A Smart City Internet of Things System Based on Wireless Communication and Its Operating Method

Abstract

The present application discloses an Internet of Things system for a smart city based on wireless communication and an operation method, belonging to the technical field of the Internet of Things, including: the server receives first monitoring data obtained from monitoring devices; inputs the first monitoring data into a first model to output at least two classification data and two prediction values; compares the prediction values and generates a confidence level through a statistical model; feeds back the generated confidence level to the server, and the server optimizes the performance of the first model according to the received confidence level. When the present invention is implemented, by inputting the first monitoring data into the first model to output at least two classification data and two prediction values, comparing the prediction values, generating a confidence level through the statistical model, and feeding back the generated confidence level to the server, and the server optimizes the performance of the first model according to the received confidence level, it is possible to appropriately classify a large amount of data, improve the data transmission efficiency, and optimize the operation of the smart city.

Description

Technical Field

[0001] The present application relates to the technical field of the Internet of Things, and particularly to an Internet of Things system for a smart city based on wireless communication and an operation method thereof. Background Art

[0002] A smart city refers to a development model that uses information technology and Internet of Things technology to digitalize, network, and intelligentize various fields of the city, and has broad development prospects. Currently, many cities are developing towards the direction of smart cities.

[0003] During the operation of a smart city, a large amount of data support is required to complete the operation of each part of the system. These data are generally transmitted through wireless communication. Currently, in the operation system of the Internet of Things system for a smart city, there are a large number of monitoring data. These monitoring data usually need to be optimized during the transmission process to improve the transmission efficiency, such as data compression, parallel transmission of data classification, etc. Among them, when classifying and transmitting data in parallel, there is often a lack of a suitable classification standard. The traditional method of classifying according to data types will result in a large difference between the classified data groups. When performing parallel transmission, it will affect the transmission efficiency, and when the data volume is large and complex, the transmission process will also be accompanied by risks such as data loss and data damage.

[0004] Therefore, it is necessary to provide an Internet of Things system for a smart city based on wireless communication and an operation method thereof to solve the above problems.

[0005] It should be noted that the above information disclosed in this background art section is only used to understand the background art of the concept of the present application, and therefore, it may include information that does not constitute the prior art. Summary of the Invention

[0006] Based on the above problems existing in the prior art, the problem to be solved by the present application is: to provide an Internet of Things system for a smart city based on wireless communication and an operation method thereof, so as to achieve the effect of classifying data through the output of a model.

[0007] The technical solution adopted by the present application to solve its technical problems is: an operation method of an Internet of Things system for a smart city based on wireless communication, the method includes:

[0008] The server receives first monitoring data obtained from monitoring devices, and the first detection data is operation data generated during the operation of the Internet of Things system for a smart city;

[0009] Input the first monitoring data into a first model, and output at least two classified data and two predicted values;

[0010] Compare the predicted values and generate a confidence level through a statistical model;

[0011] Feed the generated confidence back to the server, and the server optimizes the performance of the first model according to the received confidence.

[0012] In the implementation process of the technical solution of the present invention, by inputting the first monitoring data into the first model, at least two classification data and two prediction values are output, and the prediction values are compared. A confidence is generated through a statistical model, and the generated confidence is fed back to the server. The server optimizes the performance of the first model according to the received confidence, which can appropriately classify a large amount of data, improve data utilization, and optimize the operation of the smart city.

[0013] Further, outputting at least two classification data includes the following steps:

[0014] Preprocess the first monitoring data to generate preprocessed data;

[0015] Select at least one classification feature, and screen the preprocessed data according to the classification feature to obtain first classification feature data;

[0016] Train the first model, and input the first classification feature data into the trained first model to reclassify the first classification feature data to generate second classification feature data;

[0017] Evaluate the first model, and dynamically adjust the first model according to the evaluation result.

[0018] Further, the classification feature includes at least one attribute, and the attribute includes location information, device type, sensor data type, time, and event type.

[0019] Further, outputting at least two prediction values includes:

[0020] Input the preprocessed data into the first model, and select statistical characteristics or time-domain frequency-domain features as prediction value classification features to output a first prediction value;

[0021] Input the first classification feature data into the first model for predictive analysis to generate a second prediction value, and the second prediction value includes data prediction values corresponding to multiple attributes in the first classification feature data;

[0022] Input the second classification feature data into the first model to generate a third prediction value, and the third prediction value includes data prediction values corresponding to multiple attributes in the second classification feature data.

