Communication protection box remote alarm system and method based on Internet of Things

Abstract

The invention discloses a communication protection box remote alarm system and method based on the Internet of Things, and relates to the technical field of the Internet of Things, and the system comprises a data collection module, a classification module, a decision tree module and an early warning module. The data acquisition module is used for deploying a sensor, acquiring sensor data and environment data, and constructing a database to store the data; the classification module is used for setting a sensor threshold value and screening the collected data points according to the matching condition of the collected data and the threshold value; and the decision tree module is used for constructing a data set and a decision tree for the screened different types of data. The invention also provides a method for realizing the system. According to the invention, the situation of early warning misinformation caused by environmental problems and the like in the prior art can be effectively improved.

Description

Technical Field

[0001] The present invention relates to the technical field of Internet of Things, and in particular to a communication protection box remote alarm system and method based on the Internet of Things. Background Art

[0002] With the rapid development of IoT technology, remote alarm systems for communication protection boxes play a key role in ensuring the stable operation of communication equipment. These systems use sensors to collect real-time data from within the protection box, such as equipment operating parameters and environmental conditions, and transmit this data to a monitoring center to promptly identify potential problems and issue alerts. This effectively prevents serious impacts on the communication network caused by equipment failures or environmental anomalies, ensuring the continuity and reliability of communications.

[0003] In existing technologies, interference factors such as strong electromagnetic fields and rapid changes in temperature and humidity often cause sensors to misjudge and miscalculate in complex and changing environments, resulting in high false alarm rates. Existing anti-interference technologies can only provide limited suppression of a single type of interference and lack comprehensive, intelligent anti-interference measures. Furthermore, insufficient research has been conducted on the impact of the interactions between different environmental factors on sensors, making it difficult to establish accurate models to distinguish between true alarms and false alarms. This not only increases the workload of operations and maintenance personnel but also reduces the credibility and practicality of alarm systems. There is an urgent need to develop more advanced, intelligent, and comprehensive technologies to overcome these shortcomings. Summary of the Invention

[0004] The purpose of the present invention is to provide a communication protection box remote alarm system and method based on the Internet of Things to solve the problems raised in the prior art.

[0005] To achieve the above-mentioned object, the present invention provides the following technical solutions: a communication protection box remote alarm system based on the Internet of Things, the system comprising a data acquisition module, a classification module, a decision tree module and an early warning module; The data acquisition module is used to deploy sensors, collect sensor data and environmental data, and build a database to store the data; the classification module is used to set sensor thresholds and filter collected data points based on the matching of collected data with the thresholds; the decision tree module is used to construct data sets and decision trees for different types of filtered data; the early warning module is used to issue early warnings for newly collected data based on the matching of its sensor data with the thresholds, and output the false alarm probability of the early warning according to the corresponding decision tree; The output end of the data acquisition module is connected to the input end of the classification module; the output end of the classification module is connected to the input end of the decision tree module; the output end of the decision tree module is connected to the input end of the early warning module; the output end of the early warning module is connected to the input end of the classification module.

[0006] The data acquisition module includes a sensor unit, an API unit and a database unit; The sensor unit is used to collect sensor data arranged in the communication protection box; the API unit is used to call the API interface of the meteorological agency to collect environmental data at the corresponding time; the database unit is used to build a database to store the data; The output end of the sensor unit is connected to the input end of the API unit; the output end of the API unit is connected to the input end of the database unit; and the output end of the database unit is connected to the input end of the classification module.

[0007] The classification module includes a threshold setting unit, a matching unit and a classification unit; The threshold setting unit is used to set the threshold for the sensor to issue an early warning; the matching unit is used to determine whether the sensor data is within the set threshold range; the classification unit is used to filter the collected data points according to the matching between the collected data and the threshold; The output end of the threshold setting unit is connected to the input end of the matching unit; the output end of the matching unit is connected to the input end of the classification unit; and the output end of the classification unit is connected to the input end of the decision tree module.

[0008] The decision tree module includes a data set unit, a decision tree construction unit and a decision tree unit; The data set unit is used to construct data sets for different types of data after screening; the decision tree construction unit is used to construct decision trees based on the constructed different data sets; the decision tree unit is used to store the obtained unwarned decision trees, single-sensor decision trees and multi-sensor decision trees; The output end of the data set unit is connected to the input end of the decision tree construction unit; the output end of the decision tree construction unit is connected to the input end of the decision tree unit; the output end of the decision tree unit is connected to the input end of the early warning module.

