simulation model includes: Recording the current network state, where the network state includes node distribution information, link connection information, and link parameter information; Based on the current network state, selecting a target action from the action space, where the target action includes adding a node, deleting a node, modifying link delay, adjusting link bandwidth, introducing link faults, and changing the node mobility model; After executing the target action, updating the network state based on the action impact, and evaluating the complexity, authenticity, and testability of the generated scenario according to a preset reward function; Training a policy using a reinforcement learning algorithm so that the optimal action can be selected under different network states to generate test scenarios that meet the target complexity requirements; Outputting the finally generated multiple test scenarios when the termination condition is met, where the termination condition includes reaching a preset number of nodes, number of links, complexity level, and maximum generation steps.

3. The optimized method for stable transmission of Internet of Things data according to claim 2, wherein, The selecting a target action from the action space includes: Generating a random number r between 0 and 1 at each decision-making moment; When the random number r is greater than the preset exploration rate ε, selecting the action with the maximum action value under the current network state; When the random number r is less than or equal to the exploration rate ε, randomly selecting an action from the action space, where the exploration rate ε is used to control the balance between exploration and exploitation and decays dynamically according to the training process.

4. The method for optimizing the stable transmission of Internet of Things data according to claim 2, wherein The training a policy using a reinforcement learning algorithm so that the optimal action can be selected under different network states to generate test scenarios that meet the target complexity requirements includes: Perceiving the current network state within each decision-making cycle, where the network state includes the number of nodes, link connection conditions, and link parameter information; Based on the current network state, selecting a target action according to a preset action selection strategy that emphasizes both exploration and exploitation; Based on the action execution result, evaluating the immediate reward according to the complexity, authenticity, and testability of the scenario; Updating the policy parameters through a reinforcement learning algorithm, where the reinforcement learning algorithm includes one or more of Q-Learning, deep Q network, and proximal policy optimization.

5. The method for optimizing the stable transmission of Internet of Things data according to claim 1, characterized in that The transmission performance prediction model includes a multi-modal feature encoding layer, a feature fusion layer, a joint modeling layer, and a performance metric prediction layer; The multi-modal feature encoding layer is used to perform feature encoding on the network test data of the Internet of Things system, and the multi-modal feature encoding layer includes: A network topology encoding module, which is used to extract topology structure features based on node distribution information and link connection information; A link dynamic feature encoding module, which is used to extract temporal dynamic features based on the delay, packet loss rate, and bandwidth information of the link changing over time; A traffic statistics feature encoding module, which is used to extract static features based on the traffic statistics characteristics of nodes and links; The feature fusion layer is used to perform weighted fusion on the above multi-modal features based on the self-attention mechanism to generate a unified comprehensive feature representation; The joint modeling layer is used to perform spatial relationship modeling using a graph convolutional network and time series modeling using a bidirectional long short-term memory network based on the comprehensive feature representation; The performance metric prediction head is used to predict transmission delay, packet loss rate, and throughput performance metrics respectively according to the joint modeling results; 6. The method for optimizing the stable transmission of Internet of Things data according to claim 1, characterized in that The training process of the data mining and analysis model specifically includes: Performing data cleaning and standardization processing on the standard performance prediction data, and extracting feature information, where the feature information includes feature information such as the current value, change rate, lag feature, and local fluctuation characteristics of the performance metric; Based on the feature information, using a dynamic Bayesian network for causal structure learning to construct a dynamic causal relationship graph of the evolution of performance metrics over time; Based on the dynamic causal relationship graph, identifying abnormal nodes and abnormal metric changes, and tracing the abnormal propagation path to infer the abnormal cause chain; Introducing a contrastive learning mechanism, and using real abnormal links and random links for training optimization; Adopting an incremental update strategy, and updating the dynamic causal relationship graph and the anomaly detection model in real time based on the newly added performance prediction data to obtain the data mining and analysis model.

7. The optimized method for stable transmission of Internet of Things data according to claim 6, wherein The constructing a dynamic causal relationship graph of the evolution of performance metrics over time by using a dynamic Bayesian network based on the feature information includes: Dividing the performance feature information into multiple time slices in chronological order, where each time slice corresponds to a time step; Within each time slice, determining the synchronous dependence relationship between different performance metrics at the same time step based on conditional independence tests to generate a synchronous dependence structure; Between adjacent time slices, learning the time transfer dependence relationship between time steps based on the characteristics of performance metrics changing over time to generate a cross-time step transfer structure; Combining the synchronous dependence structure and the cross-time step transfer structure, and based on the Bayesian estimation method, learning the conditional probability distribution parameters between nodes in the dynamic causal relationship graph to form a dynamic causal relationship graph describing the temporal evolution characteristics of performance metrics.

8. An optimized system for stable transmission of Internet of Things data, characterized in that, It includes: A first generation unit, which is used to generate multiple test scenarios through a multi-network simulation model, where the multiple test scenarios include a radio access network test scenario, a core network test scenario, and an Internet test scenario; A first acquisition unit, which is used to input the multiple test scenarios into the target Internet of Things to obtain Internet of Things test data; A second acquisition unit, which is used to input the Internet of Things test data into the transmission performance prediction model to obtain performance prediction data, where the performance prediction data includes transmission delay, packet loss rate, and throughput; A third acquisition unit, configured to input the performance prediction data into a data mining and analysis model to obtain data mining information; A second generation unit, configured to generate a data stability analysis report according to the data mining information; A determination unit, configured to determine abnormal points according to the data stability analysis report to form a detection and maintenance plan.

9. An electronic device, comprising: A memory and a processor, characterized in that when the processor executes a computer program stored in the memory, the steps of the Internet of Things data stable transmission optimization method according to any one of claims 1-7 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, the steps of the Internet of Things data stable transmission optimization method according to any one of claims 1-7 are implemented.

Images