Internet of Things data analysis system based on artificial intelligence

Abstract

The invention discloses an Internet of Things data analysis system based on artificial intelligence, and relates to the technical field of telecommunication, the system analyzes the coupling effect in signal propagation through a path selection and prediction module and an AI algorithm, and outputs an environmental influence function TW; and the dimension reduction feature data set F is combined with the environmental influence function TW to calculate and output a prediction path optimization parameter Pout. The intelligent path selection model not only considers factors such as signal intensity attenuation, noise interference and building reflection, but also combines the influence of terrain and weather, so that the prediction accuracy is greatly improved. And the dynamic path adjustment switching module realizes intelligent adjustment of the communication path based on the quality score comparison of the current path Pcurr and the predicted optimal path Pout and the setting of a threshold value M. And the response speed and the decision quality of the system in the case of network condition change are enhanced.

Description

Technical Field

[0001] The present invention relates to the field of electrical communication technology, and in particular to an Internet of Things data analysis system based on artificial intelligence. Background Art

[0002] The rapid development of IoT and AI technologies has promoted the construction of intelligent communication systems, which provides new solutions for the optimization of modern wireless communication networks. In this context, cellular communication network management in the field of telecommunications technology has gradually become a hot topic in research and application, especially in 5G and future network architectures, where the optimization of communication paths between cellular towers is particularly important. The selection and optimization of communication paths for cellular towers not only affects the efficiency of data transmission, but also determines the stability of the network and user experience. Therefore, the IoT data analysis system based on artificial intelligence has become an important technical means to solve the problem of optimizing communication paths between cellular towers, achieving efficient path selection and management through multi-source data collection and intelligent analysis.

[0003] At present, the management and optimization of communication paths between cellular towers mainly rely on preset parameters and manual control, lacking real-time, adaptive intelligent adjustment capabilities. This management method that relies on preset rules is powerless in the face of dynamic environmental changes such as weather influences, geographical obstacles, and network load fluctuations. It is difficult to respond and adjust paths quickly, resulting in reduced communication quality. In addition, most existing path optimization technologies use single data source analysis, cannot integrate multi-dimensional data for in-depth analysis and prediction, and lack the ability to fully understand and process real-time environmental factors and complex signal characteristics. This limitation will especially lead to delays in path selection and transmission interruptions in scenarios with high-speed data transmission and dense users. Summary of the invention

[0004] In view of the deficiencies in the prior art, the present invention provides an Internet of Things data analysis system based on artificial intelligence, which solves the problems mentioned in the background technology.

[0005] To achieve the above objectives, the present invention is implemented through the following technical solutions: including an acquisition and processing module, a feature extraction module, a path selection and prediction module, a dynamic path adjustment and switching module and a self-learning optimization module;

[0006] The collection and processing module collects IoT communication data in real time by installing IoT sensor groups and related programming interfaces API on cellular towers, and transmits the IoT communication data to the central control center, where the IoT communication data is preprocessed to obtain a standardized data set Dfeat;

[0007] The feature extraction module extracts the standardized data set Dfeat, performs feature extraction to obtain a reduced-dimensional feature data set F, and constructs a relational database to store the reduced-dimensional feature data set F;

[0008] The path selection and prediction module obtains the environmental impact function TW by analyzing the coupling effect of signal propagation, and then uses the AI ​​algorithm to build a path selection analysis and prediction model. The environmental impact function TW is combined with the dimension reduction feature data set F, and input into the path selection analysis and prediction model, and the associated calculation is performed to output the predicted path optimization parameter Pout;

[0009] The dynamic path adjustment switching module collects the communication quality data of the current path Pcurr and the predicted path optimization parameter Pout, constructs a path quality scoring formula, outputs the current path quality score Q(Pcurr) and the predicted preferred path quality score Q(Pout), sets a threshold value M, compares and evaluates the current path quality score Q(Pcurr) and the predicted preferred path quality score Q(Pout), and generates a switching logic;

[0010] The self-learning optimization module is used to monitor the network performance data of the cellular tower after the path switching in real time after the switching logic is executed, and to perform reinforcement learning updates on the path selection analysis and prediction model based on the network performance data.

[0011] Preferably, the acquisition and processing module includes a data acquisition unit and a data processing unit;

[0012] The data collection unit collects IoT communication data of the communication quality between cellular towers in real time by installing IoT sensors on each cellular tower in combination with relevant API programming interfaces;

[0013] The IoT sensor includes a wireless signal analyzer, a spectrum scanning tool, and a traffic analysis tool;

[0014] The API programming interface includes a weather forecast API programming interface and a geographic information API programming interface;

[0015] The IoT communication data includes signal strength attenuation coefficient SAC, noise interference intensity NIS, building reflection coefficient BRC, real-time traffic fluctuation coefficient RTC, terrain blocking factor TOF and weather influence factor WIF.