[0009] The early warning module includes a data acquisition unit and an early warning unit; The data acquisition unit is used to collect new data; the early warning unit is used to issue an early warning based on the matching of the newly collected sensor data with the threshold, and output the false alarm probability of the early warning according to the corresponding decision tree; The output end of the data acquisition unit is connected to the input end of the classification module.

[0010] A remote alarm method for a communication protection box based on the Internet of Things, the method comprising the following steps: Step 1: Deploy sensors, collect sensor data and environmental data, and build a database to store the data; Step 2: Set the sensor threshold and filter the collected data points based on whether the collected data matches the threshold. Step 3: Build data sets and decision trees for different types of filtered data. Step 4: For newly collected data, an early warning is issued based on the matching of the sensor data with the threshold, and the false alarm probability of the early warning is output according to the corresponding decision tree.

[0011] In step 1, the sensor data arranged in the communication protection box is collected, and the API interface of the meteorological agency is called to collect the environmental data at the corresponding time; the sensor data is represented as: [C1, C2, ..., C n ]; Environmental data is represented as: [E1, E2,…, E m ]; where n is a positive integer, indicating the type of sensor data, C1 to C n Represents the 1st to nth types of sensor data respectively; m is a positive integer, representing the type of environmental data, E1~E m Respectively represent the 1st to mth environmental data; Construct a database in the form of: [C1,C2,…,C n ,E1,E2,…,E m ,S0 / S1,W0 / W1]; where S0 means no warning was issued, S1 means a warning was issued, W0 means no false alarm, and W1 means a false alarm; S0, S1, W0 and W1 are manually labeled according to the warning and subsequent processing.

[0012] In step 2, the preset threshold intervals of different sensor data are expressed as [l Ci ,L Ci ]; where i is a positive integer, representing the sensor sequence, C i represents the i-th sensor; l Ci Indicates sensor C i The preset lower threshold, L Ci Indicates sensor C i The preset upper threshold value of When all sensor data are within the preset threshold range, the alarm system does not issue an early warning; when there is sensor data that exceeds the preset threshold range, the alarm system issues an early warning; Filter all sensor data points that do not exceed the preset threshold and construct the data set D; When there is sensor data exceeding a preset threshold, all data points with sensor data exceeding the preset threshold are filtered out; When there is more than one type of sensor data exceeding the preset threshold, the type of sensor exceeding the preset threshold is determined; and samples whose determined sensor data exceeds the preset threshold are filtered.

