Device matching method and system for Internet of Things devices

Abstract

The present invention provides a device matching method and system for Internet of Things devices, which relate to the field of data processing technology. In the present invention, based on the historical Internet of Things device operation data corresponding to each Internet of Things device to be matched and the device deployment location information corresponding to each Internet of Things device to be matched, a corresponding first device data distribution set is constructed; using the target data node screening neural network, the first device data distribution set is selected to form a first device data distribution subset corresponding to the first device data distribution set, and the Internet of Things devices to be matched corresponding to each data distribution node included in the first device data distribution subset are combined to form a first set of Internet of Things devices to be matched. Based on the aforementioned steps, the reliability of device matching of Internet of Things devices can be improved to a certain extent.

Description

Technical Field

[0001] The present invention relates to the field of data processing technology, and in particular to a device matching method and system for Internet of Things devices. Background Art

[0002] As data processing technology continues to mature, it is being utilized in many fields for corresponding data processing. For example, in the field of the Internet of Things (IoT), data processing technology may be needed to perform device matching on IoT devices. In existing technologies, device matching is generally performed based on the deployment location of IoT devices. However, since the deployment location of IoT devices generally contains relatively little information, the basis for device matching based on this information is relatively simple, resulting in poor reliability of device matching. Summary of the Invention

[0003] In view of this, an object of the present invention is to provide a device matching method and system for IoT devices, so as to improve the reliability of device matching of IoT devices to a certain extent.

[0004] To achieve the above objectives, the embodiments of the present invention adopt the following technical solutions:

[0005] A device matching method for an Internet of Things device, comprising:

[0006] Performing historical data extraction and processing on a plurality of IoT devices to be matched to obtain historical IoT device operation data corresponding to each of the IoT devices to be matched, and then constructing a corresponding first device data distribution set based on the historical IoT device operation data corresponding to each of the IoT devices to be matched and the device deployment location information corresponding to each of the IoT devices to be matched, wherein each data distribution node in the first device data distribution set corresponds to a IoT device to be matched, and a distribution position relationship between every two data distribution nodes is correlated with the device deployment location information corresponding to the two IoT devices to be matched;

[0007] Using a target data node screening neural network, the first device data distribution set is selected to form a first device data distribution subset corresponding to the first device data distribution set, and the to-be-matched IoT devices corresponding to each data distribution node included in the first device data distribution subset are combined to form a first to-be-matched IoT device set, where every two to-be-matched IoT devices included in the first to-be-matched IoT device set have a matching relationship;

[0008] In the first device data distribution set, the first device data distribution subset corresponding to the first device data distribution set is screened to form a second device data distribution subset corresponding to the first device data distribution set, and then the target data node is used to screen the neural network to select the second device data distribution subset to form a third device data distribution subset corresponding to the second device data distribution subset, and the IoT devices to be matched corresponding to each data distribution node included in the third device data distribution subset are combined to form a second set of IoT devices to be matched, and there is a matching relationship between every two IoT devices to be matched included in the second set of IoT devices to be matched.

[0009] In some preferred embodiments, in the device matching method for the Internet of Things devices, the target data node screening neural network is formed based on network optimization of the initial data node screening neural network. The network optimization process of the initial data node screening neural network includes:

[0010] Extracting an exemplary supervised device data distribution set, an exemplary unsupervised device data distribution set, a supervised distribution node selection identifier corresponding to the exemplary supervised device data distribution set, a to-be-compared distribution node selection identifier corresponding to the exemplary unsupervised device data distribution set, and a pre-built initial data node screening neural network, wherein the initial data node screening neural network includes a first feature mining subnetwork, a second feature mining subnetwork, a feature matching fusion subnetwork, and a feature restoration subnetwork;

[0011] Using the first feature mining sub-network, a feature mining operation is performed on the exemplary supervisory device data distribution set and the data aggregation feature representation corresponding to the supervisory distribution node selection identifier to form a supervisory data feature representation corresponding to the exemplary supervisory device data distribution set;

[0012] Using the second feature mining sub-network, perform a feature mining operation on the exemplary unsupervised device data distribution set to form an unsupervised data feature representation corresponding to the exemplary unsupervised device data distribution set;

[0013] Using the feature matching fusion subnetwork, perform a feature matching fusion operation on the supervised data feature representation and the unsupervised data feature representation to form a matched fusion data feature representation corresponding to the exemplary unsupervised device data distribution set;

[0014] Using the feature restoration subnetwork, perform a feature restoration operation on the unsupervised data feature representation and the matched fusion data feature representation to output a predicted distribution node selection identifier corresponding to the exemplary unsupervised device data distribution set;

[0015] According to the predicted distribution node selection identifier corresponding to the exemplary unsupervised device data distribution set and the distribution node selection identifier to be compared, the corresponding predicted learning cost value is calculated, and then based on the predicted learning cost value, the first feature mining sub-network, the second feature mining sub-network, the feature matching fusion sub-network and the feature restoration sub-network are network optimized to form a corresponding target data node screening neural network.

[0016] In some preferred embodiments, in the above-mentioned device matching method for IoT devices, the first feature mining subnetwork includes at least one feature mining unit; the step of using the first feature mining subnetwork to perform a feature mining operation on the exemplary supervisory device data distribution set and the data aggregation feature representation corresponding to the supervisory distribution node selection identifier to form a supervisory data feature representation corresponding to the exemplary supervisory device data distribution set includes:

[0017] The at least one feature mining unit is used to perform a feature mining operation on the exemplary supervisory device data distribution set and the data aggregation feature representation corresponding to the supervisory distribution node selection identifier to form a supervisory data feature representation corresponding to the exemplary supervisory device data distribution set.

[0018] In some preferred embodiments, in the above-mentioned device matching method for IoT devices, the first feature mining subnetwork includes a cascaded input feature mining unit, an intermediate feature mining unit, and an output feature mining unit;

[0019] The step of using the at least one feature mining unit to perform a feature mining operation on the exemplary supervisory device data distribution set and the data aggregation feature representation corresponding to the supervisory distribution node selection identifier to form a supervisory data feature representation corresponding to the exemplary supervisory device data distribution set includes:

[0020] Loading the exemplary supervisory device data distribution set and the data aggregation feature representation corresponding to the supervisory distribution node selection identifier, and performing an input feature extraction operation using the input feature mining unit to form an input selected data feature representation corresponding to the exemplary supervisory device data distribution set;

[0021] Loading the input-end selected data feature representation to perform an intermediate feature extraction operation using the intermediate feature mining unit to form an intermediate selected data feature representation corresponding to the exemplary supervisory device data distribution set;

[0022] The intermediate selected data feature representation is loaded to perform an output feature extraction operation using the output feature mining unit to form an output selected data feature representation as a supervisory data feature representation corresponding to the exemplary supervisory device data distribution set.

