Building information processing method based on Internet of Things

Abstract

The invention discloses a building information processing method based on the Internet of Things, and relates to the field of data integration. According to the building information processing method based on the Internet of Things, through a professional hybrid algorithm of a convolutional neural network (CNN) and a recurrent neural network (RNN) adopted in the data preprocessing step, the effects of the CNN and the RNN in feature extraction can be balanced, so that generation of abnormal data is better reduced, the data accuracy is improved, and the method is suitable for popularization and application. Meanwhile, the training and reasoning speed of the model can be remarkably improved on the premise that the feature extraction capacity is not affected by adopting depth separable convolution, the long-term dependency relationships can be better captured by adopting an RNN structure improved by LSTM or GRU, and the building information processing capacity and data accuracy of the model are improved.

Description

Technical Field

[0001] The present invention relates to data integration technology, and in particular to a building information processing method based on the Internet of Things. Background Art

[0002] Building structure monitoring mainly includes monitoring of the construction process and operation stage of building superstructures such as reinforced concrete structures, wooden structures and steel structures. Building structure monitoring is an important means to understand the current structural health status, and can also be used as a reference for building structure identification and safety and reliability assessment. In recent years, large-scale building collapse accidents such as bridges, houses, and tunnels have occurred frequently, causing significant economic losses to the country. However, traditional building structure strain monitoring technology has poor accuracy, large errors, and cumbersome wiring, which has great limitations in application.

[0003] The Internet of Things (IoT) technology mainly refers to a network that allows all ordinary objects that can perform independent functions to be interconnected, which can optimize the monitoring process of building structures. Therefore, the wireless monitoring system of building structure strain based on the IoT has gradually been developed and applied. The current wireless monitoring system of building structure strain based on the IoT is mainly composed of a front-end acquisition module, a wireless transmission module and a background management module. The front-end acquisition module completes strain data acquisition by setting sensor nodes. In the actual acquisition process, the transmission of stress detection signals will be interfered by external factors, which will cause errors in the measurement data, thereby affecting the monitoring accuracy of the entire building data. Summary of the invention

[0004] The purpose of the present invention is to provide a building information processing method based on the Internet of Things to solve the problem in the prior art that due to the different expression habits of each financial system user, the language of each financial system user needs to be modified accordingly when integrating financial data, which will increase the workload of financial data integration.

[0005] In order to achieve the above object, the present invention provides the following technical solution: a building information processing method based on the Internet of Things, comprising the following steps:

[0006] Build an IoT sensor network that covers the entire building, collects various physical parameters and equipment operating status information in real time, and transmits it to the data processing center;

[0007] In the data processing center, the collected building information data is preprocessed, including data cleaning and data normalization operations, to remove noise and abnormal data in preparation for subsequent analysis;

[0008] The pre-processed data is analyzed using a professional hybrid algorithm of convolutional neural network (CNN) and recurrent neural network (RNN). Specifically, the spatial features in the building information data are first extracted through CNN to obtain a preliminary feature representation, and then these features are input into RNN. RNN is used to process the time series features to mine the potential patterns and trends in the building, effectively reducing the generation of abnormal data and significantly improving data accuracy.

[0009] Data compression technology is used to compress the building information data after CNN and RNN processing. Specifically, the principal component analysis (PCA) algorithm is used to reduce the dimensionality of the data, project the high-dimensional data into a low-dimensional space, retain the main feature information, and remove redundant and noisy data, thereby further improving the storage and transmission efficiency of the data without affecting the analysis results of the data.

[0010] The model structure of the compressed data is optimized, specifically: the depthwise separable convolution (DSC) structure is introduced to decompose the traditional convolution operation into two steps: depthwise convolution and pointwise convolution. The depthwise convolution is used to extract the spatial features of the data, and the pointwise convolution is used to perform linear transformation in the channel dimension. This can reduce the amount of calculation and the number of parameters while maintaining good feature extraction capabilities. At the same time, the RNN structure is optimized, and the RNN structure improved by the long short-term memory network (LSTM) or the gated recurrent unit (GRU) is used to better handle long-term dependencies and avoid the gradient vanishing problem, thereby improving the performance of the entire hybrid algorithm and data accuracy.