[0013] In step 3, for all data points whose sensor data do not exceed the preset threshold, a dataset D is constructed: {[V1(C1),V1(C2),…,V1(C n ),V1(E1),V1(E2),…,V1(E m ),W0 / W1],…,[V a (C1),V a (C2),…,V a (C n ),V a (E1),V a (E2),…,V a (E m ),W0 / W1]}; where a is a positive integer, indicating the number of data points where all sensor data do not exceed the threshold; V a (C1)~V a (C n ) represents the ath data point sensor C1~C n Corresponding data; The false alarm conditions W0 and W1 are numerically expressed, where 0 indicates no false alarm and 1 indicates a false alarm; Construct an unwarned decision tree: Calculate the entropy of the data set entropy(D) = -p0log2p0-p1log2p1; where p0 represents the proportion of non-false positive data points in the data set D to the total data points, and p1 represents the proportion of false positive data points in the data set D to the total data points; Calculate the conditional entropy of the feature, for feature F u , conditional entropy ; Among them, feature F u ∈{C1,C2,…,C n ,E1,E2,…,E m}; V(F u ) represents feature F u Value set; D v Represents feature F in dataset D u The value is a subset of v; Calculate information gain: Gain(D,F u )=entropy(D)-entropy(D|F u ); Set the minimum number of samples B; calculate the information gain of all features relative to the data set D, and select the feature with the largest information gain as the basis for dividing the root node; repeat the step of calculating the information gain for each subset divided according to the different values ​​of the selected feature, and select the feature with the largest information gain in each subset to continue dividing, generating child nodes, and continue recursively until the number of samples in the node is less than the preset minimum number of samples B; no further division is continued, and the node is regarded as a leaf node. The category of the leaf node is determined according to the category to which the majority of samples in the node belong; thus completing the construction of the unwarned decision tree; For all data points where sensor data exceeds the preset threshold, construct a data set: {[V1'(C1),V1'(C2),…,V1'(C n ),V1'(E1),V1'(E2),…,V1'(E m ),W0 / W1],…,[V b '(C1),V b '(C2),…,V b '(C n ),V b '(E1),V b '(E2),…,V b '(E m ),W0 / W1]}; where b is a positive integer, representing the number of data points where all sensor data exceeds the preset threshold; V b '(C1)~V b '(C n ) represents the bth data point sensor C1~C n Corresponding data; For sensor data C e and environmental data E f , calculate the correlation coefficient r ef ; Calculate sensor data C e Mean value of V mean (C e )=(V1'(C e )+…+V b '(C e )) / b; Calculate environmental data E f Mean value of V mean (E f )=(V1'(E f )+…+V b '(E f )) / b; Correlation coefficient: ; Where e is a positive integer, representing the sensor sequence, e∈{1,2,…,n}, C e represents the e-th sensor; The correlation coefficients between all sensor data and environmental data are combined into a matrix R; a correlation threshold θ is set. When the absolute value of the correlation coefficient between a certain sensor data and a certain environmental data is greater than θ, the sensor data and the environmental data are highly correlated. When the sensor data C e When the preset threshold is exceeded, the matrix R is extracted and the sensor data C e Environmental data with an absolute value of correlation coefficient greater than θ; sensor data C e And the extracted environmental data builds a data set, and then completes the sensor data C e Construction of decision tree: for each sensor, a corresponding single sensor decision tree is constructed; When there are more than one types of sensor data exceeding the preset threshold, the types of sensors exceeding the preset threshold are determined; samples whose sensor data have been determined to exceed the preset threshold are screened, the environmental data are determined according to the correlation matrix R, and the data set and the corresponding multi-sensor decision tree are constructed.

[0014] In step 4, new sensor data is collected and the corresponding environmental data is collected, which is expressed as X=[C1(new),C2(new),…,C n (new),E1(new),E2(new),…,E m (new)], judge the matching situation of the newly collected data with the preset threshold according to the preset threshold, and select the non-warning decision tree, single-sensor decision tree or multi-sensor decision tree according to the matching situation; re-combine the features of the newly collected data according to the selected decision tree, and start from the root node of the selected decision tree, according to the value of each feature in the feature vector, traverse the decision tree downward along the corresponding branch in sequence until reaching a leaf node; the leaf node records the distribution of false alarm and non-false alarm samples in the historical data, q0 represents the number of non-false alarm samples in the leaf node, q1 represents the number of false alarm samples, and the probability of false alarm of the newly collected data is P=q1 / (q0+q1); An early warning is issued based on the match between the newly collected data and the preset threshold, and the false alarm probability of the corresponding early warning is output.

[0015] Compared with the existing technology, the beneficial effects of the present invention are: the present invention collects environmental data by calling the meteorological agency API interface, constructs a database containing multiple data types, can more comprehensively reflect the actual status of the protective box and the surrounding environment, and provides a rich data basis for accurate alarm; the present invention constructs a single-sensor decision tree and a multi-sensor decision tree according to the situation where the sensor data exceeds the threshold, and conducts targeted analysis on different sensor data combinations and environmental data associations, so that the alarm decision is more in line with the actual scenario, thereby improving the accuracy and effectiveness of the alarm. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1 This is a flow chart of a communication protection box remote alarm system based on the Internet of Things of the present invention; Figure 2 This is a schematic diagram of the steps of a remote alarm method for a communication protection box based on the Internet of Things of the present invention. DETAILED DESCRIPTION

[0017] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. 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 making creative efforts are within the scope of protection of the present invention.

[0018] Example: Figure 1-Figure 2 As shown, the present invention provides a technical solution, a communication protection box remote alarm system based on the Internet of Things, the system includes a data acquisition module, a classification module, a decision tree module and an early warning module; The data acquisition module is used to deploy sensors, collect sensor data and environmental data, and build a database to store the data; the classification module is used to set sensor thresholds and filter collected data points based on the matching of collected data with the thresholds; the decision tree module is used to construct data sets and decision trees for different types of filtered data; the early warning module is used to issue early warnings for newly collected data based on the matching of its sensor data with the thresholds, and output the false alarm probability of the early warning according to the corresponding decision tree; The output end of the data acquisition module is connected to the input end of the classification module; the output end of the classification module is connected to the input end of the decision tree module; the output end of the decision tree module is connected to the input end of the early warning module; the output end of the early warning module is connected to the input end of the classification module.