[0016] Preferably, the data processing unit is used to transmit the collected IoT communication data to the central control center through the IoT gateway, and perform preprocessing in the central control center. The preprocessing method includes data cleaning, missing value completion and data synchronization. After the preprocessing is completed, Z-score is used for standardization to obtain a standardized data set Dfeat.

[0017] Preferably, the feature extraction module includes a feature extraction unit and a data storage unit;

[0018] The feature extraction unit extracts features with upper variance limits from the standardized data set Dfeat by principal component analysis PCA, performs dimensionality reduction processing, and then uses active window technology to perform feature decomposition on the time-varying features of the real-time traffic fluctuation coefficient RTC and the noise interference intensity NIS, extracts short-term variation trends, and then summarizes to obtain a reduced-dimensionality feature data set F;

[0019] The data storage unit is used to construct a relational database and set a write port and a write port. Through the write port, the obtained dimension reduction feature data set F is arranged in chronological order and stored in the relational database.

[0020] Preferably, the path selection and prediction module includes an influence function analysis unit and a path optimization unit;

[0021] The influence function analysis unit extracts the terrain blocking factor TOF and the weather influence factor WIF in the dimension reduction feature data set F in real time through the write port of the relational database, performs coupling calculation, and outputs the environmental influence function TW;

[0022] The environmental impact function TW is calculated and outputted by the following algorithm formula;

[0023]

[0024] Where log represents the logarithmic function, e represents the exponential function, λ represents the weather impact factor WIF adjustment parameter, and W0 represents the standard weather impact threshold.

[0025] Preferably, the path optimization unit constructs a path selection analysis and prediction model by using an LSTM network of an AI algorithm to predict the path of communication signals between future cellular towers, and then extracts the dimension reduction feature data set F in real time through the write port of the relational database, combines the environmental impact function TW and inputs it into the path selection analysis and prediction model, performs associated calculations and outputs the predicted path optimization parameter Pout;

[0026] The predicted path optimization parameter Pout is calculated and output by the following path selection analysis and prediction model;

[0027]

[0028] In the formula, p represents the path, argmin pIt represents finding the path p that makes the entire path selection analysis and prediction model reach the lower limit value, SAC(p) represents the signal strength attenuation coefficient of the p-th path, NIS(p) represents the noise interference intensity of the p-th path, BRC(p) represents the building reflection coefficient of the p-th path, RTC(p) represents the real-time traffic fluctuation coefficient of the p-th path, TW(p) represents the environmental impact function of the p-th path, α, β, γ, δ and ε represent the preset weight values ​​of the signal strength attenuation coefficient SAC, the noise interference intensity NIS, the building reflection coefficient BRC, the real-time traffic fluctuation coefficient RTC and the environmental impact function TW, respectively, and α+β+γ+δ+ε=1, and its specific value is set by the user.

[0029] Preferably, the dynamic path adjustment switching module includes a communication quality monitoring unit, a communication quality analysis unit and a switching unit;

[0030] as stated;

[0031] The communication quality monitoring unit collects the communication quality data of the current path Pcurr and the predicted path optimization parameter Pout in real time through the built-in network detection tool on the cellular tower, and then transmits it to the IoT gateway, and writes it into the relational database through the write port for storage;

[0032] The communication quality data includes the real-time data throughput DT(Pcurr) of the p-th path and the real-time network delay ND(Pcurr) of the p-th path, as well as the predicted data throughput DT(Pout) of the p-th path and the predicted network delay ND(Pout) of the p-th path.

[0033] Preferably, the communication quality analysis unit is used to construct a path quality scoring formula, input the collected communication quality data into the path quality scoring formula, calculate and output the current path quality score Q (Pcurr) and the predicted preferred path quality score Q (Pout) respectively, and analyze the communication quality of the current path Pcurr and the predicted path preferred parameter Pout;

[0034] The current path quality score Q(Pcurr) and the predicted preferred path quality score Q(Pout) are both calculated and output by the following path quality score formula;

[0035]

[0036] Preferably, the switching unit sets a threshold value M through a user, combines the threshold value with the current path quality score Q (Pcurr) and compares and evaluates the predicted preferred path quality score Q (Pout), analyzes the deviation between the current path communication quality and the preferred path communication quality, and generates a path switching logic based on the evaluation result. The specific evaluation content is as follows;

[0037] When the predicted preferred path quality score Q(Pout)> the current path quality score Q(Pcurr)+threshold value M, the first switching logic is executed and the predicted path is automatically switched;

[0038] When the predicted optimal path quality score Q(Pout)≤the current path quality score Q(Pcurr)+the threshold value M, the first switching logic is executed to continue to maintain the current communication path.