[0023] In some preferred embodiments, in the above-mentioned device matching method for IoT devices, the second feature mining subnetwork includes a primary feature mining unit, an intermediate feature mining unit, an advanced feature mining unit, and a target number of feature aggregation units;

[0024] The step of using the second feature mining sub-network to perform a feature mining operation on the exemplary unsupervised device data distribution set to form an unsupervised data feature representation corresponding to the exemplary unsupervised device data distribution set includes:

[0025] The primary feature mining unit, the intermediate feature mining unit, the high-level feature mining unit, and the target number of feature aggregation units are used to perform a feature mining operation on the exemplary unsupervised device data distribution set to form an unsupervised data feature representation corresponding to the exemplary unsupervised device data distribution set.

[0026] In some preferred embodiments, in the device matching method for the Internet of Things devices, the target number of feature aggregation units includes an input-end feature aggregation unit configured between the primary feature mining unit and the intermediate feature mining unit, an intermediate feature aggregation unit configured between the intermediate feature mining unit and the advanced feature mining unit, and an output-end feature aggregation unit configured at the output end of the advanced feature mining unit;

[0027] The step of performing a feature mining operation on the exemplary unsupervised device data distribution set using the primary feature mining unit, the intermediate feature mining unit, the high-level feature mining unit, and the target number of feature aggregation units to form an unsupervised data feature representation corresponding to the exemplary unsupervised device data distribution set includes:

[0028] Loading the exemplary unsupervised device data distribution set to perform a primary feature extraction operation using the primary feature mining unit to form a primary extracted data feature representation corresponding to the exemplary unsupervised device data distribution set;

[0029] Loading the primary decimated data feature representation and the input-end decimated data feature representation to perform a primary feature representation aggregation operation using the input-end feature aggregation unit to form an aggregated primary decimated data feature representation corresponding to the exemplary unsupervised device data distribution set;

[0030] Loading the aggregated primary selected data feature representation to perform an intermediate feature extraction operation using the intermediate feature mining unit to form an intermediate selected data feature representation corresponding to the exemplary unsupervised device data distribution set;

[0031] Loading the intermediate decimated data feature representation and the intermediate decimated data feature representation to perform an intermediate feature representation aggregation operation using the intermediate feature aggregation unit to form an aggregated intermediate decimated data feature representation corresponding to the exemplary unsupervised device data distribution set;

[0032] Loading the aggregated intermediate-level selected data feature representation to perform a high-level feature extraction operation using the high-level feature mining unit to form a high-level selected data feature representation corresponding to the exemplary unsupervised device data distribution set;

[0033] The high-level abstracted data feature representation and the supervised data feature representation are loaded to perform a high-level feature representation aggregation operation using the output-end feature aggregation unit to form a corresponding aggregated high-level abstracted data feature representation as an unsupervised data feature representation corresponding to the exemplary unsupervised device data distribution set.

[0034] In some preferred embodiments, in the device matching method for the above-mentioned IoT devices, the step of screening out the first device data distribution subset corresponding to the first device data distribution set in the first device data distribution set to form a second device data distribution subset corresponding to the first device data distribution set, then using the target data node screening neural network to perform sampling processing on the second device data distribution subset to form a third device data distribution subset corresponding to the second device data distribution subset, and combining the IoT devices to be matched corresponding to each data distribution node included in the third device data distribution subset to form a second set of IoT devices to be matched includes:

[0035] In the first device data distribution set, a first device data distribution subset corresponding to the first device data distribution set is filtered out to form a second device data distribution subset corresponding to the first device data distribution set;

[0036] Adjusting the distribution position relationship between every two data distribution nodes in the second device data distribution subset to form a new second device data distribution subset;

[0037] The target data node screening neural network is used to select the new second device data distribution subset to form a third device data distribution subset corresponding to the new second device data distribution subset, and the IoT devices to be matched corresponding to each data distribution node included in the third device data distribution subset are combined to form a second set of IoT devices to be matched corresponding to the third device data distribution subset.

[0038] In some preferred embodiments, in the device matching method for the IoT device, the step of adjusting the distribution position relationship between every two data distribution nodes in the second device data distribution subset to form a new second device data distribution subset includes:

[0039] Calculating a ratio between the number of data distribution nodes corresponding to the first device data distribution subset and the number of data distribution nodes corresponding to the first device data distribution set;

[0040] According to the result of the ratio calculation, the distribution position relationship between every two data distribution nodes in the second device data distribution subset is shortened by a corresponding proportion to adjust the distribution position relationship and form a new second device data distribution subset.

[0041] In some preferred embodiments, in the device matching method for the IoT device described above, the step of screening out the first device data distribution subset corresponding to the first device data distribution set in the first device data distribution set to form a second device data distribution subset corresponding to the first device data distribution set, then using the target data node screening neural network to perform sampling processing on the second device data distribution subset to form a third device data distribution subset corresponding to the second device data distribution subset, and combining the IoT devices to be matched corresponding to each data distribution node included in the third device data distribution subset to form a second set of IoT devices to be matched, further includes:

[0042] In the new second device data distribution subset, a third device data distribution subset corresponding to the new second device data distribution subset is filtered out to form a fourth device data distribution subset corresponding to the new second device data distribution subset;

[0043] The IoT devices to be matched corresponding to each data distribution node included in the fourth device data distribution subset are combined to form a third set of IoT devices to be matched corresponding to the fourth device data distribution subset; or, the distribution position relationship between every two data distribution nodes in the fourth device data distribution subset is adjusted to form a new fourth device data distribution subset; the new fourth device data distribution subset is selected using the target data node screening neural network to form a fifth device data distribution subset corresponding to the new fourth device data distribution subset, and the IoT devices to be matched corresponding to each data distribution node included in the fifth device data distribution subset are combined to form a third set of IoT devices to be matched corresponding to the fifth device data distribution subset, and every two IoT devices to be matched included in the third set of IoT devices to be matched have a matching relationship.