[0019] The data acquisition module includes a sensor unit, an API unit and a database unit; The sensor unit is used to collect sensor data arranged in the communication protection box; the API unit is used to call the API interface of the meteorological agency to collect environmental data at the corresponding time; the database unit is used to build a database to store the data; The output end of the sensor unit is connected to the input end of the API unit; the output end of the API unit is connected to the input end of the database unit; and the output end of the database unit is connected to the input end of the classification module.

[0020] The classification module includes a threshold setting unit, a matching unit and a classification unit; The threshold setting unit is used to set the threshold for the sensor to issue an early warning; the matching unit is used to determine whether the sensor data is within the set threshold range; the classification unit is used to filter the collected data points according to the matching between the collected data and the threshold; The output end of the threshold setting unit is connected to the input end of the matching unit; the output end of the matching unit is connected to the input end of the classification unit; and the output end of the classification unit is connected to the input end of the decision tree module.

[0021] The decision tree module includes a data set unit, a decision tree construction unit and a decision tree unit; The data set unit is used to construct data sets for different types of data after screening; the decision tree construction unit is used to construct decision trees based on the constructed different data sets; the decision tree unit is used to store the obtained unwarned decision trees, single-sensor decision trees and multi-sensor decision trees; The output end of the data set unit is connected to the input end of the decision tree construction unit; the output end of the decision tree construction unit is connected to the input end of the decision tree unit; the output end of the decision tree unit is connected to the input end of the early warning module.

[0022] The early warning module includes a data acquisition unit and an early warning unit; The data acquisition unit is used to collect new data; the early warning unit is used to issue an early warning based on the matching of the newly collected sensor data with the threshold, and output the false alarm probability of the early warning according to the corresponding decision tree; The output end of the data acquisition unit is connected to the input end of the classification module.

[0023] A remote alarm method for a communication protection box based on the Internet of Things, the method comprising the following steps: Step 1: Deploy sensors, collect sensor data and environmental data, and build a database to store the data; Step 2: Set the sensor threshold and filter the collected data points based on whether the collected data matches the threshold. Step 3: Build data sets and decision trees for different types of filtered data. Step 4: For newly collected data, an early warning is issued based on the matching of the sensor data with the threshold, and the false alarm probability of the early warning is output according to the corresponding decision tree.

[0024] In step 1, the sensor data arranged in the communication protection box is collected, and the API interface of the meteorological agency is called to collect the environmental data at the corresponding time; the sensor data is represented as: [C1, C2, ..., C n ]; Environmental data is represented as: [E1, E2,…, E m ]; where n is a positive integer, indicating the type of sensor data, C1 to C n Represents the 1st to nth types of sensor data respectively; m is a positive integer, representing the type of environmental data, E1~E m Respectively represent the 1st to mth environmental data; Construct a database in the form of: [C1,C2,…,C n ,E1,E2,…,E m ,S0 / S1,W0 / W1]; where S0 means no warning was issued, S1 means a warning was issued, W0 means no false alarm, and W1 means a false alarm; S0, S1, W0 and W1 are manually labeled according to the warning and subsequent processing.

[0025] In step 2, the preset threshold intervals of different sensor data are expressed as [l Ci ,L Ci ]; where i is a positive integer, representing the sensor sequence, C i represents the i-th sensor; l Ci Indicates sensor C i The preset lower threshold, L Ci Indicates sensor C i The preset upper threshold value of When all sensor data are within the preset threshold range, the alarm system does not issue an early warning; when there is sensor data that exceeds the preset threshold range, the alarm system issues an early warning; Filter all sensor data points that do not exceed the preset threshold and construct the data set D; When there is sensor data exceeding a preset threshold, all data points with sensor data exceeding the preset threshold are filtered out; When there is more than one type of sensor data exceeding the preset threshold, the type of sensor exceeding the preset threshold is determined; and samples whose determined sensor data exceeds the preset threshold are filtered.