[0011] Based on the analysis results, an intelligent model of the building is established. The model can accurately predict the future building status based on the current building information data, including temperature change trends, equipment failure probability, energy consumption requirements, and provide corresponding early warnings and optimization suggestions;

[0012] The prediction results and optimization suggestions of the intelligent model are sent to various devices and systems in the building through the Internet of Things platform to achieve intelligent control and optimization management of the building, such as automatically adjusting the air conditioning temperature and optimizing the equipment operation strategy;

[0013] Regularly evaluate and optimize the building information processing system, adjust and improve data compression technology and model structure optimization according to actual operation conditions, so as to continuously improve the accuracy and adaptability of the system;

[0014] Provide a user interface that enables users to easily view the operating status, analysis results and optimization suggestions of the building information processing system, and perform corresponding operations and management, including setting early warning thresholds and adjusting optimization strategies.

[0015] Furthermore, when constructing the IoT sensor network, a variety of high-precision sensors are used, including temperature sensors, humidity sensors, light sensors, and air pressure sensors, to ensure the accuracy and comprehensiveness of the collected building information data.

[0016] Furthermore, for the professional hybrid algorithm of convolutional neural network (CNN) and recurrent neural network (RNN) used in the data preprocessing step, the specific calculation formula is:

[0017] Z=alphaCNN(X)+(1-alpha)RNN(X)

[0018] Where Z is the final feature representation, CNN(X) is the result of spatial feature extraction of input data X through convolutional neural network (CNN), RNN(X) is the result of time series feature processing of input data X through recurrent neural network (RNN), alpha is the weight coefficient, and 0 <alpha<1;

[0019] In this formula, X represents the preprocessed building information data, and its dimension is mtimesntimesp, where m represents the number of samples, n represents the spatial dimension, and p represents the number of channels. The calculation process of CNN(X) is as follows: first, a series of convolutional layers are used to perform convolution operations on the input data X, and the convolution kernel size is khtimeskwtimesc, where kh is the height of the convolution kernel, kw is the width of the convolution kernel, c is the number of channels of the convolution kernel, and the step size is shtimessw, where sh is the step size in the vertical direction, sw is the step size in the horizontal direction, and the padding method is same or valid, and the intermediate feature map is obtained; then the intermediate feature map is downsampled through the pooling layer, and the pooling kernel size is phtimespw, where ph is the height of the pooling kernel, pw is the width of the pooling kernel, and the step size is qhtimesqw, where qh is the step size in the vertical direction, and qw is the step size in the horizontal direction, and CNN(X) is obtained;

[0020] The calculation process of RNN(X) is as follows: expand the input data X according to the time series, and process the time series features through the recursive calculation of the hidden layer state. The update formula of the hidden layer state is Ht=sigma(W{xh}Xt+W{hh}H{t-1}+bh), where Ht is the hidden layer state at the current moment, Xt is the input at the current moment, W{xh} is the input weight matrix, W{hh} is the hidden layer weight matrix, bh is the bias vector, sigma is the activation function, and the output formula is Ot=sigma(W{ho}Ht+bo), where Ot is the output at the current moment, W{ho} is the output weight matrix, and bo is the output bias vector. RNN(X) is obtained, and the final feature representation Z is obtained by adding alphaCNN(X) and (1-alpha)RNN(X). The value of alpha is adjusted according to the characteristics of the data and the analysis requirements to balance the role of CNN and RNN in feature extraction, thereby better reducing the generation of abnormal data and improving data accuracy.

[0021] Furthermore, in the data compression step, the specific process of using the principal component analysis (PCA) algorithm to reduce the dimension of the data is as follows: first, the covariance matrix Cov(X) of the input data X is calculated, and then the covariance matrix is ​​decomposed to obtain the eigenvalues ​​lambda1, lambda2, cdots, lambdap and the corresponding eigenvectors v1, v2, cdots, vp, and the eigenvectors are sorted according to the size of the eigenvalues, and the eigenvectors corresponding to the first k largest eigenvalues ​​are selected to form a projection matrix P, where k is the dimension after dimensionality reduction, and the input data X is projected onto the projection matrix P to obtain the compressed data Y=P^TX.