[0039] Preferably, the self-learning optimization module collects the communication quality data of the path after the switching logic is executed through a network detection tool, marks the timestamp, writes the data into a relational database through a write port, and stores the data in the order of the timestamp;

[0040] When using the policy gradient algorithm to calculate the gradient of the output loss function, the signal strength attenuation coefficient weight value α, noise interference intensity weight value β, building reflection coefficient weight value γ, real-time traffic fluctuation coefficient weight value δ and environmental impact function weight value ε of the path selection analysis and prediction model are updated, and the weight update value △w of the i-th weight value is obtained. i ;

[0041] The weight update amount Δw i The output is calculated by the following policy gradient algorithm;

[0042]

[0043] In the formula, η represents the learning rate, which is used to control the step size of each weight update, wi represents the weight vector of the i-th weight value of the path selection analysis and prediction model, represents the partial derivative, Q(Pout∨Pcurr) represents the current path quality score Q(Pcurr) or the predicted preferred path quality score Q(Pout), and Qtarget represents the target score.

[0044] The present invention provides an Internet of Things data analysis system based on artificial intelligence. It has the following beneficial effects:

[0045] (1) The system collects communication quality-related data in real time through the IoT sensor group and API interface deployed on the cellular tower, and performs data preprocessing in the central control center to obtain the standardized data set Dfeat. The integration of this acquisition and processing module and the efficient data transmission process enable the system to quickly adapt to the dynamically changing network environment. Standardized data and subsequent feature extraction modules enable the system to effectively reduce computational complexity and ensure the timeliness and accuracy of data analysis. By acquiring and processing multi-source data in real time, the system improves the real-time performance of path selection and ensures the stability and efficiency of the communication path.

[0046] (2) The system uses the combination of the reduced-dimensional feature data set F and the environmental impact function TW through the path selection and prediction module, analyzes the coupling effect in signal propagation through the AI ​​algorithm, and outputs the predicted path optimization parameter Pout. This intelligent path selection model not only takes into account factors such as signal strength attenuation, noise interference, and building reflection, but also combines the influence of terrain and weather, greatly improving the accuracy of the prediction. The dynamic path adjustment switching module realizes the intelligent adjustment of the communication path based on the comparison of the quality scores of the current path Pcurr and the predicted preferred path Pout, as well as the setting of the threshold value M. This module enhances the system's response speed and decision-making quality when facing changes in network conditions.

[0047] (3) After the path switching logic is executed, the system's self-learning optimization module monitors and records the network performance data after the path switching, and uses this data to reinforce learning to update the weight parameters of the path selection model. This real-time learning mechanism continuously updates the model through the policy gradient algorithm, allowing the system to gradually optimize its path selection logic. Through this module, the system has the ability to adapt and continuously optimize, effectively reducing the volatility of network performance and the uncertainty in path selection, and improving the accuracy of future path selection and communication quality. This not only improves the stability of the network, but also reduces the maintenance costs caused by manual adjustment and troubleshooting. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] Figure 1 The present invention is a flowchart of an Internet of Things data analysis system based on artificial intelligence. DETAILED DESCRIPTION

[0049] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0050] Example 1

[0051] See also Figure 1 The present invention provides an IoT data analysis system based on artificial intelligence. To achieve the above purpose, the present invention is implemented through the following technical solutions: including an acquisition and processing module, a feature extraction module, a path selection and prediction module, a dynamic path adjustment and switching module, and a self-learning optimization module;

[0052] The collection and processing module collects IoT communication data in real time by installing IoT sensor groups and related programming interface APIs on cellular towers, and transmits the IoT communication data to the central control center, where the IoT communication data is preprocessed to obtain a standardized data set Dfeat;

[0053] The feature extraction module extracts the standardized data set Dfeat, performs feature extraction to obtain the reduced dimension feature data set F, and constructs a relational database to store the reduced dimension feature data set F;

[0054] The path selection and prediction module obtains the environmental impact function TW by analyzing the coupling effect of signal propagation, and then uses the AI ​​algorithm to build a path selection analysis and prediction model. The environmental impact function TW is combined with the dimension reduction feature data set F, and input into the path selection analysis and prediction model to perform correlation calculations and output the predicted path optimization parameter Pout;

[0055] The dynamic path adjustment switching module collects the communication quality data of the current path Pcurr and the predicted path optimization parameter Pout, constructs the path quality scoring formula, outputs the current path quality score Q(Pcurr) and the predicted preferred path quality score Q(Pout), sets the threshold value M, compares and evaluates the current path quality score Q(Pcurr) and the predicted preferred path quality score Q(Pout), and generates the switching logic;

[0056] The self-learning optimization module is used to monitor the network performance data of the cellular tower after the path switching in real time after the switching logic is executed, and to perform reinforcement learning updates on the path selection analysis and prediction model based on the network performance data.