[0023] Further, the generation of the confidence includes:

[0024] Compare the first predicted value with the second predicted value and the third predicted value respectively according to the same attribute to generate a first confidence level and a second confidence level, where the first confidence level is the comparison result of the first predicted value and the second predicted value, and the second confidence level is the comparison result of the first predicted value and the third predicted value;

[0025] Compare the first confidence level with the second confidence level and select one of them as the attribute confidence level under the corresponding attribute.

[0026] Furthermore, the method for determining the attribute confidence level is as follows:

[0027] When the first confidence level is equal to the second confidence level, select either the first confidence level or the second confidence level as the attribute confidence level;

[0028] When the first confidence level is less than the second confidence level, select the second confidence level as the attribute confidence level;

[0029] When the first confidence level is greater than the second confidence level, calculate the difference between the first confidence level and the second confidence level, and compare the set threshold with the result of the difference calculation. When the result of the difference calculation is less than or equal to the threshold, select the second confidence level as the attribute confidence level; when the difference result is greater than the threshold, select the first confidence level as the attribute confidence level.

[0030] Furthermore, the threshold is ten percent of the second confidence level.

[0031] A smart city Internet of Things system based on wireless communication, the system includes:

[0032] A data receiving module, used for the server to receive the first monitoring data obtained from the monitoring device, and the first detection data is the operation data generated during the operation of the smart city Internet of Things system;

[0033] A model output module, used for inputting the first monitoring data into the first model and outputting at least two classification data and two predicted values;

[0034] A confidence level generation module, used for comparing the predicted values and generating a confidence level through a statistical model;

[0035] A performance optimization module, used for feeding back the generated confidence level to the server, and the server optimizes the performance of the first model according to the received confidence level.

[0036] Furthermore, the model output module includes:

[0037] A preprocessing module, used for preprocessing the first monitoring data to generate preprocessed data;

[0038] A classification feature selection module for selecting at least one classification feature and screening the preprocessed data according to the classification feature to obtain first classification feature data;

[0039] A model training module for training the first model and inputting the first classification feature data into the trained first model to reclassify the first classification feature data and generate second classification feature data;

[0040] A model evaluation module for evaluating the first model and dynamically adjusting the first model according to the evaluation result.

[0041] Further, the confidence level generation module includes:

[0042] A predicted value comparison module for comparing the first predicted value with the second predicted value and the third predicted value respectively according to the same attribute to generate a first confidence level and a second confidence level, where the first confidence level is the comparison result of the first predicted value and the second predicted value, and the second confidence level is the comparison result of the first predicted value and the third predicted value;

[0043] A confidence level comparison module for comparing the first confidence level with the second confidence level and selecting one of them as the attribute confidence level under the corresponding attribute.

[0044] The beneficial effects of this application are: A smart city Internet of Things system and operation method based on wireless communication provided by this application, by inputting the first monitoring data into the first model, outputting at least two classification data and two predicted values, comparing the predicted values, generating a confidence level through a statistical model, and feeding back the generated confidence level to the server. The server optimizes the performance of the first model according to the received confidence level, can classify a large amount of data appropriately, improve data utilization rate, and optimize the operation of the smart city.

[0045] In addition to the purposes, features and advantages described above, this application has other purposes, features and advantages. The following will refer to the drawings for a further detailed description of this application. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] The specification drawings forming a part of this application are used to provide a further understanding of this application. The schematic embodiments of this application and their descriptions are used to explain this application and do not constitute an improper limitation to this application.

[0047] In the drawings:

[0048] Figure 1 is a schematic flow chart of an operation method of a smart city Internet of Things system based on wireless communication in this application;

[0049] Figure 2It is a schematic diagram of the module composition of an Internet of Things system for smart cities based on wireless communication in this application. Specific implementation manners

[0050] It should be noted that, without conflict, the embodiments in this application and the features in the embodiments may be combined with each other. The following will describe this application in detail with reference to the drawings and in conjunction with the embodiments.

[0051] In order to enable those skilled in the art to better understand the solution of this application, the technical solutions in the embodiments of this application will be clearly and completely described below in conjunction with the drawings in the embodiments of this application. Obviously, the described embodiments are only a part of the embodiments of this application, rather than all the embodiments. Based on the embodiments in this application, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of this application.