[0022] Furthermore, the specific way to optimize the model by introducing the depthwise separable convolution (DSC) structure is as follows: the depthwise convolution operation is to perform a convolution operation on each channel of the input data separately, while the pointwise convolution operation is to perform a linear transformation on the channel dimension. Specifically, for a depthwise separable convolution with an input channel number in, an output channel number out, and a convolution kernel size of ktimesk, a depthwise convolution is performed first, that is, in convolution kernels of size ktimesk are used to perform a convolution operation on each channel of the input data separately to obtain in intermediate feature maps; then a pointwise convolution is performed, that is, out convolution kernels of size 1times1 are used to perform a linear transformation on the in intermediate feature maps to obtain the output feature map.

[0023] Furthermore, the RNN structure is optimized, and the specific principles of the RNN structure improved by the long short-term memory network (LSTM) and the gated recurrent unit (GRU) are as follows: the traditional RNN is prone to gradient vanishing or gradient exploding problems when dealing with long-term dependencies, which makes it difficult for the model to learn long-term time series features. LSTM can better control the flow and storage of information and avoid the gradient vanishing problem by introducing input gate, forget gate and output gate mechanisms. Specifically, the input gate is used to control how much of the input information at the current moment can be updated to the cell state, the forget gate is used to control how much of the cell state at the previous moment can be forgotten, and the output gate is used to control how much of the cell state at the current moment can be output as the hidden state at the current moment. GRU further simplifies the structure of LSTM, merges the input gate and the forget gate into an update gate, and controls the updating and forgetting of information through the update gate. In building information processing, since the equipment operation time series data often has long-term dependencies, the RNN structure improved by LSTM or GRU can better capture these long-term dependencies and improve the model's processing capabilities for building information and data accuracy.

[0024] Compared with the prior art, the invention provides a method for processing building information based on the Internet of Things. By using a professional hybrid algorithm of the convolutional neural network (CNN) and the recurrent neural network (RNN) used in the data preprocessing step, the effects of CNN and RNN in feature extraction can be balanced, thereby better reducing the generation of abnormal data and improving data accuracy. At the same time, by using deep separable convolution, the training and reasoning speed of the model can be significantly improved without affecting the feature extraction capability. The RNN structure improved by LSTM or GRU can better capture these long-term dependencies, thereby improving the model's processing capability for building information and data accuracy. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in the present invention. For ordinary technicians in this field, other drawings can also be obtained based on these drawings.

[0026] Figure 1 A schematic diagram of the overall structure provided for an embodiment of the present invention. DETAILED DESCRIPTION

[0027] In order to enable those skilled in the art to better understand the technical solution of the present invention, the present invention will be further described in detail below with reference to the accompanying drawings.

[0028] In the description of the present invention, it should be understood that the terms "center", "longitudinal", "lateral", "length", "width", "thickness", "up", "down", "front", "back", "left", "right", "vertical", "horizontal", "top", "bottom", "inside", "outside", "clockwise" and "counterclockwise" indicate orientations or positional relationships based on the orientations or positional relationships shown in the accompanying drawings, and are only for the convenience of describing the present invention and simplifying the description, and do not indicate or imply that the referred device or element must have a specific orientation, be constructed and operated in a specific orientation, and therefore should not be understood as limiting the present invention.

[0029] In addition, the terms "first" and "second" are only used for descriptive purposes and cannot be understood as indicating or implying relative importance or implicitly indicating the number of technical features indicated. Therefore, the features defined as "first" and "second" may explicitly or implicitly include one or more of the features. In the description of the present invention, "multiple" means two or more, unless otherwise clearly and specifically defined. In addition, the terms "installed", "connected" and "connected" should be understood in a broad sense, for example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be a direct connection, or it can be indirectly connected through an intermediate medium, or it can be the internal connection of two components. For those of ordinary skill in the art, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.

[0030] Example embodiments will be described more fully below with reference to the accompanying drawings, but the example embodiments may be embodied in different forms and should not be construed as limited to the embodiments set forth herein. On the contrary, the purpose of providing these embodiments is to make the present disclosure thorough and complete and to enable those skilled in the art to fully understand the scope of the present disclosure.