[0057] In this embodiment, the system collects IoT communication data in real time on the cellular tower through the acquisition and processing module and performs preprocessing to obtain the standardized data set Dfeat, providing high-quality data input for subsequent analysis. The feature extraction module performs dimensionality reduction processing on the data set, extracts the core dimensionality reduction feature data set F and stores it in the relational database, providing improved computational efficiency for path selection. The path selection and prediction module obtains the environmental impact function TW by analyzing the coupling effect of signal propagation, and then uses the AI ​​algorithm to build a path selection analysis and prediction model. The optimal path Pout is analyzed and predicted by combining the dimensionality reduction feature data set F with the environmental impact function TW, so that the system can respond to complex network conditions more intelligently. The dynamic path adjustment switching module flexibly switches the communication path by comparing the quality score of the current path Pcurr with the predicted path in real time to ensure communication stability and efficiency. Finally, the self-learning optimization module realizes continuous optimization and adaptability improvement of the model by monitoring the performance of the path after switching and performing reinforcement learning updates. Compared with the prior art, the system significantly improves the automation and real-time performance of path selection. It solves the problem of lack of real-time multi-source data analysis and dynamic adjustment capabilities in traditional methods, realizes adaptive path optimization, and improves the stability, response speed and resource utilization of the communication network. Overall, the system can quickly respond to environmental changes and network requirements in complex environments, reduce communication interruptions and maintenance costs, and significantly improve communication quality and user experience.

[0058] Example 2

[0059] This embodiment is explained in Example 1, please refer to Figure 1 ,Specifically: the acquisition and processing module includes a data acquisition unit and a data processing unit;

[0060] The data collection unit collects IoT communication data on the communication quality between cellular towers in real time by installing IoT sensors on each cellular tower in combination with relevant API programming interfaces;

[0061] IoT sensors include wireless signal analyzers, spectrum scanning tools, and traffic analysis tools;

[0062] The API programming interface includes the weather forecast API programming interface and the geographic information API programming interface;

[0063] IoT communication data includes signal strength attenuation coefficient SAC, noise interference intensity NIS, building reflection coefficient BRC, real-time traffic fluctuation coefficient RTC, terrain blocking factor TOF and weather impact factor WIF;

[0064] The strength attenuation coefficient SAC is obtained by using the wireless signal analyzer of the base station signal monitoring equipment to monitor and record the strength changes of the signal during propagation;

[0065] Noise Interference Strength NIS scans the radio spectrum through spectrum scanning tools to identify and quantify the strength of interference sources;

[0066] The real-time traffic fluctuation coefficient (RTC) monitors the data traffic and traffic fluctuations on network devices through the traffic analysis tool SolarWinds;

[0067] The building reflection coefficient BRC and terrain blocking factor TOF are included in the LIDAR light detection and ranging of the geographic information system GIS software through the geographic information system API programming interface to obtain the building height, material and density of the structural data of urban buildings, predict the characteristics of reflection and multipath propagation to obtain the building reflection coefficient BRC;

[0068] At the same time, ArcGIS, a geographic information system (GIS) software, is used to process and analyze the terrain data around the cell tower to obtain the terrain blocking factor (TOF);

[0069] The weather impact factor WIF is seamlessly connected to the meteorological platform through the weather forecast API programming interface to update data in real time.

[0070] The data processing unit is used to transmit the collected IoT communication data to the central control center through the IoT gateway, and perform preprocessing in the central control center. The preprocessing methods include data cleaning, missing value completion and data synchronization. After the preprocessing is completed, the Z-score is used for standardization to obtain the standardized data set Dfeat.

[0071] In this embodiment, the system realizes the real-time collection of key communication parameters such as signal strength attenuation coefficient SAC, noise interference intensity NIS, building reflection coefficient BRC, real-time traffic fluctuation coefficient RTC, terrain blocking factor TOF and weather influence factor WIF through the IoT sensors installed on each cellular tower and the integrated programming interface API. The data processing unit is responsible for transmitting data to the central control center through the IoT gateway, where data cleaning, missing value completion and data synchronization are performed, and finally Z-score standardization is used to generate a high-quality standardized data set Dfeat. This module realizes real-time collection and efficient preprocessing of multi-dimensional data, greatly improving the data acquisition accuracy and processing speed of the communication network. Compared with traditional technical means, the system has significant improvements in collection accuracy, data types and preprocessing efficiency. Through the seamless integration of IoT sensors and programming interface API, the system realizes comprehensive monitoring of environmental and network factors, and ensures the accuracy and consistency of data input through data cleaning and standardization.

[0072] Example 3

[0073] This embodiment is explained in Example 2. Please refer to Figure 1 ,Specifically: the feature extraction module includes a feature extraction unit and a data storage unit;

[0074] The feature extraction unit extracts the features with the upper limit of variance from the standardized data set Dfeat through principal component analysis PCA, performs dimensionality reduction processing to reduce the computational complexity, and then uses the active window technology to decompose the time-varying features of the real-time traffic fluctuation coefficient RTC and the noise interference intensity NIS, extracts the short-term change trend, and then summarizes to obtain the reduced-dimensionality feature data set F;

[0075] The data storage unit is used to construct a relational database and set a write port and a write port. Through the write port, the obtained dimension reduction feature data set F is arranged in chronological order and stored in the relational database.