[0026] In step 3, for all data points whose sensor data do not exceed the preset threshold, a dataset D is constructed: {[V1(C1),V1(C2),…,V1(C n ),V1(E1),V1(E2),…,V1(E m ),W0 / W1],…,[V a (C1),V a (C2),…,Va (C n ),V a (E1),V a (E2),…,V a (E m ),W0 / W1]}; where a is a positive integer, indicating the number of data points where all sensor data do not exceed the threshold; V a (C1)~V a (C n ) represents the ath data point sensor C1~C n Corresponding data; The false alarm conditions W0 and W1 are numerically expressed, where 0 indicates no false alarm and 1 indicates a false alarm; Construct an unwarned decision tree: Calculate the entropy of the data set entropy(D) = -p0log2p0-p1log2p1; where p0 represents the proportion of non-false positive data points in the data set D to the total data points, and p1 represents the proportion of false positive data points in the data set D to the total data points; Calculate the conditional entropy of the feature, for feature F u , conditional entropy ; Among them, feature F u ∈{C1,C2,…,C n ,E1,E2,…,E m}; V(F u ) represents feature F u Value set; D v Represents feature F in dataset D u The value is a subset of v; Calculate information gain: Gain(D,F u )=entropy(D)-entropy(D|F u ); Set the minimum number of samples B; calculate the information gain of all features relative to the data set D, and select the feature with the largest information gain as the basis for dividing the root node; repeat the step of calculating the information gain for each subset divided according to the different values ​​of the selected feature, and select the feature with the largest information gain in each subset to continue dividing, generating child nodes, and continue recursively until the number of samples in the node is less than the preset minimum number of samples B; no further division is continued, and the node is regarded as a leaf node. The category of the leaf node is determined according to the category to which the majority of samples in the node belong; thus completing the construction of the unwarned decision tree; For all data points where sensor data exceeds the preset threshold, construct a data set: {[V1'(C1),V1'(C2),…,V1'(C n ),V1'(E1),V1'(E2),…,V1'(E m ),W0 / W1],…,[Vb '(C1),V b '(C2),…,V b '(C n ),V b '(E1),V b '(E2),…,V b '(E m ),W0 / W1]}; where b is a positive integer, representing the number of data points where all sensor data exceeds the preset threshold; V b '(C1)~V b '(C n ) represents the bth data point sensor C1~C n Corresponding data; For sensor data C e and environmental data E f , calculate the correlation coefficient r ef ; Calculate sensor data C e Mean value of V mean (C e )=(V1'(C e )+…+V b '(C e )) / b; Calculate environmental data E f Mean value of V mean (E f )=(V1'(E f )+…+V b '(E f )) / b; Correlation coefficient: ; Where e is a positive integer, representing the sensor sequence, e∈{1,2,…,n}, C e represents the e-th sensor; The correlation coefficients between all sensor data and environmental data are combined into a matrix R; a correlation threshold θ is set. When the absolute value of the correlation coefficient between a certain sensor data and a certain environmental data is greater than θ, the sensor data and the environmental data are highly correlated. When the sensor data C e When the preset threshold is exceeded, the matrix R is extracted and the sensor data C e Environmental data with an absolute value of correlation coefficient greater than θ; sensor data C e And the extracted environmental data builds a data set, and then completes the sensor data C e Construction of decision tree: for each sensor, a corresponding single sensor decision tree is constructed; When there are more than one types of sensor data exceeding the preset threshold, the types of sensors exceeding the preset threshold are determined; samples whose sensor data have been determined to exceed the preset threshold are screened, the environmental data are determined according to the correlation matrix R, and the data set and the corresponding multi-sensor decision tree are constructed.

[0027] In step 4, new sensor data is collected and the corresponding environmental data is collected, which is expressed as X=[C1(new),C2(new),…,C n (new),E1(new),E2(new),…,E m (new)], judge the matching situation of the newly collected data with the preset threshold according to the preset threshold, and select the non-warning decision tree, single-sensor decision tree or multi-sensor decision tree according to the matching situation; re-combine the features of the newly collected data according to the selected decision tree, and start from the root node of the selected decision tree, according to the value of each feature in the feature vector, traverse the decision tree downward along the corresponding branch in sequence until reaching a leaf node; the leaf node records the distribution of false alarm and non-false alarm samples in the historical data, q0 represents the number of non-false alarm samples in the leaf node, q1 represents the number of false alarm samples, and the probability of false alarm of the newly collected data is P=q1 / (q0+q1); An early warning is issued based on the match between the newly collected data and the preset threshold, and the false alarm probability of the corresponding early warning is output.