[0031] In the absence of conflict, the various embodiments of the present disclosure and the various features therein may be combined with each other.

[0032] As used herein, the term "and / or" includes any and all combinations of one or more of the associated listed items.

[0033] The terms used herein are only used to describe specific embodiments and are not intended to limit the present disclosure. As used herein, the singular forms "a", "an" and "the" are also intended to include the plural forms, unless the context clearly indicates otherwise. It will also be understood that when the terms "comprising" and / or "made of" are used in this specification, the presence of the features, wholes, steps, operations, elements and / or components is specified, but the presence or addition of one or more other features, wholes, steps, operations, elements, components and / or groups thereof is not excluded.

[0034] The embodiments herein may be described with reference to plan views and / or cross-sectional views with the aid of idealized schematic diagrams of the present disclosure. Therefore, the example illustrations may be modified according to manufacturing techniques and / or tolerances. Therefore, the embodiments are not limited to the embodiments shown in the drawings, but include modifications of the configurations formed based on the manufacturing process. Therefore, the regions illustrated in the drawings have schematic properties, and the shapes of the regions shown in the drawings illustrate the specific shapes of the regions of the elements, but are not intended to be limiting.

[0035] Unless otherwise defined, all terms (including technical and scientific terms) used herein have the same meaning as commonly understood by those of ordinary skill in the art. It will also be understood that terms such as those defined in commonly used dictionaries should be interpreted as having a meaning consistent with their meaning in the context of the relevant art and the present disclosure, and will not be interpreted as having an idealized or overly formal meaning unless explicitly defined as such herein.

[0036] See also Figure 1 , a building information processing method based on the Internet of Things, comprising the following steps:

[0037] Build an IoT sensor network that covers the entire building, collects various physical parameters and equipment operating status information in real time, and transmits it to the data processing center;

[0038] In the data processing center, the collected building information data is preprocessed, including data cleaning and data normalization operations, to remove noise and abnormal data in preparation for subsequent analysis;

[0039] The pre-processed data is analyzed using a professional hybrid algorithm of convolutional neural network (CNN) and recurrent neural network (RNN). Specifically, the spatial features in the building information data are first extracted through CNN to obtain a preliminary feature representation, and then these features are input into RNN. RNN is used to process the time series features to mine the potential patterns and trends in the building, effectively reducing the generation of abnormal data and significantly improving data accuracy.

[0040] Data compression technology is used to compress the building information data after CNN and RNN processing. Specifically, the principal component analysis (PCA) algorithm is used to reduce the dimensionality of the data, project the high-dimensional data into a low-dimensional space, retain the main feature information, and remove redundant and noisy data, thereby further improving the storage and transmission efficiency of the data without affecting the analysis results of the data.

[0041] The model structure of the compressed data is optimized, specifically: the depthwise separable convolution (DSC) structure is introduced to decompose the traditional convolution operation into two steps: depthwise convolution and pointwise convolution. The depthwise convolution is used to extract the spatial features of the data, and the pointwise convolution is used to perform linear transformation in the channel dimension. This can reduce the amount of calculation and the number of parameters while maintaining good feature extraction capabilities. At the same time, the RNN structure is optimized, and the RNN structure improved by the long short-term memory network (LSTM) or the gated recurrent unit (GRU) is used to better handle long-term dependencies and avoid the gradient vanishing problem, thereby improving the performance of the entire hybrid algorithm and data accuracy.

[0042] Based on the analysis results, an intelligent model of the building is established. The model can accurately predict the future building status based on the current building information data, including temperature change trends, equipment failure probability, energy consumption requirements, and provide corresponding early warnings and optimization suggestions;

[0043] The prediction results and optimization suggestions of the intelligent model are sent to various devices and systems in the building through the Internet of Things platform to achieve intelligent control and optimization management of the building, such as automatically adjusting the air conditioning temperature and optimizing the equipment operation strategy;

[0044] Regularly evaluate and optimize the building information processing system, adjust and improve data compression technology and model structure optimization according to actual operation conditions, so as to continuously improve the accuracy and adaptability of the system;

[0045] Provide a user interface that enables users to easily view the operating status, analysis results and optimization suggestions of the building information processing system, and perform corresponding operations and management, including setting early warning thresholds and adjusting optimization strategies.