[0044] An embodiment of the present invention further provides a device matching processing platform, including:

[0045] a device data distribution set construction module, configured to extract and process historical data from a plurality of IoT devices to be matched, to obtain historical IoT device operation data corresponding to each IoT device to be matched, and then construct a corresponding first device data distribution set based on the historical IoT device operation data corresponding to each IoT device to be matched and the device deployment location information corresponding to each IoT device to be matched, wherein each data distribution node in the first device data distribution set corresponds to a IoT device to be matched, and the distribution position relationship between each two data distribution nodes is correlated with the device deployment location information corresponding to the two IoT devices to be matched;

[0046] a first device matching and combining module, configured to select the first device data distribution set using a target data node screening neural network to form a first device data distribution subset corresponding to the first device data distribution set, and to combine the to-be-matched IoT devices corresponding to each data distribution node included in the first device data distribution subset to form a first to-be-matched IoT device set, wherein each two to-be-matched IoT devices included in the first to-be-matched IoT device set have a matching relationship;

[0047] The second device matching and combination module is used to screen out the first device data distribution subset corresponding to the first device data distribution set in the first device data distribution set to form a second device data distribution subset corresponding to the first device data distribution set, and then use the target data node screening neural network to select the second device data distribution subset to form a third device data distribution subset corresponding to the second device data distribution subset, and combine the IoT devices to be matched corresponding to each data distribution node included in the third device data distribution subset to form a second set of IoT devices to be matched, where there is a matching relationship between every two IoT devices to be matched included in the second set of IoT devices to be matched.

[0048] An embodiment of the present invention provides a device matching method and system for Internet of Things devices. The method and system construct a corresponding first device data distribution set based on historical Internet of Things device operation data corresponding to each to-be-matched Internet of Things device and device deployment location information corresponding to each to-be-matched Internet of Things device; utilize a target data node screening neural network to select the first device data distribution set to form a first device data distribution subset corresponding to the first device data distribution set, and combine the to-be-matched Internet of Things devices corresponding to each data distribution node included in the first device data distribution subset to form a first to-be-matched Internet of Things device set; and, in the first device data distribution set, screen out the first device data distribution subset corresponding to the first device data distribution set to form a second device data distribution subset corresponding to the first device data distribution set; then utilize the target data node screening neural network to select the second device data distribution subset to form a third device data distribution subset corresponding to the second device data distribution subset; and combine the to-be-matched Internet of Things devices corresponding to each data distribution node included in the third device data distribution subset to form a second to-be-matched Internet of Things device set. Based on the aforementioned steps, a neural network can be used to comprehensively analyze and process the historical IoT device operation data and device deployment location information corresponding to the IoT device to be matched, so that the reliability of data processing is higher, and the reliability of device matching of IoT devices can be improved to a certain extent.

[0049] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, preferred embodiments are given below and described in detail with reference to the accompanying drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0050] Figure 1 This is a structural block diagram of the device matching processing platform provided by an embodiment of the present invention.

[0051] Figure 2A flowchart of the steps of a device matching method for IoT devices provided in an embodiment of the present invention is provided.

[0052] Figure 3 A schematic diagram of the modules included in the device matching system for IoT devices provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0053] To make the objectives, technical solutions, and advantages of the embodiments of the present invention more clear, the technical solutions of the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Generally, the components of the embodiments of the present invention described and shown in the drawings herein can be arranged and designed in various different configurations.

[0054] Therefore, the following detailed description of the embodiments of the present invention provided in the accompanying drawings is not intended to limit the scope of the invention as claimed, but rather merely represents selected embodiments of the present invention. All other embodiments derived by persons of ordinary skill in the art based on the embodiments of the present invention without creative effort shall fall within the scope of protection of the present invention.

[0055] like Figure 1 As shown, an embodiment of the present invention provides a device matching processing platform, wherein the device matching processing platform may include a memory and a processor.

[0056] In detail, the memory and the processor are electrically connected directly or indirectly to achieve data transmission or interaction. For example, they can be electrically connected to each other via one or more communication buses or signal lines. The memory can store at least one software function module (computer program) that can exist in the form of software or firmware. The processor can be used to execute the executable computer program stored in the memory, thereby implementing the device matching method for the Internet of Things device provided in the embodiment of the present invention.

[0057] For example, in some embodiments, the memory may be, but is not limited to, a random access memory (RAM), a read-only memory (ROM), a programmable read-only memory (PROM), an erasable programmable read-only memory (EPROM), an electrically erasable programmable read-only memory (EEPROM), etc. The processor may be a general-purpose processor, including a central processing unit (CPU), a network processor (NP), a system on a chip (SoC), etc.; it may also be a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field programmable gate array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components.

[0058] For example, in some implementations, the device matching processing platform may be a server with data processing capabilities (which may also include a communication unit, etc.).

[0059] Combine Figure 2 The present invention also provides a device matching method for IoT devices, which can be applied to the aforementioned device matching processing platform. The method steps defined in the process related to the device matching method for IoT devices can be implemented by the device matching processing platform.

[0060] The following will Figure 2 The specific process shown is explained in detail.

[0061] Step S110: Perform historical data extraction processing on multiple IoT devices to be matched to obtain historical IoT device operation data corresponding to each of the IoT devices to be matched, and then construct a corresponding first device data distribution set based on the historical IoT device operation data corresponding to each of the IoT devices to be matched and the device deployment location information corresponding to each of the IoT devices to be matched.

[0062] In an embodiment of the present invention, the device matching processing platform can perform historical data extraction processing on multiple IoT devices to be matched to obtain the historical IoT device operation data corresponding to each IoT device to be matched, and then construct a corresponding first device data distribution set based on the historical IoT device operation data corresponding to each IoT device to be matched and the device deployment location information corresponding to each IoT device to be matched. Each data distribution node in the first device data distribution set corresponds to an IoT device to be matched, and the distribution position relationship between each two data distribution nodes is correlated with the device deployment location information corresponding to the two IoT devices to be matched (for example, the node attribute information of each data distribution node in the first device data distribution set can be the historical IoT device operation data of the corresponding IoT device to be matched, and the historical IoT device operation data can be the device operation log data of the corresponding IoT device to be matched, etc.).