[0052] This application provides an operation method for an Internet of Things system for smart cities based on wireless communication. This method is generally applied to an Internet of Things system for smart cities. An Internet of Things system for smart cities is a technology that connects various sensors, devices, and items to the Internet to achieve data collection, interaction, and intelligent management among various fields of the city. The Internet of Things system for smart cities can be used in multiple fields, such as transportation, energy, environment, security, medical care, education, etc. During the operation of the Internet of Things system for smart cities, a large amount of operation data will be generated. These operation data need to be transmitted in a wired or wireless manner. To improve the transmission efficiency, currently, a wireless communication method is usually adopted. The wireless communication method includes, but is not limited to, wireless local area network, cellular mobile communication, Bluetooth, infrared communication, radio frequency communication, etc. It is not limited in this embodiment. Figure 1 The flowchart of the steps of an operation method for an Internet of Things system for smart cities based on wireless communication in this application is shown. This method includes the following steps:

[0053] Step 101: The server receives the first monitoring data obtained from the monitoring device. The first monitoring data is the operation data generated during the operation of the Internet of Things system for smart cities.

[0054] To ensure the normal operation of the Internet of Things system for smart cities, the monitoring device is an essential part. The monitoring device can collect various environmental data and the operation status data of facilities and equipment, and provide a data source for urban planning and resource allocation. In this embodiment, the monitoring device may include, but is not limited to, various sensors, image acquisition devices, positioning devices, access control systems, environmental monitoring devices, hardware monitoring devices, etc. in different scenarios. The data obtained by these monitoring devices can all be used as the first monitoring data.

[0055] It should be noted that in this embodiment, the server refers to a module with functions such as data reception, analysis and processing, temporary storage, and data output. And the server can exist in various stages of data, such as the data sending end, the data receiving end, the data transmission end, etc., rather than the server device in the conventional sense;

[0056] Step 102: Process the first monitoring data and output at least two classification data and two prediction values;

[0057] There is a first model in the server for processing the first monitoring data. It can be understood that the first model has a classification unit and a prediction unit, and both the classification unit and the prediction unit have input and output ports. Among them, the input end of the classification unit is used to receive data, and the output end of the classification unit outputs the classified data. Similarly, the input end of the prediction unit is used to receive the transmitted data and outputs through the output end;

[0058] Therefore, the first model is used to classify the data in the first monitoring data according to certain rules, and at the same time judge the future development trend of the monitoring data of the same category;

[0059] The first model can be a model for data classification and prediction. This model can be a neural network model and various algorithms, such as at least one of support vector machine (SVM), random forest (Random Forest), long short-term memory network (LSTM), convolutional neural network (CNM), etc. or a combination of several of them. Among them, the steps for outputting classification data include the following:

[0060] Step 201: Preprocess the first monitoring data to generate preprocessed data;

[0061] In order to improve the classification effect, it is necessary to preprocess the first monitoring data and generate preprocessed data. Among them, the preprocessing includes common data cleaning, denoising, missing value processing, etc., which will not be described in detail in this embodiment. The specific method can refer to the prior art or the partial description in step one of the patent with the publication number CN110543904A;

[0062] Step 203: Select at least one classification feature and screen the preprocessed data according to this classification feature to obtain the first classification feature data;

[0063] Since the first monitoring data includes various types of monitoring data, classification criteria need to be considered during classification. The traditional classification method is to divide according to the categories of the first monitoring data. For example, temperature data is classified into one category, noise data is classified into one category, and so on. The data classified in this way will have some problems. For example, there is a lack of correlation between different classification combinations, and the data volume varies greatly. It is difficult to participate in subsequent analysis and processing after classification. There is also a classification method based on the data volume. This classification method is to judge the data volume of the data and classify the data with a small data volume difference into one category. Although this method can ensure a small data volume difference between each classification combination, there will still be a problem of lack of correlation. Therefore, in this embodiment, classification features are used as the classification criteria to ensure a small data volume difference between classification combinations and, under the linkage of the classification features, increase the correlation between each classification combination to prevent low correlation between data and affect the analysis results during subsequent analysis and processing;

[0064] Specifically, the classification feature consists of at least one attribute, such as attributes like location information, device type, sensor data type, time, event type, etc. In this embodiment, the classification feature can be a combination of one or more attributes. For example, the current classification feature is location information L plus time T. Here, the location information L can be a certain range or area, and the time T is a certain time period. When screening the preprocessed data, the data in the preprocessed data that simultaneously includes the location information L and the time T will be screened out as a classification combination. The data within this classification combination represents the monitoring data obtained by the monitoring devices in the location L area during the time period T;