[0046] Furthermore, when constructing the IoT sensor network, a variety of high-precision sensors are used, including temperature sensors, humidity sensors, light sensors, and air pressure sensors, to ensure the accuracy and comprehensiveness of the collected building information data.

[0047] Furthermore, for the professional hybrid algorithm of convolutional neural network (CNN) and recurrent neural network (RNN) used in the data preprocessing step, the specific calculation formula is:

[0048] Z=alphaCNN(X)+(1-alpha)RNN(X)

[0049] Where Z is the final feature representation, CNN(X) is the result of spatial feature extraction of input data X through convolutional neural network (CNN), RNN(X) is the result of time series feature processing of input data X through recurrent neural network (RNN), alpha is the weight coefficient, and 0 <alpha<1;

[0050] In this formula, X represents the preprocessed building information data, and its dimension is mtimesntimesp, where m represents the number of samples, n represents the spatial dimension, and p represents the number of channels. The calculation process of CNN(X) is as follows: first, a series of convolutional layers are used to perform convolution operations on the input data X, and the convolution kernel size is khtimeskwtimesc, where kh is the height of the convolution kernel, kw is the width of the convolution kernel, c is the number of channels of the convolution kernel, and the step size is shtimessw, where sh is the step size in the vertical direction, sw is the step size in the horizontal direction, and the padding method is same or valid, and the intermediate feature map is obtained; then the intermediate feature map is downsampled through the pooling layer, and the pooling kernel size is phtimespw, where ph is the height of the pooling kernel, pw is the width of the pooling kernel, and the step size is qhtimesqw, where qh is the step size in the vertical direction, and qw is the step size in the horizontal direction, and CNN(X) is obtained;

[0051] The calculation process of RNN(X) is as follows: expand the input data X according to the time series, and process the time series features through the recursive calculation of the hidden layer state. The update formula of the hidden layer state is Ht=sigma(W{xh}Xt+W{hh}H{t-1}+bh), where Ht is the hidden layer state at the current moment, Xt is the input at the current moment, W{xh} is the input weight matrix, W{hh} is the hidden layer weight matrix, bh is the bias vector, sigma is the activation function, and the output formula is Ot=sigma(W{ho}Ht+bo), where Ot is the output at the current moment, W{ho} is the output weight matrix, and bo is the output bias vector. RNN(X) is obtained, and the final feature representation Z is obtained by adding alphaCNN(X) and (1-alpha)RNN(X). The value of alpha is adjusted according to the characteristics of the data and the analysis requirements to balance the role of CNN and RNN in feature extraction, thereby better reducing the generation of abnormal data and improving data accuracy.

[0052] Furthermore, in the data compression step, the specific process of using the principal component analysis (PCA) algorithm to reduce the dimension of the data is as follows: first, the covariance matrix Cov(X) of the input data X is calculated, and then the covariance matrix is ​​decomposed to obtain the eigenvalues ​​lambda1, lambda2, cdots, lambdap and the corresponding eigenvectors v1, v2, cdots, vp, the eigenvectors are sorted according to the size of the eigenvalues, and the eigenvectors corresponding to the first k largest eigenvalues ​​are selected to form a projection matrix P, where k is the dimension after dimensionality reduction, and the input data X is projected onto the projection matrix P to obtain the compressed data Y=P^TX. In this way, the dimension of the data is reduced from p dimension to k dimension, most of the redundant and noise data are removed, while the main feature information is retained, thereby improving the data compression ratio and storage and transmission efficiency.

[0053] Furthermore, the specific way to optimize the model by introducing the depthwise separable convolution (DSC) structure is as follows: the depthwise convolution operation is to perform convolution operation on each channel of the input data separately, while the pointwise convolution operation is to perform linear transformation on the channel dimension. Specifically, for a depthwise separable convolution with in input channels, out output channels, and ktimesk convolution kernel size, firstly, the depthwise convolution is performed, that is, in convolution kernels of size ktimesk are used to perform convolution operation on each channel of the input data separately to obtain in intermediate feature maps; then the pointwise convolution is performed, that is, out convolution kernels of size 1times1 are used to perform linear transformation on in intermediate feature maps to obtain output feature maps. Compared with the traditional convolution operation, the depthwise separable convolution greatly reduces the amount of calculation and the number of parameters, because the depthwise convolution only performs local convolution on each channel, while the pointwise convolution performs simple linear transformation on the channel dimension. In building information processing, due to the high dimensionality of the data, the use of depthwise separable convolution can significantly improve the training and reasoning speed of the model without affecting the feature extraction capability.

