An information query method and system based on the Internet of Things

Abstract

This invention provides an information query method and system based on the Internet of Things (IoT), belonging to the field of IoT information query. The method includes: S1, receiving multimodal query input and converting it into a structured query representation; S2, using an improved data credibility assessment algorithm to evaluate the credibility of IoT data and obtain a credibility score for each data item; S3, calculating the complexity of the structured query representation based on a nonlinear evaluation model to obtain a query complexity score; S4, determining a query execution strategy based on the query complexity score and the status of IoT resources; S5, executing the information query according to the query execution strategy to obtain initial query results; and S6, fusing the query results based on the credibility score to obtain the final information query result. This invention achieves adaptive processing and quality optimization of information query in the IoT environment, and can dynamically select the optimal execution strategy based on query complexity and resource status.

Description

Technical Field

[0001] This invention relates to the field of Internet of Things (IoT) information query technology, and in particular to an information query method and system based on IoT. Background Technology

[0002] The Internet of Things (IoT), as a crucial component of the next generation of information technology, is based on the core concept of connecting any object to the internet through various information sensing devices, such as RFID devices, infrared sensors, GPS, and laser scanners, according to agreed-upon protocols. This enables information exchange and communication, achieving intelligent identification, location, tracking, monitoring, and management. IoT-based information retrieval refers to the process of extracting, processing, and presenting valuable information based on user needs from massive, distributed, and heterogeneous physical world sensor data using the IoT architecture. This type of retrieval differs from traditional internet information retrieval, as its data source possesses stronger real-time characteristics, dynamism, and a direct correlation with the physical world. With the widespread adoption of IoT applications, efficiently, accurately, and intelligently extracting necessary information from IoT big data has become a key technological challenge.

[0003] Existing IoT information query technologies primarily employ a centralized cloud processing architecture. User query requests are transmitted over the network to a cloud server, which retrieves relevant information from its database and returns the results. To improve efficiency, some technologies utilize data indexing, caching mechanisms, and distributed storage. Regarding query understanding, most technologies only support keyword matching and simple structured query languages, offering limited capabilities to understand complex query intents. In terms of data security, existing technologies mainly rely on transmission encryption and access control, with relatively simple authentication of data sources and assessment of the credibility of data content.

[0004] Chinese invention patent application number 202411081388.2 discloses an IoT data processing method, system and computer-readable storage medium. Based on a data credibility assessment algorithm component library, it realizes IoT data content analysis, identity recognition, abnormal information blocking and improves IoT information security protection capabilities. However, it does not take into account the complexity of multimodal query input in the modern IoT environment and can only process structured data queries. It cannot meet the increasingly complex intelligent query and analysis needs of modern IoT applications. Summary of the Invention

[0005] This invention provides an information query method and system based on the Internet of Things (IoT) to address the shortcomings of existing technologies that cannot perform multimodal queries, and to achieve adaptive processing and quality optimization of information queries in the IoT environment.

[0006] This invention provides an information query method based on the Internet of Things, comprising: S1. Receive multimodal query input and convert the multimodal query input into a structured query representation; S2. An improved data credibility assessment algorithm is used to assess the credibility of IoT data and obtain a credibility score for each data item. S3. Calculate the complexity of the structured query representation based on a nonlinear evaluation model to obtain a query complexity score; S4. Determine the query execution strategy based on the query complexity score and the IoT resource status; S5. Perform an information query according to the query execution strategy to obtain the initial query results; S6. The query results are fused based on the credibility score to obtain the final information query results.

[0007] According to the present invention, an information query method based on the Internet of Things is provided, wherein the structured query representation adopts a five-tuple structure, including query subject, time constraint, space constraint, context information and data credibility requirements.

[0008] According to the information query method based on the Internet of Things provided by the present invention, the improved data credibility evaluation algorithm includes a feature extraction module, a feature learning module, a scoring module, an information verification module, and an analysis module.

[0009] According to the information query method based on the Internet of Things provided by the present invention, step S2 specifically includes: The feature extraction module is used to extract features from IoT data to obtain device hardware fingerprint features, network behavior features, and data content features; The device hardware fingerprint features, network behavior features, and data content features are reconstructed by using a feature learning module to encode and decode feature vectors, thereby obtaining the data potential representation and reconstruction error. The scoring module calculates an initial credibility score based on the data's latent representation and reconstruction error; Obtain external validation information and use the information validation module to calculate the cross-validation strength; Acquire historical IoT data and use the analysis module to analyze the historical data to obtain a time series consistency score; The credibility score is calculated based on the initial credibility score, cross-validation strength, and temporal consistency score.