[0065] It should be noted that the above is just a relatively simple classification feature. In actual implementation, the classification feature can be composed of more attributes. Moreover, the more attributes included in a classification feature, the fewer the corresponding monitoring data that meet the conditions. Therefore, the selection of the classification feature can control the data quantity of a classification combination, and under the control of the classification feature composed of each attribute, the extracted monitoring data will have a certain correlation, and the data will not be isolated, facilitating subsequent data analysis and processing;

[0066] Step 205: Train the first model, and input the first classification feature data into the trained first model (hereinafter referred to as the second model) to reclassify the first classification feature data and generate second classification feature data;

[0067] During the use of the first model, model training needs to be performed according to different conditions to achieve better output results. In this embodiment, when performing model training on the first model, the unclassified first monitoring data can be used as the input parameter of the first model. Since the first monitoring data has not been classified yet, it has higher complexity;

[0068] Alternatively, the preprocessed data can be used as the input parameter of the first model to prevent the data in the first monitoring data that has not undergone data cleaning, denoising, and missing value processing from increasing the training difficulty and training time of the first model. At the same time, part of the first monitoring data and part of the preprocessed data can also be selected as the input parameters of the first model respectively to improve the training effect of the first model. For example, in this embodiment, 40% of the first monitoring data and 60% of the preprocessed data can be selected as the input parameters of the first model;

[0069] In this embodiment, different configured data can be repeatedly input into the first model to improve the training efficiency of the first model;

[0070] After the first model is trained, the first classification feature data is input into the trained first model (the second model). The second model will perform model analysis on the first classification feature data and reclassify the first classification data according to the analysis results. Because in step 203, the selection of classification features is usually manually selected or selected by fixed attributes, while the second model can automatically select more appropriate attributes as classification features and make adjustments based on the first classification feature data to generate second classification feature data;

[0071] For example, the attribute of a certain classification feature is location information L and time T. When the first classification feature data under this classification feature is input into the second model, the second model determines that the first classification feature data still lacks the data type attribute A, resulting in a lack of correlation between the first classification feature data under the data type attribute A. Then the second model will first retrieve the data types of all the data in the first classification feature data and classify the data with the same data type into one category. Since the classification feature attributes of the first classification feature data are location information L and time T;

[0072] Therefore, no matter how it is divided, the classified data will always contain both location information L and time T. At this time, the first model will input other feature classification data and select at least the classification data containing location information L and data type attribute A or time T and data type attribute A and incorporate it into the classified data to reclassify the first classification feature data and generate second classification feature data. At this time, the classification feature attributes included in the second classification feature data are location information L, time T, and data type attribute A. Through the reclassification process, it is possible to increase the classification feature attributes without reducing the data quantity and keep a certain correlation between the data;

[0073] Step 207: Evaluate the first model and dynamically adjust the first model according to the evaluation results;

[0074] After completing the training of the first model and the reclassification output, it is also necessary to evaluate the model according to certain metrics to facilitate its dynamic adjustment. Specifically, the evaluation metrics can be accuracy, recall rate, F1 value, etc. The specific model evaluation method can refer to the Chinese invention patent with the publication number CN114254264A, which will not be elaborated in this embodiment;

[0075] Using the first model to output prediction values includes:

[0076] Step 202: Input the preprocessed data into the first model, and select statistical characteristics or time-domain frequency-domain features as the prediction value classification features, and output the first prediction value;

[0077] When outputting the prediction value of the first monitoring data, preprocessing is also required first. The preprocessing process is the same as that of step 201, and in this step, it is necessary to add a process of classifying the type of the first monitoring data, and classify the data of the same type into one category to facilitate the comparison between data of the same category;

[0078] The statistical characteristics can be mean, variance, standard deviation, etc., and the time-domain frequency-domain features can be peak value, spectral density, etc. These classification features can all predict the future trend of the data and are applicable to the subsequent steps of outputting prediction values. In this embodiment, the classification features of the prediction values are not limited, as long as they can reflect the future trend of various types of data;

[0079] In this step, first input the preprocessed data into the first model. Since the preprocessed data has not been classified by features, the first prediction value generated contains the prediction values of all preprocessed data of the same type. And it can be understood that the prediction value usually contains multiple data to reflect the long-term evolution of various types of data, rather than isolated single data;

[0080] Step 204: Input the first classification feature data into the first model, perform prediction analysis, and generate the second prediction value. The second prediction value contains the data prediction values corresponding to multiple attributes in the first classification feature data;

[0081] The first classification feature data is the data after feature classification. Therefore, when inputting it into the first model, the prediction values should also be output respectively according to the corresponding classification features;