[0054] Furthermore, the RNN structure is optimized, and the specific principles of the RNN structure improved by the long short-term memory network (LSTM) and the gated recurrent unit (GRU) are as follows: the traditional RNN is prone to gradient vanishing or gradient exploding problems when dealing with long-term dependencies, which makes it difficult for the model to learn long-term time series features. LSTM can better control the flow and storage of information and avoid the gradient vanishing problem by introducing input gate, forget gate and output gate mechanisms. Specifically, the input gate is used to control how much of the input information at the current moment can be updated to the cell state, the forget gate is used to control how much of the cell state at the previous moment can be forgotten, and the output gate is used to control how much of the cell state at the current moment can be output as the hidden state at the current moment. GRU further simplifies the structure of LSTM, merges the input gate and the forget gate into an update gate, and controls the updating and forgetting of information through the update gate. In building information processing, since the equipment operation time series data often has long-term dependencies, the RNN structure improved by LSTM or GRU can better capture these long-term dependencies and improve the model's processing capabilities for building information and data accuracy.

[0055] The above description is only by way of illustration of certain exemplary embodiments of the present invention. It is undoubted that those skilled in the art can modify the described embodiments in various ways without departing from the spirit and scope of the present invention. Therefore, the above drawings and descriptions are illustrative in nature and should not be construed as limiting the scope of protection of the claims of the present invention.

Claims

1. A building information processing method based on the Internet of Things, characterized in that: The following steps are involved: Build an IoT sensor network that covers the entire building, collects various physical parameters and equipment operating status information in real time, and transmits it to the data processing center; In the data processing center, the collected building information data is preprocessed, including data cleaning and data normalization operations, to remove noise and abnormal data in preparation for subsequent analysis; The preprocessed data is analyzed using a professional hybrid algorithm of convolutional neural network (CNN) and recurrent neural network (RNN). Specifically, the spatial features in the building information data are first extracted through CNN to obtain a preliminary feature representation, and then these features are input into RNN, which is used to process the time series features; Data compression technology is used to compress the building information data after CNN and RNN processing, specifically: principal component analysis (PCA) algorithm is used to reduce the dimension of the data; The model structure is optimized for the compressed data, specifically: the depthwise separable convolution (DSC) structure is introduced to decompose the traditional convolution operation into two steps: depthwise convolution and pointwise convolution. The depthwise convolution is used to extract the spatial features of the data, and the pointwise convolution is used to perform linear transformation in the channel dimension. At the same time, the RNN structure is optimized, and the RNN structure improved by the long short-term memory network (LSTM) or the gated recurrent unit (GRU) is adopted. Based on the analysis results, an intelligent model of the building is established. The model can accurately predict the future building status based on the current building information data, including temperature change trends, equipment failure probability, energy consumption requirements, and provide corresponding early warnings and optimization suggestions; The prediction results and optimization suggestions of the intelligent model are sent to various devices and systems in the building through the Internet of Things platform to achieve intelligent control and optimized management of the building; Regularly evaluate and optimize the building information processing system, and adjust and improve data compression technology and model structure optimization according to actual operation conditions; Provide a user interface that enables users to easily view the operating status, analysis results and optimization suggestions of the building information processing system, and perform corresponding operations and management, including setting early warning thresholds and adjusting optimization strategies.

2. The method for processing building information based on the Internet of Things according to claim 1, characterized in that: When building the IoT sensor network, a variety of high-precision sensors are used, including temperature sensors, humidity sensors, light sensors, and air pressure sensors.