[0010] According to the information query method based on the Internet of Things provided by the present invention, the formula for calculating the credibility score is as follows: ; in, Represents data items Credibility score, This represents the i-th IoT data item to be evaluated. Index representing the evaluation dimension This represents the total number of evaluation dimensions. This represents the weight coefficient of the j-th evaluation dimension. Represents data items The base credibility score on the j-th evaluation dimension, This represents the sensitivity parameter for cross-validation. Represents data items Cross-validation strength, This represents the threshold parameter for cross-validation. This represents the normalized information entropy function.

[0011] According to the information query method based on the Internet of Things provided by the present invention, the nonlinear evaluation model includes a query parsing module, a complexity calculation module, an analysis module, and a comprehensive scoring module, wherein... The query parsing module is used to parse the structured query representation and extract basic factors, wherein the basic factors include the number of entities, relation complexity, time span, and spatial range; The complexity calculation module is used to perform nonlinear transformations on the basic factors and calculate the basic complexity. The analysis module is used to construct a query semantic network and calculate the mutual information and degree centrality of semantic relations based on the query semantic network. The comprehensive scoring module is used to fuse the basic complexity and semantic network complexity to obtain the final complexity score.

[0012] According to the present invention, an information query method based on the Internet of Things (IoT) includes a query execution strategy comprising a local processing strategy, an edge collaborative processing strategy, and a cloud processing strategy. The local processing strategy is used to complete query processing within a single edge node; The edge collaborative processing strategy is used to collaboratively complete query processing across multiple edge nodes; The cloud processing strategy is used to report queries to the cloud service layer for processing.

[0013] According to the information query method based on the Internet of Things provided by the present invention, step S4 specifically includes: The computing resource status is obtained based on the CPU utilization, memory usage, remaining storage space, and network bandwidth usage of local edge nodes. The availability, completeness, and timeliness of the data are evaluated locally through structured queries. Based on the processing capabilities of local edge nodes and historical query statistics, a first complexity threshold and a second complexity threshold are set, where the first complexity threshold corresponds to the upper limit of queries that local resources can process independently, and the second complexity threshold corresponds to the upper limit of queries that edge collaborative processing can handle. When the query complexity score is lower than the first complexity threshold and the local data completeness meets the query requirements, the local processing strategy is selected. When the complexity score is greater than the first complexity threshold but less than the second complexity threshold, or when the local data is incomplete, the preset neighbor node status table is queried to obtain the processing capabilities, data types, and network connection quality information of the surrounding edge nodes, and the edge collaborative processing strategy is executed. When the complexity score is higher than the second complexity threshold or the combination of adjacent edge nodes still cannot meet the query requirements, the cloud processing strategy is selected.

[0014] According to the information query method based on the Internet of Things provided by the present invention, step S6 specifically includes: The initial query results are deduplicated to obtain a set of deduplicated data items; The data in the deduplicated data item set is matched with the structured query representation to obtain a relevance score for each data item; Calculate the similarity and difference between data items in the deduplicated data item set to obtain the information evaluation result of each data item; Based on the credibility score, data integrity index, and measurement accuracy index of each data item, a weighting algorithm is used to calculate the data quality weight of each data item. The timeliness discount factor for each data item is calculated based on the data generation time and query time. The contribution of each data item to the query result is calculated based on the relevance score, information evaluation result, data quality weight and timeliness discount factor. Based on the credibility score and contribution of each data item, a weighted fusion process is used to obtain the final information query results.

[0015] This invention also provides an information query system based on the Internet of Things, implementing the information query method described above, including: The query input module is used to receive multimodal query input and convert the multimodal query input into a structured query representation; The credibility assessment module is used to assess the credibility of IoT data using an improved data credibility assessment algorithm to obtain a credibility score for each data item. The complexity calculation module is used to calculate the complexity of the structured query representation based on a nonlinear evaluation model to obtain a query complexity score. The execution strategy determination module is used to determine the query execution strategy based on the query complexity score and the IoT resource status. The information query module is used to perform information queries according to the query execution strategy and obtain initial query results; The fusion module is used to fuse the query results based on the credibility score to obtain the final information query results.