[0076] In this embodiment, the feature extraction unit of the system extracts the upper limit feature of the variance from the standardized data set Dfeat through principal component analysis PCA, realizes dimensionality reduction processing, and significantly reduces the computational complexity of the data. At the same time, the module uses active window technology to perform feature decomposition on time-varying features such as real-time traffic fluctuation coefficient RTC and noise interference intensity NIS, extracts short-term change trends and summarizes them to form a reduced dimension feature data set F. This method ensures the efficiency of key feature extraction and data dimensionality reduction, which helps to carry out subsequent analysis and modeling faster. The data storage unit provides stable data storage and retrieval functions by building a relational database, and stores the reduced dimension feature data set F in chronological order, which is convenient for subsequent real-time data extraction and analysis. Compared with existing feature extraction and storage technologies, this module combines PCA dimensionality reduction and active window technology to achieve efficient analysis and processing of dynamic data, reduce the time of model training and calculation, and improve the ability to capture data changes. The optimization of the data storage unit strengthens the stability of the system in data access and processing, ensuring the timeliness and consistency of the data. This improvement greatly enhances the system's real-time response capability, provides more accurate input data for path selection and prediction, and ultimately improves the system's overall performance and intelligence level in communication optimization.

[0077] Example 4

[0078] This embodiment is explained in Example 3, please refer to Figure 1 ,Specifically: Path selection and prediction module includes an ,influence function analysis unit and a path optimization unit;

[0079] The influence function analysis unit extracts the terrain blocking factor TOF and the weather influence factor WIF in the dimension reduction feature data set F in real time through the write port of the relational database, performs coupling calculation, and outputs the environmental influence function TW;

[0080] The environmental impact function TW is calculated and output by the following algorithm formula;

[0081]

[0082] Wherein, log represents the logarithmic function, e represents the exponential function, λ represents the weather impact factor WIF adjustment parameter, and W0 represents the standard weather impact threshold, for example, 20km / h. When the WIF wind speed approaches or exceeds 20km / h, the value of the function will change rapidly, indicating that the impact of weather factors on signal propagation begins to become significant.

[0083] The path optimization unit uses the LSTM network of the AI ​​algorithm to build a path selection analysis and prediction model to predict the optimal path of the communication signal between the future cellular towers. Then, through the write port of the relational database, it extracts the dimension reduction feature data set F in real time, combines it with the environmental impact function TW and inputs it into the path selection analysis and prediction model to perform correlation calculations and output the predicted path optimization parameter Pout.

[0084] The predicted path optimization parameter Pout is calculated and output by the following path selection analysis and prediction model;

[0085]

[0086] Where p represents the path, which represents the different communication paths between cell towers, and argmin p It represents finding the path p that makes the entire path selection analysis and prediction model reach the lower limit value, SAC(p) represents the signal strength attenuation coefficient of the p-th path, NIS(p) represents the noise interference intensity of the p-th path, BRC(p) represents the building reflection coefficient of the p-th path, RTC(p) represents the real-time traffic fluctuation coefficient of the p-th path, TW(p) represents the environmental impact function of the p-th path, α, β, γ, δ and ε represent the preset weight values ​​of the signal strength attenuation coefficient SAC, the noise interference intensity NIS, the building reflection coefficient BRC, the real-time traffic fluctuation coefficient RTC and the environmental impact function TW, respectively, and α+β+γ+δ+ε=1, and its specific value is set by the user.

[0087] In this embodiment, the system realizes intelligent analysis and prediction of communication paths between cellular towers through a path selection and prediction module composed of an influence function analysis unit and a path optimization unit. The influence function analysis unit extracts the terrain blocking factor TOF and the weather influence factor WIF in the dimensionality reduction feature data set F in real time, and performs coupling calculation to output the environmental influence function TW. This function comprehensively analyzes the impact of signal propagation and weather conditions on communication through complex mathematical formulas, ensuring accurate modeling of environmental changes on different paths. The path optimization unit constructs a path selection analysis and prediction model by using an AI algorithm based on an LSTM network, inputs the dimensionality reduction feature data set F and the environmental influence function TW into the model for correlation calculation, and outputs the predicted path optimization parameter Pout to determine the optimal communication path. Through this path selection analysis, the system can intelligently and dynamically evaluate the signal transmission characteristics and environmental impacts of different paths, and finally select the path with the best comprehensive performance. Compared with the existing path selection technology, this module realizes the automation and intelligence of multi-factor comprehensive analysis. It introduces real-time calculation of complex environmental variables such as terrain and weather in the analysis, predicts future communication conditions through the LSTM algorithm, and improves the accuracy and adaptability of path selection. Compared with the traditional static path selection method, this module can more effectively predict and respond to dynamic factors in signal propagation, greatly reducing the risk of communication interruption and signal quality degradation. Overall, this improvement provides a more stable and efficient path selection solution for the communication network, significantly improving network stability and user experience.