3. The method for processing building information based on the Internet of Things according to claim 2, characterized in that: For the professional hybrid algorithm of convolutional neural network (CNN) and recurrent neural network (RNN) used in the data preprocessing step, the specific calculation formula is: Z=alphaCNN(X)+(1-alpha)RNN(X) Where Z is the final feature representation, CNN(X) is the result of spatial feature extraction of input data X through convolutional neural network (CNN), RNN(X) is the result of time series feature processing of input data X through recurrent neural network (RNN), alpha is the weight coefficient, and 0 <alpha<1; In this formula, X represents the preprocessed building information data, and its dimension is mtimesntimesp, where m represents the number of samples, n represents the spatial dimension, and p represents the number of channels. The calculation process of CNN(X) is as follows: first, a series of convolutional layers are used to perform convolution operations on the input data X, and the convolution kernel size is khtimeskwtimesc, where kh is the height of the convolution kernel, kw is the width of the convolution kernel, c is the number of channels of the convolution kernel, and the step size is shtimessw, where sh is the step size in the vertical direction, sw is the step size in the horizontal direction, and the padding method is same or valid, and the intermediate feature map is obtained; then the intermediate feature map is downsampled through the pooling layer, and the pooling kernel size is phtimespw, where ph is the height of the pooling kernel, pw is the width of the pooling kernel, and the step size is qhtimesqw, where qh is the step size in the vertical direction, and qw is the step size in the horizontal direction, and CNN(X) is obtained; The calculation process of RNN(X) is as follows: expand the input data X according to the time series, and process the time series features through the recursive calculation of the hidden layer state. The update formula of the hidden layer state is Ht=sigma(W{xh}Xt+W{hh}H{t-1}+bh), where Ht is the hidden layer state at the current moment, Xt is the input at the current moment, W{xh} is the input weight matrix, W{hh} is the hidden layer weight matrix, bh is the bias vector, sigma is the activation function, and the output formula is Ot=sigma(W{ho}Ht+bo), where Ot is the output at the current moment, W{ho} is the output weight matrix, and bo is the output bias vector. RNN(X) is obtained, and the final feature representation Z is obtained by adding alphaCNN(X) and (1-alpha)RNN(X), where the value of alpha is adjusted according to the characteristics of the data and analysis requirements.

4. The method for processing building information based on the Internet of Things according to claim 3 is characterized in that: In the data compression step, the specific process of using the principal component analysis (PCA) algorithm to reduce the dimensionality of the data is as follows: first, the covariance matrix Cov(X) of the input data X is calculated, and then the covariance matrix is ​​decomposed to obtain the eigenvalues ​​lambda1, lambda2, cdots, lambdap and the corresponding eigenvectors v1, v2, cdots, vp. The eigenvectors are sorted according to the size of the eigenvalues, and the eigenvectors corresponding to the first k largest eigenvalues ​​are selected to form a projection matrix P, where k is the dimension after dimensionality reduction. The input data X is projected onto the projection matrix P to obtain the compressed data Y=P^TX.

5. The method for processing building information based on the Internet of Things according to claim 4, characterized in that: The specific way to optimize the model by introducing the depthwise separable convolution (DSC) structure is as follows: the depthwise convolution operation is to perform a convolution operation on each channel of the input data separately, while the pointwise convolution operation is to perform a linear transformation on the channel dimension. Specifically, for a depthwise separable convolution with an input channel number in, an output channel number out, and a convolution kernel size of ktimesk, a depthwise convolution is performed first, that is, in convolution kernels of size ktimesk are used to perform a convolution operation on each channel of the input data separately to obtain in intermediate feature maps; then a pointwise convolution is performed, that is, out convolution kernels of size 1times1 are used to perform a linear transformation on the in intermediate feature maps to obtain the output feature map.

6. The method for processing building information based on the Internet of Things according to claim 5, characterized in that: The RNN structure is optimized and the specific principles of the RNN structure improved by using the long short-term memory network (LSTM) and the gated recurrent unit (GRU) are as follows: LSTM introduces input gate, forget gate and output gate mechanism. The input gate is used to control how much of the input information at the current moment can be updated to the cell state, the forget gate is used to control how much of the cell state at the previous moment can be forgotten, and the output gate is used to control how much of the cell state at the current moment can be output as the hidden state at the current moment. GRU further simplifies the structure of LSTM, merges the input gate and the forget gate into an update gate, and controls the updating and forgetting of information through the update gate.

Images