[0063] Step S120: Use the target data node screening neural network to select the first device data distribution set to form a first device data distribution subset corresponding to the first device data distribution set, and combine the IoT devices to be matched corresponding to each data distribution node included in the first device data distribution subset to form a first set of IoT devices to be matched.

[0064] In an embodiment of the present invention, the device matching processing platform can utilize a target data node screening neural network to select the first device data distribution set to form a first device data distribution subset corresponding to the first device data distribution set, and combine the IoT devices to be matched corresponding to each data distribution node included in the first device data distribution subset to form a first set of IoT devices to be matched. Each two IoT devices to be matched included in the first set of IoT devices to be matched have a matching relationship.

[0065] Step S130: In the first device data distribution set, the first device data distribution subset corresponding to the first device data distribution set is screened out to form a second device data distribution subset corresponding to the first device data distribution set, and then the target data node is used to screen the neural network to perform sampling processing on the second device data distribution subset to form a third device data distribution subset corresponding to the second device data distribution subset, and the IoT devices to be matched corresponding to each data distribution node included in the third device data distribution subset are combined to form a second set of IoT devices to be matched.

[0066] In an embodiment of the present invention, the device matching processing platform may screen out the first device data distribution subset corresponding to the first device data distribution set in the first device data distribution set to form a second device data distribution subset corresponding to the first device data distribution set, then utilize the target data node screening neural network to perform sampling processing on the second device data distribution subset to form a third device data distribution subset corresponding to the second device data distribution subset, and combine the IoT devices to be matched corresponding to each data distribution node included in the third device data distribution subset to form a second set of IoT devices to be matched. There is a matching relationship between each two IoT devices to be matched included in the second set of IoT devices to be matched.

[0067] Based on the aforementioned steps, a neural network can be used to comprehensively analyze and process the historical IoT device operation data and device deployment location information corresponding to the IoT devices to be matched, so that the reliability of data processing is higher, and the reliability of device matching of IoT devices can be improved to a certain extent, thereby improving the deficiencies in the existing technology.

[0068] For example, in some embodiments, step S120 may further include:

[0069] Extracting an exemplary supervised device data distribution set, an exemplary unsupervised device data distribution set, a supervised distribution node sampling identifier corresponding to the exemplary supervised device data distribution set (exemplarily, in the exemplary supervised device data distribution set, exemplary IoT devices corresponding to the supervised distribution node sampling identifier have a matching relationship), a to-be-compared distribution node sampling identifier corresponding to the exemplary unsupervised device data distribution set (exemplarily, in the exemplary unsupervised device data distribution set, exemplary IoT devices corresponding to the to-be-compared distribution node sampling identifier have a matching relationship), and a pre-built initial data node screening neural network, wherein the initial data node screening neural network includes a first feature mining subnetwork, a second feature mining subnetwork, a feature matching fusion subnetwork, and a feature restoration subnetwork;

[0070] Using the first feature mining sub-network, a feature mining operation is performed on the data aggregation feature representation corresponding to the exemplary supervisory device data distribution set and the supervisory distribution node selection identifier to form a supervisory data feature representation corresponding to the exemplary supervisory device data distribution set (exemplarily, the exemplary supervisory device data distribution set and the supervisory distribution node selection identifier can first be aggregated, such as by concatenating, to form a corresponding data aggregation feature representation, and then, the first feature mining sub-network is used to perform a feature mining operation on the data aggregation feature representation to form a supervisory data feature representation corresponding to the exemplary supervisory device data distribution set);

[0071] Using the second feature mining sub-network, perform a feature mining operation on the exemplary unsupervised device data distribution set to form an unsupervised data feature representation corresponding to the exemplary unsupervised device data distribution set;

[0072] Using the feature matching fusion subnetwork, perform a feature matching fusion operation on the supervised data feature representation and the unsupervised data feature representation to form a matched fusion data feature representation corresponding to the exemplary unsupervised device data distribution set;

[0073] Using the feature restoration subnetwork, perform a feature restoration operation on the unsupervised data feature representation and the matched fusion data feature representation to output a predicted distribution node selection identifier corresponding to the exemplary unsupervised device data distribution set;

[0074] According to the predicted distribution node selection identifier corresponding to the exemplary unsupervised device data distribution set and the distribution node selection identifier to be compared, the corresponding predicted learning cost value is calculated, and then based on the predicted learning cost value, the first feature mining sub-network, the second feature mining sub-network, the feature matching fusion sub-network and the feature restoration sub-network are network optimized to form a corresponding target data node screening neural network.

[0075] For example, in some embodiments, the first feature mining subnetwork includes at least one feature mining unit. Based on this, the step of using the first feature mining subnetwork to perform a feature mining operation on the exemplary supervisory device data distribution set and the data aggregation feature representation corresponding to the supervisory distribution node selection identifier to form a supervisory data feature representation corresponding to the exemplary supervisory device data distribution set may further include:

[0076] The at least one feature mining unit is used to perform a feature mining operation on the exemplary supervisory device data distribution set and the data aggregation feature representation corresponding to the supervisory distribution node selection identifier to form a supervisory data feature representation corresponding to the exemplary supervisory device data distribution set.

[0077] For example, in some embodiments, the first feature mining subnetwork includes a cascade of an input feature mining unit, an intermediate feature mining unit, and an output feature mining unit. Based on this, the step of using the at least one feature mining unit to perform a feature mining operation on the exemplary supervisory device data distribution set and the data aggregation feature representation corresponding to the supervisory distribution node selection identifier to form a supervisory data feature representation corresponding to the exemplary supervisory device data distribution set may further include:

[0078] Loading the data aggregation feature representation corresponding to the exemplary supervisory device data distribution set and the supervisory distribution node selection identifier, so as to perform an input feature extraction operation (i.e., extracting data according to a certain rule or ratio) using the input feature mining unit to form an input selected data feature representation corresponding to the exemplary supervisory device data distribution set;

[0079] Loading the input-end selected data feature representation to perform an intermediate feature extraction operation (as described above) using the intermediate feature mining unit to form an intermediate selected data feature representation corresponding to the exemplary supervisory device data distribution set;

[0080] The intermediate selected data feature representation is loaded to perform an output feature extraction operation using the output feature mining unit (as described above. In addition, each feature mining unit may include a filtering layer, a data mapping layer, and an activation function, and the data mapping layer may be used to map from 0 to 1) to form an output selected data feature representation as the supervisory data feature representation corresponding to the exemplary supervisory device data distribution set.