[0028] In this embodiment, a communication protection box of an important communication base station in a certain city is equipped with a variety of sensors, including temperature sensors, humidity sensors, smoke sensors, and electromagnetic intensity sensors. These sensors are used to monitor the environmental conditions and equipment operating status inside the box in real time. The box is also connected to a remote monitoring center via a network to implement remote alarm functions and ensure the stable operation of the communication equipment. Various sensors are placed at key locations inside the communication protection box. These sensors collect data every 5 minutes. The sensor data is represented as [temperature, humidity, smoke concentration, electromagnetic intensity], namely [C1, C2, C3, C4]. At the same time, by calling the API interface of the local meteorological agency, the environmental data of the protection box location is collected every hour, including atmospheric pressure, wind speed, precipitation, etc. The environmental data is represented as [atmospheric pressure, wind speed, precipitation], namely [E1, E2, E3]. Storing the collected sensor data and environmental data in a database; Preset threshold ranges for different sensor data: the temperature sensor threshold is [10°C, 40°C], the humidity sensor threshold is [30%, 60%], the smoke concentration threshold is [0ppm, 5ppm], and the electromagnetic intensity threshold is [5μT, 30μT]. When all sensor data are within the preset threshold range, the alarm system does not issue an alarm, and the data point is filtered out to build data set D; when there is sensor data that exceeds the preset threshold range, the alarm system issues an alarm and filters out all data points with sensor data that exceeds the preset threshold for further analysis; Construct corresponding unwarned decision trees, single-sensor decision trees and multi-sensor decision trees; Collect new sensor data and corresponding environmental data. The newly collected data is [35°C, 50%, 0ppm, 25μT, 1005hPa, 4m / s, 0mm]. The matching between the newly collected data and the preset threshold is determined based on the preset threshold. If the electromagnetic intensity exceeds the lower threshold, the decision tree corresponding to the electromagnetic intensity is selected based on the correlation matrix. According to the selected decision tree, starting from the root node of the decision tree, according to the value of each feature in the feature vector, traverse the decision tree down along the corresponding branch in sequence until a leaf node is reached; the leaf node reached records that there are 30 false positive samples in the historical data (q1) and 70 non-false positive samples (q0). Therefore, the probability of false positive in the newly collected data is P=30 / (30+70)=0.3; The monitoring center issues an early warning based on the match between the newly collected data and the preset threshold, and simultaneously outputs a false alarm probability of 0.3 for the corresponding early warning, so that operation and maintenance personnel can better judge the authenticity and urgency of the alarm and take corresponding measures in a timely manner.

[0029] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above and that the invention can be embodied in other specific forms without departing from the spirit or essential characteristics of the invention. Therefore, the embodiments should be considered in all respects as illustrative and non-restrictive, and the scope of the invention is defined by the appended claims, not the foregoing description, and all variations within the meaning and range of equivalents of the claims are intended to be included therein. Any reference sign in a claim should not be construed as limiting the claim to which it relates.

Claims

1. A communication protection box remote alarm system based on the Internet of Things, characterized by: The system includes a data acquisition module, a classification module, a decision tree module and an early warning module; The data acquisition module is used to deploy sensors, collect sensor data and environmental data, and build a database to store the data; The classification module is used to set the sensor threshold and filter the collected data points according to the matching between the collected data and the threshold; The decision tree module is used to construct data sets and decision trees for different types of data after screening; The warning module is used to issue a warning for the newly collected data according to the matching of its sensor data with the threshold, and output the false alarm probability of the warning situation according to the corresponding decision tree; The output end of the data acquisition module is connected to the input end of the classification module; The output end of the classification module is connected to the input end of the decision tree module; The output end of the decision tree module is connected to the input end of the early warning module; The output end of the early warning module is connected to the input end of the classification module.

2. According to the Internet of Things-based communication protection box remote alarm system of claim 1, it is characterized by: The data acquisition module includes a sensor unit, an API unit and a database unit; The sensor unit is used to collect sensor data arranged in the communication protection box; The API unit is used to call the API interface of the meteorological agency to collect environmental data at the corresponding time; The database unit is used to construct a database to store data; The output end of the sensor unit is connected to the input end of the API unit; The output end of the API unit is connected to the input end of the database unit; The output end of the database unit is connected to the input end of the classification module.