[0016] This invention provides an information query method and system based on the Internet of Things (IoT). Through multimodal query input, improved data credibility assessment algorithm, nonlinear complexity assessment model, intelligent execution strategy decision-making, and credibility fusion processing, it achieves adaptive processing and quality optimization of information query in the IoT environment. It can dynamically select the optimal execution strategy according to query complexity and resource status, and switch between local processing, edge collaborative processing, and cloud processing. It effectively balances the relationship between query response time, resource utilization efficiency, and result quality, and significantly improves the overall performance and user experience of IoT information query. Attached Figure Description

[0017] To more clearly illustrate the technical solutions in this invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0018] Figure 1 This is a flowchart illustrating the information query method based on the Internet of Things provided by this invention. Detailed Implementation

[0019] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this invention. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without creative effort are within the scope of protection of this invention.

[0020] like Figure 1 As shown, the present invention provides an information query method based on the Internet of Things, comprising the following steps: S1. Receive multimodal query input and convert the multimodal query input into a structured query representation, wherein the structured query representation adopts a five-tuple structure, including query subject, time constraint, space constraint, context information and data credibility requirements.

[0021] Understandably, when a user initiates a query request through a terminal device, the multimodal query understanding module first classifies, identifies, and preprocesses different types of input.

[0022] In one embodiment of the present invention, for voice input, the user's voice signal is acquired through a microphone array, and preprocessing operations including noise reduction, echo cancellation and volume normalization are performed to ensure the audio quality of subsequent processing. An automatic speech recognition technology based on deep neural networks is adopted to convert speech signals into corresponding text sequences. The deep neural network is trained using the Transformer architecture combined with the Connection Temporal Classification (CTC) loss function.

[0023] In one embodiment of the present invention, for image input processing, image data captured by the user through a camera is received, and image preprocessing is performed on the image data, including operations such as size adjustment, brightness and contrast correction and noise filtering, to ensure that the image quality meets the requirements of subsequent analysis. A convolutional neural network-based target detection algorithm is used to identify various IoT devices, sensors, or related physical objects in images. The convolutional neural network-based target detection algorithm uses a pre-trained network model that has been fine-tuned on an IoT device dataset. The system extracts and classifies the features of the detected targets, identifies key information such as device type, status indicator color, and instrument reading, and converts the extracted visual features into corresponding semantic labels.

[0024] In one embodiment of the present invention, for gesture input, the three-dimensional position and posture information of the user's hand is captured by a depth camera or an infrared sensor; A three-dimensional coordinate sequence of key hand points is constructed, and a gesture sequence analysis algorithm based on a long short-term memory network is used to identify the types of gestures performed by the user, including basic operations such as pointing, selecting, zooming, and swiping. Based on the recognized gesture type, the corresponding parameter information is extracted, such as the direction vector of the pointing gesture, the area coordinates of the circled gesture, and the scaling factor of the zoom gesture. The gesture information is then spatially mapped to the current display interface to determine the specific object or area that the user is pointing at or operating on, and the gesture recognition result is converted into a query operation command.

[0025] In one embodiment of the present invention, the quintuple structure is as follows: ,in, It indicates the subject of the query, namely a specific device type, data attribute, or physical phenomenon; It represents time constraints, defining the time range and granularity of the query, including specific time points, time periods, or relative time expressions; This indicates spatial constraints, specifying the geographic location or region range for the query; It represents contextual information, including background information such as user identity, environment status, and device status; This indicates the data credibility requirement, specifying the minimum credibility threshold required for the query results.

[0026] Understandably, structured query representation can uniformly handle complex query requirements from different modalities, effectively support users in expressing their query requirements through various natural interaction methods, and at the same time ensure the accuracy of query understanding and the efficiency of processing.

[0027] This invention addresses the limitation of traditional IoT query systems that only support a single input mode by converting multimodal inputs into standardized 5-tuple structured query representations, thereby improving the naturalness and convenience of human-computer interaction and enhancing the efficiency of IoT information query functions.

[0028] S2. An improved data credibility assessment algorithm is used to assess the credibility of IoT data and obtain a credibility score for each data item. The improved data credibility assessment algorithm includes a feature extraction module, a feature learning module, a scoring module, an information verification module, and an analysis module.

[0029] The Internet of Things (IoT) data includes sensor data, device status data, and network communication data. Sensor data includes environmental monitoring data and location data; device status data includes device identification information, device operating status, and device configuration information; and network communication data includes communication protocol data, network performance data, and data traffic statistics.