[0081] For example, in some embodiments, the second feature mining subnetwork includes a primary feature mining unit, an intermediate feature mining unit, a high-level feature mining unit, and a target number (e.g., 1, 2, 3, 4, 5, 6, 7, 8, 9, etc.) of feature aggregation units. Based on this, the step of using the second feature mining subnetwork to perform a feature mining operation on the exemplary unsupervised device data distribution set to form an unsupervised data feature representation corresponding to the exemplary unsupervised device data distribution set may further include:

[0082] The primary feature mining unit, the intermediate feature mining unit, the high-level feature mining unit, and the target number of feature aggregation units are used to perform a feature mining operation on the exemplary unsupervised device data distribution set to form an unsupervised data feature representation corresponding to the exemplary unsupervised device data distribution set.

[0083] For example, in some embodiments, the target number of feature aggregation units includes an input-end feature aggregation unit configured between the primary feature mining unit and the intermediate feature mining unit, an intermediate feature aggregation unit configured between the intermediate feature mining unit and the high-level feature mining unit, and an output-end feature aggregation unit configured at the output end of the high-level feature mining unit. Based on this, the step of performing a feature mining operation on the exemplary unsupervised device data distribution set using the primary feature mining unit, the intermediate feature mining unit, the high-level feature mining unit, and the target number of feature aggregation units to form an unsupervised data feature representation corresponding to the exemplary unsupervised device data distribution set may further include:

[0084] Loading the exemplary unsupervised device data distribution set to perform a primary feature extraction operation (as described above) using the primary feature mining unit to form a primary selected data feature representation corresponding to the exemplary unsupervised device data distribution set;

[0085] Loading the primary decimated data feature representation and the input-end decimated data feature representation to perform a primary feature representation aggregation operation (which may be used to aggregate, such as concatenate or superimpose, the feature representations) using the input-end feature aggregation unit to form an aggregated primary decimated data feature representation corresponding to the exemplary unsupervised device data distribution set;

[0086] Loading the aggregated primary selected data feature representation to perform a mid-level feature extraction operation (as described above) using the mid-level feature mining unit to form a mid-level selected data feature representation corresponding to the exemplary unsupervised device data distribution set;

[0087] Loading the intermediate decimated data feature representation and the intermediate decimated data feature representation to perform an intermediate feature representation aggregation operation (as described above) using the intermediate feature aggregation unit to form an aggregated intermediate decimated data feature representation corresponding to the exemplary unsupervised device data distribution set;

[0088] Loading the aggregated intermediate-level selected data feature representation to perform a high-level feature extraction operation (as described above) using the high-level feature mining unit to form a high-level selected data feature representation corresponding to the exemplary unsupervised device data distribution set;

[0089] The high-level abstracted data feature representation and the supervised data feature representation are loaded to perform a high-level feature representation aggregation operation (as described above) using the output-end feature aggregation unit to form a corresponding aggregated high-level abstracted data feature representation as the unsupervised data feature representation corresponding to the exemplary unsupervised device data distribution set.

[0090] For example, in some embodiments, the input-end feature aggregation unit includes a second feature fusion unit and a second feature filtering unit. Based on this, the step of loading the primary decimated data feature representation and the input-end decimated data feature representation to perform a primary feature representation aggregation operation using the input-end feature aggregation unit to form an aggregated primary decimated data feature representation corresponding to the exemplary unsupervised device data distribution set may further include:

[0091] The primary extracted data feature representation and the input-end extracted data feature representation are loaded to determine the corresponding extracted data fusion feature representation using the second feature fusion unit; and the extracted data fusion feature representation is loaded to determine the aggregated primary extracted data feature representation corresponding to the exemplary unsupervised device data distribution set using the second feature filtering unit (for specific implementation methods, please refer to the relevant description above).

[0092] For example, in some embodiments, the feature restoration subnetwork includes a cascaded input feature restoration unit, an intermediate feature restoration unit, an output feature restoration unit, and a third feature filtering unit. Based on this, the step of using the feature restoration subnetwork to perform a feature restoration operation on the unsupervised data feature representation and the matched fusion data feature representation to output a predicted distribution node selection identifier corresponding to the exemplary unsupervised device data distribution set may further include:

[0093] The unsupervised data feature representation and the matched fusion data feature representation are loaded to determine an initial data restoration feature representation using the input feature restoration unit (illustratively, the processing process of each feature restoration unit may be opposite to the processing process of the aforementioned feature mining unit, and may include an inverse filtering layer, a data mapping layer, and an activation function);

[0094] Loading the initial data restoration feature representation and the intermediate selected data feature representation to determine an intermediate data restoration feature representation using the intermediate feature restoration unit;

[0095] Loading the intermediate data restoration feature representation and the primary selected data feature representation to determine the output-end data restoration feature representation using the output-end feature restoration unit;

[0096] The output end data restoration feature representation is loaded to utilize the third feature filtering unit (exemplarily, the third feature filtering unit may include a filtering layer and an activation function, the activation function may be a nonlinear activation, and the obtained predicted distribution node selection identifier includes a selection probability value corresponding to each data distribution node, such as 0 or 1, or 0 to 1) to determine the predicted distribution node selection identifier corresponding to the exemplary unsupervised device data distribution set.

[0097] For example, in some embodiments, the step of calculating the corresponding predicted learning cost based on the predicted distribution node selection identifier corresponding to the exemplary unsupervised device data distribution set and the to-be-compared distribution node selection identifier may further include:

[0098] Based on the predicted distribution node sampling identifier corresponding to the exemplary unsupervised device data distribution set and the distribution node sampling identifier to be compared, the corresponding first dimension predicted learning cost value and the second dimension predicted learning cost value are analyzed and output; and, based on the first dimension predicted learning cost value and the second dimension predicted learning cost value, the corresponding predicted learning cost value is calculated and output (exemplarily, based on any two learning cost value calculation functions, i.e., the loss calculation function, two learning cost values can be calculated for the predicted distribution node sampling identifier corresponding to the exemplary unsupervised device data distribution set and the distribution node sampling identifier to be compared, and then, the two learning cost values can be weighted averaged or averaged to obtain the predicted learning cost value).