3. According to the Internet of Things-based communication protection box remote alarm system of claim 2, it is characterized by: The classification module includes a threshold setting unit, a matching unit and a classification unit; The threshold setting unit is used to set the threshold for the sensor to issue an early warning; The matching unit is used to determine whether the sensor data is within a set threshold range; The classification unit is used to screen the collected data points according to the matching of the collected data with the threshold; The output end of the threshold setting unit is connected to the input end of the matching unit; The output end of the matching unit is connected to the input end of the classification unit; The output end of the classification unit is connected to the input end of the decision tree module.

4. According to the Internet of Things-based communication protection box remote alarm system of claim 3, it is characterized by: The decision tree module includes a data set unit, a decision tree construction unit and a decision tree unit; The data set unit is used to construct data sets for different types of data after screening; The decision tree construction unit is used to construct a decision tree according to the constructed different data sets; The decision tree unit is used to store the obtained unwarned decision tree, single sensor decision tree and multi-sensor decision tree; The output end of the data set unit is connected to the input end of the decision tree construction unit; The output end of the decision tree construction unit is connected to the input end of the decision tree unit; The output end of the decision tree unit is connected to the input end of the early warning module.

5. According to the Internet of Things-based communication protection box remote alarm system of claim 4, it is characterized by: The early warning module includes a data acquisition unit and an early warning unit; The data acquisition unit is used to collect new data; The warning unit is used to issue a warning according to the matching of the newly collected sensor data with the threshold value, and output the false alarm probability of the warning situation according to the corresponding decision tree; The output end of the data acquisition unit is connected to the input end of the classification module.

6. A remote alarm method for a communication protection box based on the Internet of Things, characterized in that: The method comprises the following steps: Step 1: Deploy sensors, collect sensor data and environmental data, and build a database to store the data; Step 2: Set the sensor threshold and filter the collected data points according to the matching between the collected data and the threshold; Step 3: Build data sets and decision trees for different types of data after screening; Step 4: For the newly collected data, an early warning is issued based on the matching of the sensor data with the threshold, and the false alarm probability of the early warning is output according to the corresponding decision tree.

7. The remote alarm method for a communication protection box based on the Internet of Things according to claim 6 is characterized in that: In step 1, the sensor data arranged in the communication protection box is collected, and the API interface of the meteorological agency is called to collect the environmental data at the corresponding time; the sensor data is represented as: [C1, C2, …, C n ]; environmental data is represented as: [E1, E2, …, E m ]; where n is a positive integer, indicating the type of sensor data, C1~C n They represent the 1st to nth types of sensor data respectively; m is a positive integer, indicating the type of environmental data, E1~E m Respectively represent the 1st to mth environmental data; Construct a database in the form of: [C1,C2,…,C n ,E1,E2,…,E m ,S0 / S1,W0 / W1]; where S0 means no warning was issued, S1 means a warning was issued, W0 means no false alarm, and W1 means a false alarm; S0, S1, W0 and W1 are manually labeled according to the warning and subsequent processing.

8. The remote alarm method for a communication protection box based on the Internet of Things according to claim 7 is characterized in that: In step 2, the preset threshold intervals of different sensor data are expressed as [l Ci ,L Ci ]; where i is a positive integer, representing the sensor sequence, C i represents the i-th sensor; l Ci Indicates sensor C i The preset lower threshold, L Ci Indicates sensor C i The preset upper threshold of When all sensor data are within the preset threshold range, the alarm system does not issue an early warning; when there is sensor data that exceeds the preset threshold range, the alarm system issues an early warning; Filter all data points whose sensor data do not exceed the preset threshold and construct data set D; When there is sensor data exceeding a preset threshold, all data points where there is sensor data exceeding the preset threshold are filtered out; When there are more than one type of sensor data exceeding the preset threshold, the type of sensor exceeding the preset threshold is determined; and samples whose determined sensor data exceeds the preset threshold are screened.