[0088] Example 5

[0089] This embodiment is explained in Example 4. Please refer to Figure 1 ,Specifically: the dynamic path adjustment switching module includes a ,communication quality monitoring unit, a communication quality analysis unit and a ,switching unit;

[0091] The communication quality monitoring unit collects the communication quality data of the current path Pcurr and the predicted path optimization parameter Pout in real time through the built-in network detection tool on the cellular tower, and then transmits it to the IoT gateway, and writes it into the relational database for storage through the write port;

[0092] The communication quality data includes the real-time data throughput DT(Pcurr) of the p-th path and the real-time network delay ND(Pcurr) of the p-th path, as well as the predicted data throughput DT(Pout) of the p-th path and the predicted network delay ND(Pout) of the p-th path.

[0093] The communication quality analysis unit is used to construct a path quality scoring formula, input the collected communication quality data into the path quality scoring formula, calculate and output the current path quality score Q (Pcurr) and the predicted preferred path quality score Q (Pout), and analyze the communication quality of the current path Pcurr and the predicted path preferred parameter Pout;

[0094] The current path quality score Q(Pcurr) and the predicted optimal path quality score Q(Pout) are both calculated and output by the following path quality score formula;

[0095]

[0096] The switching unit sets a threshold value M through the user, combines the threshold value with the current path quality score Q (Pcurr) and compares and evaluates the predicted preferred path quality score Q (Pout), analyzes the deviation between the current path communication quality and the preferred path communication quality, and generates the path switching logic based on the evaluation results. The specific evaluation contents are as follows;

[0097] When the predicted preferred path quality score Q(Pout)> the current path quality score Q(Pcurr)+threshold value M, it means that the predicted preferred path quality score Q(Pout) is significantly higher than the current path, and still has a large advantage after adding the threshold value M. Therefore, the system should perform path switching to improve communication quality and network performance. At this time, the first switching logic is executed to automatically switch to the predicted path;

[0098] When the predicted preferred path quality score Q(Pout) ≤ the current path quality score Q(Pcurr) + threshold value M, it means that the difference between the score of the optimal path and the score of the current path is not enough to exceed the threshold value M. This means that although the performance of the current path may be slightly worse than the optimal path, the difference is not large enough to switch. At this time, the first switching logic is executed to continue to maintain the current communication path.

[0099] In this embodiment, the system is composed of a communication quality monitoring unit, a communication quality analysis unit and a switching unit through a dynamic path adjustment switching module, which greatly improves the real-time management and adaptive adjustment capabilities of inter-cell tower communication. The communication quality monitoring unit collects the communication quality data of the current path Pcurr and the predicted path Pout in real time through the network detection tool built into the cellular tower, including data throughput DT and network delay ND. These data are transmitted and stored in a relational database through the IoT gateway for further analysis. The communication quality analysis unit calculates and outputs the current path quality score Q (Pcurr) and the predicted preferred path quality score Q (Pout) through the path quality scoring formula, and makes a detailed evaluation of the communication quality of the current path and the predicted path. The switching unit compares the path score according to the preset threshold value M to determine whether the path switching logic needs to be executed to optimize the communication quality. When the quality score of the predicted path is significantly higher than that of the current path, the system automatically switches to a better path to ensure network performance and communication stability. Compared with traditional technologies, this module realizes comprehensive monitoring, accurate analysis and intelligent decision-making of path quality, overcoming the problems of slow response and inflexible switching logic of the existing system. Through real-time data monitoring and adaptive threshold value comparison, the module can quickly identify changes in the quality of the communication path and automatically perform adjustments, avoiding unnecessary frequent switching and effectively reducing the risk of communication interruption and network performance degradation. The overall improvement brings higher network stability, optimized communication quality and smoother user experience.

[0100] Example 6

[0101] This embodiment is explained in Example 5. Please refer to Figure 1 ,Specifically: After the switching logic is executed, the self-learning optimization module collects the ,communication quality data of the path after switching through the ,network detection tool, and marks the timestamp, writes it into the ,relational database through the write port, and stores it in the order of the ,timestamp;

[0102] When using the policy gradient algorithm to calculate the gradient of the output loss function, the signal strength attenuation coefficient weight value α, noise interference intensity weight value β, building reflection coefficient weight value γ, real-time traffic fluctuation coefficient weight value δ and environmental impact function weight value ε of the path selection analysis and prediction model are updated, and the weight update value △w of the i-th weight value is obtained. i ;

[0103] Weight update amount △w i The output is calculated by the following policy gradient algorithm;

[0104]

[0105] In the formula, η represents the learning rate, which is used to control the step size of each weight update, wi represents the weight vector of the i-th weight value of the path selection analysis and prediction model, represents the partial derivative, Q(Pout∨Pcurr) represents the current path quality score Q(Pcurr) or the predicted preferred path quality score Q(Pout), and Qtarget represents the target score.