[0099] For example, in some embodiments, the feature matching fusion subnetwork includes a feature matching unit, a first feature fusion unit, and a first feature filtering unit. Based on this, the step of using the feature matching fusion subnetwork to perform a feature matching fusion operation on the supervised data feature representation and the unsupervised data feature representation to form a matching fused data feature representation corresponding to the exemplary unsupervised device data distribution set may further include:

[0100] Loading the supervised data feature representation and the unsupervised data feature representation to determine, using the feature matching unit, a matching data feature representation between the supervised data feature representation and the unsupervised data feature representation (exemplarily, a cross product may be performed on the supervised data feature representation and the unsupervised data feature representation to determine the matching data feature representation between the supervised data feature representation and the unsupervised data feature representation);

[0101] Loading the matching data feature representation and the unsupervised data feature representation to determine an initial data fusion feature representation using the first feature fusion unit (exemplarily, the first feature fusion unit can be used for splicing or superposition, etc.);

[0102] The initial data fusion feature representation is loaded to utilize the first feature filtering unit (the first feature filtering unit includes a convolution kernel to perform a convolution operation on the initial data fusion feature representation to achieve feature filtering), analyze and output the matching fusion data feature representation corresponding to the exemplary unsupervised device data distribution set.

[0103] For example, in some embodiments, step S130 may further include:

[0104] In the first device data distribution set, a first device data distribution subset corresponding to the first device data distribution set is filtered out to form a second device data distribution subset corresponding to the first device data distribution set;

[0105] Adjusting the distribution position relationship between every two data distribution nodes in the second device data distribution subset to form a new second device data distribution subset;

[0106] The target data node screening neural network is used to select the new second device data distribution subset (refer to the relevant description above) to form a third device data distribution subset corresponding to the new second device data distribution subset, and the IoT devices to be matched corresponding to each data distribution node included in the third device data distribution subset are combined to form a second set of IoT devices to be matched corresponding to the third device data distribution subset.

[0107] For example, in some embodiments, the step of adjusting the distribution position relationship between every two data distribution nodes in the second device data distribution subset to form a new second device data distribution subset may further include:

[0108] Calculating a ratio between the number of data distribution nodes corresponding to the first device data distribution subset and the number of data distribution nodes corresponding to the first device data distribution set (exemplarily, the ratio may be the former divided by the latter);

[0109] Based on the result of the ratio calculation, the distribution position relationship between every two data distribution nodes in the second device data distribution subset is shortened by a corresponding proportion to adjust the distribution position relationship (for example, the position vector corresponding to the distribution position relationship can be weighted based on the result of the ratio calculation to obtain a weighted position vector corresponding to the distribution position relationship, that is, a new distribution position relationship), forming a new second device data distribution subset.

[0110] For example, in some implementations, step S130 may further include:

[0111] In the new second device data distribution subset, a third device data distribution subset corresponding to the new second device data distribution subset is filtered out to form a fourth device data distribution subset corresponding to the new second device data distribution subset;

[0112] The IoT devices to be matched corresponding to each data distribution node included in the fourth device data distribution subset are combined to form a third set of IoT devices to be matched corresponding to the fourth device data distribution subset; or, the distribution position relationship between every two data distribution nodes in the fourth device data distribution subset is adjusted to form a new fourth device data distribution subset; the new fourth device data distribution subset is selected using the target data node screening neural network to form a fifth device data distribution subset corresponding to the new fourth device data distribution subset, and the IoT devices to be matched corresponding to each data distribution node included in the fifth device data distribution subset are combined to form a third set of IoT devices to be matched corresponding to the fifth device data distribution subset, and every two IoT devices to be matched included in the third set of IoT devices to be matched have a matching relationship.

[0113] Combine Figure 3 The embodiment of the present invention further provides a device matching system for IoT devices, which can be applied to the above-mentioned device matching processing platform. The device matching system for IoT devices includes:

[0114] a device data distribution set construction module, configured to extract and process historical data from a plurality of IoT devices to be matched, to obtain historical IoT device operation data corresponding to each IoT device to be matched, and then construct a corresponding first device data distribution set based on the historical IoT device operation data corresponding to each IoT device to be matched and the device deployment location information corresponding to each IoT device to be matched, wherein each data distribution node in the first device data distribution set corresponds to a IoT device to be matched, and the distribution position relationship between each two data distribution nodes is correlated with the device deployment location information corresponding to the two IoT devices to be matched;

[0115] a first device matching and combining module, configured to select the first device data distribution set using a target data node screening neural network to form a first device data distribution subset corresponding to the first device data distribution set, and to combine the to-be-matched IoT devices corresponding to each data distribution node included in the first device data distribution subset to form a first to-be-matched IoT device set, wherein each two to-be-matched IoT devices included in the first to-be-matched IoT device set have a matching relationship;

[0116] The second device matching and combination module is used to screen out the first device data distribution subset corresponding to the first device data distribution set in the first device data distribution set to form a second device data distribution subset corresponding to the first device data distribution set, and then use the target data node screening neural network to select the second device data distribution subset to form a third device data distribution subset corresponding to the second device data distribution subset, and combine the IoT devices to be matched corresponding to each data distribution node included in the third device data distribution subset to form a second set of IoT devices to be matched, where there is a matching relationship between every two IoT devices to be matched included in the second set of IoT devices to be matched.

[0117] In summary, the present invention provides a device matching method and system for IoT devices, which constructs a corresponding first device data distribution set based on the historical IoT device operation data corresponding to each IoT device to be matched and the device deployment location information corresponding to each IoT device to be matched; uses a target data node screening neural network to select the first device data distribution set to form a first device data distribution subset corresponding to the first device data distribution set, and combines the IoT devices to be matched corresponding to each data distribution node included in the first device data distribution subset to form a first set of IoT devices to be matched; and, in the first device data distribution set, the first device data distribution subset corresponding to the first device data distribution set is screened out to form a second device data distribution subset corresponding to the first device data distribution set; then uses the target data node screening neural network to select the second device data distribution subset to form a third device data distribution subset corresponding to the second device data distribution subset, and combines the IoT devices to be matched corresponding to each data distribution node included in the third device data distribution subset to form a second set of IoT devices to be matched. Based on the aforementioned steps, a neural network can be used to comprehensively analyze and process the historical IoT device operation data and device deployment location information corresponding to the IoT device to be matched, so that the reliability of data processing is higher, and the reliability of device matching of IoT devices can be improved to a certain extent.