[0082] For example, the classification feature includes the location information L and the time T attributes. The first classification feature data containing these attributes is input into the first model. After selecting the predicted value classification feature, the data predicted values corresponding to the location information attribute L and the data predicted values corresponding to the time attribute T are respectively output. The predicted values corresponding to these attributes are unified as the second predicted value;

[0083] Step 206: Input the second classification feature data into the first model to generate a third predicted value. The third predicted value includes the data predicted values corresponding to multiple attributes in the second classification feature data;

[0084] The second classification feature data is the data generated after reclassification of the first classification feature data, and the data attributes included in the second classification feature data are usually more than those in the first classification feature data. For example, the second classification feature data includes three attributes: location information L, time T, and data type A. After inputting the second classification feature data into the first model, the data predicted values of location information L, time T, and data type A will be respectively output. Therefore, the predicted value generated with the second classification feature data as the input is the third predicted value;

[0085] Step 103: Compare the predicted values and generate a confidence level through a statistical model;

[0086] After generating the predicted values, it is also necessary to compare them to determine the confidence level of the output predicted values, so as to judge the accuracy and reliability of each predicted value. Specifically, the generation of the confidence level includes:

[0087] Compare the first predicted value with the second predicted value and the third predicted value respectively according to the same attribute to generate a first confidence level and a second confidence level. The first confidence level is the comparison result of the first predicted value and the second predicted value, and the second confidence level is the comparison result of the first predicted value and the third predicted value;

[0088] And the calculation of the confidence level needs to be carried out separately according to the attributes. For example, taking the location information L as an example, the data containing the location information L in the first predicted value is sequentially compared with the data containing the location information L in the second predicted value in the statistical model to generate the first confidence level; the data containing the location information L in the first predicted value is sequentially compared with the data containing the location information L in the third predicted value in the statistical model to generate the second confidence level;

[0089] The statistical model is a common method for calculating the confidence level, which can represent the confidence level of the predicted value by calculating the confidence interval or probability distribution according to the distribution of the known data. Common statistical models include the normal distribution model, t-distribution model, binomial distribution model, Poisson distribution model, etc., which are not limited in this embodiment;

[0090] It should be noted that, whether it is the first prediction value or the second prediction value, since both the first classification feature data and the second classification feature data are composed of one or more attributes, when calculating the confidence level, it is necessary to select data with the same attribute for comparison;

[0091] After the confidence level is generated, it is also necessary to compare the first confidence level with the second confidence level and select one of them as the attribute confidence level corresponding to the attribute.

[0092] In the actual application process, the first confidence level is generally not equal to the second confidence level. Therefore, it is necessary to determine the final attribute confidence level by comparison. The specific determination method is as follows:

[0093] When the first confidence level is equal to the second confidence level, select either the first confidence level or the second confidence level as the attribute confidence level;

[0094] When the first confidence level is less than the second confidence level, select the second confidence level as the attribute confidence level;

[0095] When the first confidence level is greater than the second confidence level, calculate the difference between the first confidence level and the second confidence level, and compare the set threshold with the result of the difference calculation. When the result of the difference calculation is less than or equal to the threshold, select the second confidence level as the attribute confidence level; when the result of the difference is greater than the threshold, select the first confidence level as the attribute confidence level;

[0096] In this embodiment, ten percent of the second confidence level can be used as the threshold;

[0097] In the case where the first confidence level is greater than the second confidence level, since the second confidence level is related to the second classification data, and the second classification data is obtained by reclassifying the first classification data, the correlation between the data is stronger. Therefore, the confidence level weight calculated is higher and can better represent the accuracy of the prediction value under a certain attribute. Therefore, when the gap between the first confidence level and the second confidence level is less than or equal to the set threshold, select the second confidence level as the attribute confidence level. However, when the gap between the first confidence level and the second confidence level is too large and exceeds the set threshold, it means that the second confidence level can no longer represent the accuracy of the prediction value of this attribute. Therefore, it is necessary to select the first confidence level as the attribute confidence level;

[0098] Step 104: Feed back the generated confidence level to the server, and the server optimizes the performance of the first model according to the received confidence level.

[0099] After generating the confidence level, the server can evaluate the performance of the first model based on the received confidence level, determine the reliability of the first model, and adjust and optimize the first model according to the evaluation results. In this embodiment, a model comparison method is adopted to select a model by comparing the confidence levels of different models on the same data, and the selection criterion is to preferentially select the model with a larger confidence level.