[0106] In this embodiment, after the self-learning optimization module of the system completes the execution of the path switching logic, it uses a network detection tool to collect the communication quality data after the switch in real time, and writes it into a relational database for orderly storage after marking the timestamp. Through this continuous path monitoring and data storage, the module achieves accurate tracking of the effect of communication path switching. Using the policy gradient algorithm, the module calculates and outputs the gradient of the loss function, and updates various weight parameters of the path selection analysis and prediction model, including the signal strength attenuation coefficient, noise interference intensity, building reflection coefficient, real-time traffic fluctuation coefficient, and environmental impact function weight. By adjusting the learning rate, the weight update amount Δw is calculated i , achieving adaptive adjustment and optimization of model parameters. Compared with traditional static or periodic adjustment methods, the self-learning optimization module achieves dynamic self-optimization of the path selection model through real-time feedback and reinforcement learning, enabling the model to continuously learn and adjust according to the latest network performance data. This improvement makes the system more adaptable and more accurate in predictions in complex and changing network environments. By continuously updating weight parameters, the system can quickly respond to changes in network quality, optimize path selection and switching logic, improve the stability and efficiency of the communication network, reduce manual intervention and maintenance costs, and ultimately enhance the network's intelligence level and user experience.

[0107] While the embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that many changes, modifications, substitutions and variations can be made to the embodiments without departing from the principles and spirit of the invention.

Claims

1. An IoT data analysis system based on artificial intelligence, characterized by: It includes acquisition and processing module, feature extraction module, path selection and prediction module, dynamic path adjustment and switching module and self-learning optimization module; The collection and processing module collects IoT communication data in real time by installing IoT sensor groups and related programming interfaces API on cellular towers, and transmits the IoT communication data to the central control center, where the IoT communication data is preprocessed to obtain a standardized data set Dfeat; The feature extraction module extracts the standardized data set Dfeat, performs feature extraction to obtain a reduced-dimensional feature data set F, and constructs a relational database to store the reduced-dimensional feature data set F; The path selection and prediction module obtains the environmental impact function TW by analyzing the coupling effect of signal propagation, and then uses the AI ​​algorithm to build a path selection analysis and prediction model. The environmental impact function TW is combined with the dimension reduction feature data set F, and input into the path selection analysis and prediction model, and the associated calculation is performed to output the predicted path optimization parameter Pout; The dynamic path adjustment switching module collects the communication quality data of the current path Pcurr and the predicted path optimization parameter Pout, constructs a path quality scoring formula, outputs the current path quality score Q(Pcurr) and the predicted preferred path quality score Q(Pout), sets a threshold value M, compares and evaluates the current path quality score Q(Pcurr) and the predicted preferred path quality score Q(Pout), and generates a switching logic; The self-learning optimization module is used to monitor the network performance data of the cellular tower after the path switching in real time after the switching logic is executed, and to perform reinforcement learning updates on the path selection analysis and prediction model based on the network performance data.

2. The IoT data analysis system based on artificial intelligence according to claim 1, characterized in that: The acquisition and processing module includes a data acquisition unit and a data processing unit; The data collection unit collects IoT communication data of the communication quality between cellular towers in real time by installing IoT sensors on each cellular tower in combination with relevant API programming interfaces; The IoT sensor includes a wireless signal analyzer, a spectrum scanning tool, and a traffic analysis tool; The API programming interface includes a weather forecast API programming interface and a geographic information API programming interface; The IoT communication data includes signal strength attenuation coefficient SAC, noise interference intensity NIS, building reflection coefficient BRC, real-time traffic fluctuation coefficient RTC, terrain blocking factor TOF and weather influence factor WIF.

3. The IoT data analysis system based on artificial intelligence according to claim 2, characterized in that: The data processing unit is used to transmit the collected IoT communication data to the central control center through the IoT gateway, and perform preprocessing in the central control center. The preprocessing method includes data cleaning, missing value completion and data synchronization. After the preprocessing is completed, Z-score is used for standardization to obtain a standardized data set Dfeat.

4. The IoT data analysis system based on artificial intelligence according to claim 3 is characterized in that: The feature extraction module includes a feature extraction unit and a data storage unit; The feature extraction unit extracts features with upper variance limits from the standardized data set Dfeat by principal component analysis PCA, performs dimensionality reduction processing, and then uses active window technology to perform feature decomposition on the time-varying features of the real-time traffic fluctuation coefficient RTC and the noise interference intensity NIS, extracts short-term variation trends, and then summarizes to obtain a reduced-dimensionality feature data set F; The data storage unit is used to construct a relational database and set a write port and a write port. Through the write port, the obtained dimension reduction feature data set F is arranged in chronological order and stored in the relational database.