[0118] The foregoing description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. Those skilled in the art will readily appreciate that various modifications and variations of the present invention are possible. Any modifications, equivalent substitutions, or improvements made within the spirit and principles of the present invention are intended to be within the scope of protection of the present invention.

Claims

1. A device matching method for an Internet of Things device, characterized in that: include: Performing historical data extraction and processing on a plurality of IoT devices to be matched to obtain historical IoT device operation data corresponding to each of the IoT devices to be matched, and then constructing a corresponding first device data distribution set based on the historical IoT device operation data corresponding to each of the IoT devices to be matched and the device deployment location information corresponding to each of the IoT devices to be matched, wherein each data distribution node in the first device data distribution set corresponds to a IoT device to be matched, and a distribution position relationship between every two data distribution nodes is correlated with the device deployment location information corresponding to the two IoT devices to be matched; Using a target data node screening neural network, the first device data distribution set is selected to form a first device data distribution subset corresponding to the first device data distribution set, and the to-be-matched IoT devices corresponding to each data distribution node included in the first device data distribution subset are combined to form a first to-be-matched IoT device set, where every two to-be-matched IoT devices included in the first to-be-matched IoT device set have a matching relationship; In the first device data distribution set, the first device data distribution subset corresponding to the first device data distribution set is screened to form a second device data distribution subset corresponding to the first device data distribution set, and then the target data node is used to screen the neural network to select the second device data distribution subset to form a third device data distribution subset corresponding to the second device data distribution subset, and the IoT devices to be matched corresponding to each data distribution node included in the third device data distribution subset are combined to form a second set of IoT devices to be matched, and there is a matching relationship between every two IoT devices to be matched included in the second set of IoT devices to be matched.

2. The device matching method of the Internet of Things device according to claim 1, characterized in that: The target data node screening neural network is formed based on network optimization of the initial data node screening neural network, and the network optimization process of the initial data node screening neural network includes: Extracting an exemplary supervised device data distribution set, an exemplary unsupervised device data distribution set, a supervised distribution node selection identifier corresponding to the exemplary supervised device data distribution set, a to-be-compared distribution node selection identifier corresponding to the exemplary unsupervised device data distribution set, and a pre-built initial data node screening neural network, wherein the initial data node screening neural network includes a first feature mining subnetwork, a second feature mining subnetwork, a feature matching fusion subnetwork, and a feature restoration subnetwork; Using the first feature mining sub-network, a feature mining operation is performed on the exemplary supervisory device data distribution set and the data aggregation feature representation corresponding to the supervisory distribution node selection identifier to form a supervisory data feature representation corresponding to the exemplary supervisory device data distribution set; Using the second feature mining sub-network, perform a feature mining operation on the exemplary unsupervised device data distribution set to form an unsupervised data feature representation corresponding to the exemplary unsupervised device data distribution set; Using the feature matching fusion subnetwork, perform a feature matching fusion operation on the supervised data feature representation and the unsupervised data feature representation to form a matched fusion data feature representation corresponding to the exemplary unsupervised device data distribution set; Using the feature restoration subnetwork, perform a feature restoration operation on the unsupervised data feature representation and the matched fusion data feature representation to output a predicted distribution node selection identifier corresponding to the exemplary unsupervised device data distribution set; According to the predicted distribution node selection identifier corresponding to the exemplary unsupervised device data distribution set and the distribution node selection identifier to be compared, the corresponding predicted learning cost value is calculated, and then based on the predicted learning cost value, the first feature mining sub-network, the second feature mining sub-network, the feature matching fusion sub-network and the feature restoration sub-network are network optimized to form a corresponding target data node screening neural network.

3. The device matching method of the Internet of Things device according to claim 2, characterized in that: The first feature mining subnetwork includes at least one feature mining unit; the step of using the first feature mining subnetwork to perform a feature mining operation on the exemplary supervisory device data distribution set and the data aggregation feature representation corresponding to the supervisory distribution node selection identifier to form a supervisory data feature representation corresponding to the exemplary supervisory device data distribution set includes: The at least one feature mining unit is used to perform a feature mining operation on the exemplary supervisory device data distribution set and the data aggregation feature representation corresponding to the supervisory distribution node selection identifier to form a supervisory data feature representation corresponding to the exemplary supervisory device data distribution set.

4. The device matching method for an Internet of Things device according to claim 3, wherein: The first feature mining sub-network includes a cascaded input feature mining unit, an intermediate feature mining unit and an output feature mining unit; The step of using the at least one feature mining unit to perform a feature mining operation on the exemplary supervisory device data distribution set and the data aggregation feature representation corresponding to the supervisory distribution node selection identifier to form a supervisory data feature representation corresponding to the exemplary supervisory device data distribution set includes: Loading the exemplary supervisory device data distribution set and the data aggregation feature representation corresponding to the supervisory distribution node selection identifier, and performing an input feature extraction operation using the input feature mining unit to form an input selected data feature representation corresponding to the exemplary supervisory device data distribution set; Loading the input-end selected data feature representation to perform an intermediate feature extraction operation using the intermediate feature mining unit to form an intermediate selected data feature representation corresponding to the exemplary supervisory device data distribution set; The intermediate selected data feature representation is loaded to perform an output feature extraction operation using the output feature mining unit to form an output selected data feature representation as a supervisory data feature representation corresponding to the exemplary supervisory device data distribution set.

5. The device matching method for an Internet of Things device according to claim 4, characterized in that: The second feature mining sub-network includes a primary feature mining unit, an intermediate feature mining unit, an advanced feature mining unit and a target number of feature aggregation units; The step of using the second feature mining sub-network to perform a feature mining operation on the exemplary unsupervised device data distribution set to form an unsupervised data feature representation corresponding to the exemplary unsupervised device data distribution set includes: The primary feature mining unit, the intermediate feature mining unit, the high-level feature mining unit, and the target number of feature aggregation units are used to perform a feature mining operation on the exemplary unsupervised device data distribution set to form an unsupervised data feature representation corresponding to the exemplary unsupervised device data distribution set.