[0100] Embodiment 2:

[0101] As Figure 2 shown, the present application also proposes a smart city Internet of Things system based on wireless communication. This system runs the operation method in Embodiment 1, and this system includes:

[0102] A data receiving module, configured to receive, by the server, first monitoring data obtained from monitoring devices, where the first detection data is operation data generated during the operation of the smart city Internet of Things system;

[0103] A model output module, configured to input the first monitoring data into the first model and output at least two classification data and two predicted values;

[0104] A confidence level generation module, configured to compare the predicted values and generate a confidence level through a statistical model;

[0105] A performance optimization module, configured to feedback the generated confidence level to the server, and the server optimizes the performance of the first model according to the received confidence level;

[0106] The model output module includes:

[0107] A preprocessing module, configured to preprocess the first monitoring data to generate preprocessed data;

[0108] A classification feature selection module, configured to select at least one classification feature and screen the preprocessed data according to the classification feature to obtain first classification feature data;

[0109] A model training module, configured to train the first model and input the first classification feature data into the trained first model to reclassify the first classification feature data to generate second classification feature data;

[0110] A model evaluation module, configured to evaluate the first model and dynamically adjust the first model according to the evaluation results;

[0111] The confidence level generation module includes:

[0112] A predicted value comparison module, configured to compare a first predicted value with a second predicted value and a third predicted value respectively according to the same attribute, and generate a first confidence level and a second confidence level, where the first confidence level is the comparison result between the first predicted value and the second predicted value, and the second confidence level is the comparison result between the first predicted value and the third predicted value;

[0113] A confidence level comparison module, configured to compare the first confidence level with the second confidence level, and select one of them as the attribute confidence level under the corresponding attribute.

[0114] The above are only the preferred embodiments of the present application and are not intended to limit the present application. For those skilled in the art, various changes and modifications can be made to the present application. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application shall be included in the protection scope of the present application.

Claims

1. A running method of an Internet of Things system for a smart city based on wireless communication, characterized in that: The method includes: The server receives first monitoring data obtained from a monitoring device, where the first monitoring data is operation data generated during the operation of the smart city Internet of Things system; Preprocess the first monitoring data to generate preprocessed data; Select at least one classification feature, and filter the preprocessed data according to the classification feature to obtain first classification feature data. The classification feature includes at least one attribute, and the attribute includes location information, device type, sensor data type, time, or event type; Select 40% of the first monitoring data and 60% of the preprocessed data as input parameters for the first model, train the first model, and input the first classification feature data into the trained first model to reclassify the first classification feature data to generate second classification feature data; among them, the trained first model can automatically select the corresponding attribute as the classification feature and make adjustments based on the first classification feature data to generate second classification feature data; After completing the training and reclassification output of the first model, evaluate the first model and make dynamic adjustments to the first model according to the evaluation results; Input the preprocessed data into the first model, and select statistical characteristics or time-domain and frequency-domain characteristics as prediction value classification features to output a first prediction value. The prediction value classification feature can reflect the future trend of various types of data; the first prediction value contains the prediction values of all the same type of preprocessed data; Input the first classification feature data into the first model for predictive analysis to generate a second prediction value. The second prediction value contains the data prediction values corresponding to multiple attributes in the first classification feature data; Input the second classification feature data into the first model to generate a third prediction value. The third prediction value contains the data prediction values corresponding to multiple attributes in the second classification feature data; Compare the first prediction value with the second prediction value and the third prediction value respectively according to the same attribute to generate a first confidence level and a second confidence level. The first confidence level is the comparison result of the first prediction value and the second prediction value, and the second confidence level is the comparison result of the first prediction value and the third prediction value; Compare the first confidence level with the second confidence level and select one of them as the attribute confidence level under the corresponding attribute; feedback the generated attribute confidence level to the server, and the server optimizes the performance of the first model according to the received attribute confidence level.

2. The operating method of a smart city Internet of Things system based on wireless communication according to claim 1, characterized in that: The method for determining the attribute confidence level is: When the first confidence level is equal to the second confidence level, select either the first confidence level or the second confidence level as the attribute confidence level; When the first confidence level is less than the second confidence level, select the second confidence level as the attribute confidence level; When the first confidence level is greater than the second confidence level, calculate the difference between the first confidence level and the second confidence level, and compare the set threshold with the difference calculation result. When the difference calculation result is less than or equal to the threshold, select the second confidence level as the attribute confidence level; when the difference result is greater than the threshold, select the first confidence level as the attribute confidence level.

3. The operating method of a smart city Internet of Things system based on wireless communication according to claim 2, characterized in that: The threshold is 10% of the second confidence level.

Images