5. The IoT data analysis system based on artificial intelligence according to claim 4 is characterized in that: The path selection and prediction module includes an influence function analysis unit and a path optimization unit; The influence function analysis unit extracts the terrain blocking factor TOF and the weather influence factor WIF in the dimension reduction feature data set F in real time through the write port of the relational database, performs coupling calculation, and outputs the environmental influence function TW; The environmental impact function TW is calculated and outputted by the following algorithm formula; Where log represents the logarithmic function, e represents the exponential function, λ represents the weather impact factor WIF adjustment parameter, and W0 represents the standard weather impact threshold.

6. The IoT data analysis system based on artificial intelligence according to claim 5, characterized in that: The path optimization unit constructs a path selection analysis and prediction model by using the LSTM network of the AI ​​algorithm to predict the path of the communication signal between the future cellular towers, and then extracts the dimension reduction feature data set F in real time through the write port of the relational database, combines the environmental impact function TW and inputs it into the path selection analysis and prediction model, performs associated calculations and outputs the predicted path optimization parameter Pout; The predicted path optimization parameter Pout is calculated and output by the following path selection analysis and prediction model; In the formula, p represents the path, argmin p It represents finding the path p that makes the entire path selection analysis and prediction model reach the lower limit value, SAC(p) represents the signal strength attenuation coefficient of the p-th path, NIS(p) represents the noise interference intensity of the p-th path, BRC(p) represents the building reflection coefficient of the p-th path, RTC(p) represents the real-time traffic fluctuation coefficient of the p-th path, TW(p) represents the environmental impact function of the p-th path, α, β, γ, δ and ε represent the preset weight values ​​of the signal strength attenuation coefficient SAC, the noise interference intensity NIS, the building reflection coefficient BRC, the real-time traffic fluctuation coefficient RTC and the environmental impact function TW, respectively, and α+β+γ+δ+ε=1, and its specific value is set by the user.

7. The IoT data analysis system based on artificial intelligence according to claim 4 is characterized by: The dynamic path adjustment switching module includes a communication quality monitoring unit, a communication quality analysis unit and a switching unit; The communication quality monitoring unit collects the communication quality data of the current path Pcurr and the predicted path optimization parameter Pout in real time through the built-in network detection tool on the cellular tower, and then transmits it to the IoT gateway, and writes it into the relational database through the write port for storage; The communication quality data includes the real-time data throughput DT(Pcurr) of the p-th path and the real-time network delay ND(Pcurr) of the p-th path, as well as the predicted data throughput DT(Pout) of the p-th path and the predicted network delay ND(Pout) of the p-th path.

8. The IoT data analysis system based on artificial intelligence according to claim 7, characterized in that: The communication quality analysis unit is used to construct a path quality scoring formula, input the collected communication quality data into the path quality scoring formula, calculate and output the current path quality score Q (Pcurr) and the predicted preferred path quality score Q (Pout), and analyze the communication quality of the current path Pcurr and the predicted path preferred parameter Pout; The current path quality score Q(Pcurr) and the predicted preferred path quality score Q(Pout) are both calculated and output by the following path quality score formula; 9. The IoT data analysis system based on artificial intelligence according to claim 7, characterized in that: The switching unit sets a threshold value M through the user, combines the threshold value with the current path quality score Q (Pcurr) and compares and evaluates the predicted preferred path quality score Q (Pout), analyzes the deviation between the current path communication quality and the preferred path communication quality, and generates a path switching logic based on the evaluation result. The specific evaluation content is as follows; When the predicted preferred path quality score Q(Pout)> the current path quality score Q(Pcurr)+threshold value M, the first switching logic is executed and the predicted path is automatically switched; When the predicted optimal path quality score Q(Pout)≤the current path quality score Q(Pcurr)+the threshold value M, the first switching logic is executed to continue to maintain the current communication path.

10. The IoT data analysis system based on artificial intelligence according to claim 9, characterized in that: The self-learning optimization module collects the communication quality data of the path after the switching logic is executed through the network detection tool, marks the timestamp, writes it into the relational database through the write port, and stores it in the order of the timestamp; When using the policy gradient algorithm to calculate the gradient of the output loss function, the signal strength attenuation coefficient weight value α, noise interference intensity weight value β, building reflection coefficient weight value γ, real-time traffic fluctuation coefficient weight value δ and environmental impact function weight value ε of the path selection analysis and prediction model are updated, and the weight update value △w of the i-th weight value is obtained. i ; The weight update amount Δw i The output is calculated by the following policy gradient algorithm; In the formula, η represents the learning rate, which is used to control the step size of each weight update, wi represents the weight vector of the i-th weight value of the path selection analysis and prediction model, represents the partial derivative, and Qtarget represents the target score.

Images