6. The device matching method of the Internet of Things device according to claim 5, characterized in that: The target number of feature aggregation units includes an input feature aggregation unit configured between the primary feature mining unit and the intermediate feature mining unit, an intermediate feature aggregation unit configured between the intermediate feature mining unit and the high-level feature mining unit, and an output feature aggregation unit configured at the output end of the high-level feature mining unit; The step of performing a feature mining operation on the exemplary unsupervised device data distribution set using the primary feature mining unit, the intermediate feature mining unit, the high-level feature mining unit, and the target number of feature aggregation units to form an unsupervised data feature representation corresponding to the exemplary unsupervised device data distribution set includes: Loading the exemplary unsupervised device data distribution set to perform a primary feature extraction operation using the primary feature mining unit to form a primary extracted data feature representation corresponding to the exemplary unsupervised device data distribution set; Loading the primary decimated data feature representation and the input-end decimated data feature representation to perform a primary feature representation aggregation operation using the input-end feature aggregation unit to form an aggregated primary decimated data feature representation corresponding to the exemplary unsupervised device data distribution set; Loading the aggregated primary selected data feature representation to perform an intermediate feature extraction operation using the intermediate feature mining unit to form an intermediate selected data feature representation corresponding to the exemplary unsupervised device data distribution set; Loading the intermediate decimated data feature representation and the intermediate decimated data feature representation to perform an intermediate feature representation aggregation operation using the intermediate feature aggregation unit to form an aggregated intermediate decimated data feature representation corresponding to the exemplary unsupervised device data distribution set; Loading the aggregated intermediate-level selected data feature representation to perform a high-level feature extraction operation using the high-level feature mining unit to form a high-level selected data feature representation corresponding to the exemplary unsupervised device data distribution set; The high-level abstracted data feature representation and the supervised data feature representation are loaded to perform a high-level feature representation aggregation operation using the output-end feature aggregation unit to form a corresponding aggregated high-level abstracted data feature representation as an unsupervised data feature representation corresponding to the exemplary unsupervised device data distribution set.

7. The device matching method for an Internet of Things device according to any one of claims 1 to 6, characterized in that: The step of screening out a first device data distribution subset corresponding to the first device data distribution set from the first device data distribution set to form a second device data distribution subset corresponding to the first device data distribution set, then selecting the second device data distribution subset using the target data node screening neural network to form a third device data distribution subset corresponding to the second device data distribution subset, and combining the to-be-matched IoT devices corresponding to each data distribution node included in the third device data distribution subset to form a second to-be-matched IoT device set includes: In the first device data distribution set, a first device data distribution subset corresponding to the first device data distribution set is filtered out to form a second device data distribution subset corresponding to the first device data distribution set; Adjusting the distribution position relationship between every two data distribution nodes in the second device data distribution subset to form a new second device data distribution subset; The target data node screening neural network is used to select the new second device data distribution subset to form a third device data distribution subset corresponding to the new second device data distribution subset, and the IoT devices to be matched corresponding to each data distribution node included in the third device data distribution subset are combined to form a second set of IoT devices to be matched corresponding to the third device data distribution subset.

8. The device matching method for an Internet of Things device according to claim 7, wherein: The step of adjusting the distribution position relationship between every two data distribution nodes in the second device data distribution subset to form a new second device data distribution subset includes: Calculating a ratio between the number of data distribution nodes corresponding to the first device data distribution subset and the number of data distribution nodes corresponding to the first device data distribution set; According to the result of the ratio calculation, the distribution position relationship between every two data distribution nodes in the second device data distribution subset is shortened by a corresponding proportion to adjust the distribution position relationship and form a new second device data distribution subset.

9. The device matching method for an Internet of Things device according to claim 7, wherein: The step of screening out a first device data distribution subset corresponding to the first device data distribution set from the first device data distribution set to form a second device data distribution subset corresponding to the first device data distribution set, then selecting the second device data distribution subset using the target data node screening neural network to form a third device data distribution subset corresponding to the second device data distribution subset, and combining the to-be-matched IoT devices corresponding to each data distribution node included in the third device data distribution subset to form a second to-be-matched IoT device set, further includes: In the new second device data distribution subset, a third device data distribution subset corresponding to the new second device data distribution subset is filtered out to form a fourth device data distribution subset corresponding to the new second device data distribution subset; The IoT devices to be matched corresponding to each data distribution node included in the fourth device data distribution subset are combined to form a third set of IoT devices to be matched corresponding to the fourth device data distribution subset; or, the distribution position relationship between every two data distribution nodes in the fourth device data distribution subset is adjusted to form a new fourth device data distribution subset; the new fourth device data distribution subset is selected using the target data node screening neural network to form a fifth device data distribution subset corresponding to the new fourth device data distribution subset, and the IoT devices to be matched corresponding to each data distribution node included in the fifth device data distribution subset are combined to form a third set of IoT devices to be matched corresponding to the fifth device data distribution subset, and every two IoT devices to be matched included in the third set of IoT devices to be matched have a matching relationship.

10. A device matching processing platform, characterized in that: include: a device data distribution set construction module, configured to extract and process historical data from a plurality of IoT devices to be matched, to obtain historical IoT device operation data corresponding to each IoT device to be matched, and then construct a corresponding first device data distribution set based on the historical IoT device operation data corresponding to each IoT device to be matched and the device deployment location information corresponding to each IoT device to be matched, wherein each data distribution node in the first device data distribution set corresponds to a IoT device to be matched, and the distribution position relationship between each two data distribution nodes is correlated with the device deployment location information corresponding to the two IoT devices to be matched; a first device matching and combining module, configured to select the first device data distribution set using a target data node screening neural network to form a first device data distribution subset corresponding to the first device data distribution set, and to combine the to-be-matched IoT devices corresponding to each data distribution node included in the first device data distribution subset to form a first to-be-matched IoT device set, wherein each two to-be-matched IoT devices included in the first to-be-matched IoT device set have a matching relationship; The second device matching and combination module is used to screen out the first device data distribution subset corresponding to the first device data distribution set in the first device data distribution set to form a second device data distribution subset corresponding to the first device data distribution set, and then use the target data node screening neural network to select the second device data distribution subset to form a third device data distribution subset corresponding to the second device data distribution subset, and combine the IoT devices to be matched corresponding to each data distribution node included in the third device data distribution subset to form a second set of IoT devices to be matched, where there is a matching relationship between every two IoT devices to be matched included in the second set of IoT devices to be matched.

Images