9. The remote alarm method for a communication protection box based on the Internet of Things according to claim 8 is characterized in that: In step 3, for all data points whose sensor data do not exceed the preset threshold, a dataset D is constructed: {[V1(C1),V1(C2),…,V1(C n ),V1(E1),V1(E2),…,V1(E m ),W0 / W1],…,[V a (C1),V a (C2),…,V a (C n ),V a (E1),V a (E2),…,V a (E m ),W0 / W1]}; where a is a positive integer, indicating the number of data points where all sensor data do not exceed the threshold; V a (C1)~V a (C n ) represents the ath data point sensor C1~C n Corresponding data; The false alarm situations W0 and W1 are numerically expressed, 0 means no false alarm, and 1 means false alarm; Construct an unwarned decision tree: Calculate the entropy of the data set entropy(D) = -p0log2p0-p1log2p1; where p0 represents the proportion of non-false positive data points in the data set D to the total data points, and p1 represents the proportion of false positive data points in the data set D to the total data points; Calculate the conditional entropy of the feature, for feature F u , conditional entropy ; Among them, feature F u ∈{C1,C2,…,C n ,E1,E2,…,E m }; V(F u ) represents the feature F u Value set; D v Represents feature F in dataset D u The value is a subset of v; Calculate information gain: Gain(D,F u )=entropy(D)-entropy(D|F u ); Set the minimum number of samples B; calculate the information gain of all features relative to the data set D, and select the feature with the largest information gain as the basis for dividing the root node; repeat the steps of calculating the information gain for each subset divided according to different values ​​of the selected feature, and select the feature with the largest information gain in each subset to continue dividing, generate child nodes, and continue to recurse until the number of samples in the node is less than the preset minimum number of samples B; no longer divide, the node is used as a leaf node, and the category of the leaf node is determined according to the category to which the majority of samples in the node belong; complete the construction of the unwarned decision tree; For all data points where sensor data exceeds the preset threshold, construct a data set: {[V1'(C1),V1'(C2),…,V1'(C n ),V1'(E1),V1'(E2),…,V1'(E m ),W0 / W1],…,[V b '(C1),V b '(C2),…,V b '(C n ),V b '(E1),V b '(E2),…,V b '(E m ),W0 / W1]}; where b is a positive integer, indicating the number of data points where all sensor data exceeds the preset threshold; V b '(C1)~V b '(C n ) represents the bth data point sensor C1~C n Corresponding data; For sensor data C e and environmental data E f , calculate the correlation coefficient r ef ; Calculate sensor data C e Mean value: V mean (C e )=(V1'(C e )+…+V b '(C e )) / b; Calculate environmental data E f Mean value: V mean (E f )=(V1'(E f )+…+V b '(E f )) / b; Correlation coefficient: ; Where e is a positive integer, representing the sensor sequence, e∈{1,2,…,n}, C e represents the e-th sensor; The correlation coefficients between all sensor data and environmental data are combined into a matrix R; a correlation threshold θ is set. When the absolute value of the correlation coefficient between a certain sensor data and a certain environmental data is greater than θ, the sensor data and the environmental data are highly correlated. When the sensor data C e When the preset threshold is exceeded, the matrix R is used to extract the sensor data C e Environmental data with an absolute value of correlation coefficient greater than θ; sensor data C e And the extracted environmental data builds a data set, and then completes the sensor data C e Construction of decision tree: for each sensor, a corresponding single sensor decision tree is constructed; When there are more than one type of sensor data exceeding the preset threshold, determine the type of sensor that exceeds the preset threshold; filter the samples whose sensor data exceeds the preset threshold, determine the environmental data according to the correlation matrix R, and construct the data set and the corresponding multi-sensor decision tree.

10. The remote alarm method for a communication protection box based on the Internet of Things according to claim 9 is characterized in that: In step 4, new sensor data is collected and the corresponding environmental data is collected, which is expressed as X=[C1(new),C2(new),…,C n (new),E1(new),E2(new),…,E m (new)], according to the preset threshold, determine the match between the newly collected data and the preset threshold, and select the unwarned decision tree, single sensor decision tree or multi-sensor decision tree according to the match; re-combine the features of the newly collected data according to the selected decision tree, starting from the root node of the selected decision tree, according to the value of each feature in the feature vector, traverse the decision tree downward along the corresponding branch in sequence until reaching a leaf node; the leaf node records the distribution of false alarm and non-false alarm samples in the historical data, q0 represents the number of non-false alarm samples in the leaf node, q1 represents the number of false alarm samples, then the probability of false alarm of the newly collected data P=q1 / (q0+q1); An early warning is issued based on the match between the newly collected data and the preset threshold, and the false alarm probability of the corresponding early warning is output.

Images