[0030] Specifically, step S2 includes: The feature extraction module is used to extract features from IoT data to obtain device hardware fingerprint features, network behavior features, and data content features; The device hardware fingerprint features, network behavior features, and data content features are reconstructed by using a feature learning module to encode and decode feature vectors, thereby obtaining the data potential representation and reconstruction error. The scoring module calculates an initial credibility score based on the latent representation and reconstruction error of the data; wherein the calculation method for the initial credibility score is as follows: Based on the error value of the reconstruction error, calculate the overall error rate and feature similarity between the device hardware fingerprint features, network behavior features, and data content features and the reconstructed device hardware fingerprint features, network behavior features, and data content features; The initial confidence level is obtained based on the error rate and feature similarity, i.e., initial confidence level = (1 - error rate) × similarity × 100; External validation information is obtained, and the cross-validation strength is calculated using the information validation module. The calculation method is as follows: Obtain external verification information and identify key data items; Set a reasonable range and an error tolerance range of ±10% for each key data item, calculate the numerical deviation between the current data and the external verification information provided by each verification data source, and calculate the ratio of the number of verification sources with less than 10% to the total number of verification sources to obtain the consistency ratio. Calculate the verification strength score based on the consistency ratio, i.e., verification strength score = consistency ratio × number of verification sources × 20. Historical IoT data is acquired and analyzed using an analytics module to obtain a time-series consistency score. The calculation method for time-series consistency is as follows: Obtain IoT data for the same time period over the past 30 days, and calculate the difference between the historical average and the historical standard deviation; Set the upper and lower limits of the normal range based on the historical average and historical standard deviation; Compare the current data value with the upper and lower limits of the normal range to determine the degree of deviation. If the current value is within the normal range: Timing consistency score = 100 points If the deviation from the normal range is within 20%: Timing consistency score = 80 points If the deviation from the normal range is 20%-50%: Timing consistency score = 50 points If the deviation from the normal range exceeds 50%: Timing consistency score = 0 points The credibility score is calculated based on the initial credibility score, cross-validation strength, and temporal consistency score.

[0031] In one embodiment of the present invention, a weighted average method is used to calculate the credibility, with the initial credibility score having a weight coefficient of 0.4, the cross-validation strength having a weight coefficient of 0.35, and the temporal consistency score having a weight coefficient of 0.25.

[0032] This invention improves the data credibility assessment algorithm, enabling multi-dimensional credibility assessment of IoT data and significantly enhancing the accuracy and comprehensiveness of credibility assessment.

[0033] In one embodiment of the present invention, the processing steps of the feature extraction module are as follows: Preprocessing operations are performed on the input IoT data, including basic operations such as data format standardization, outlier detection, and missing value handling, to ensure that the data quality meets the requirements for feature extraction; By analyzing the device's unique identifiers such as MAC address, CPU serial number, hardware version information, and firmware version number, a digital fingerprint feature vector of the device is constructed to obtain the device's hardware fingerprint features. The network communication mode of the monitoring equipment includes key indicators such as data transmission frequency, communication protocol usage, and timing characteristics of network connection establishment and disconnection. It also analyzes network-level behavioral characteristics such as data packet size distribution, transmission time interval statistics, and network latency change trends to identify the normal behavior baseline and abnormal behavior patterns of the equipment and obtain network behavior characteristics. Statistical analysis is performed on sensor data to calculate statistical indicators such as data distribution characteristics, coefficient of variation, and temporal stability. By comparing and analyzing the data with historical data and data from similar devices, the rationality and consistency of the data are identified, and the characteristics of the data content are obtained.

[0034] In one embodiment of the present invention, the feature learning module receives multidimensional feature data output by the feature extraction module and performs feature encoding learning and representation optimization through a deep neural network architecture. The feature learning module employs an autoencoder network structure, including an encoder and a decoder. The encoder is responsible for mapping high-dimensional original features to a low-dimensional latent space, learning the compact representation and intrinsic structure of the data. The decoder attempts to reconstruct the original features from the latent representation, and the quality of feature learning is evaluated through the reconstruction error. During the encoding process, the system uses a multilayer perceptron or convolutional neural network structure, selecting an appropriate network architecture based on the type and dimension of the input features. Each layer introduces expressive power through a non-linear activation function, enabling the network to learn complex feature relationships and patterns. The decoder then reconstructs the original features from the latent representation; the accuracy of the reconstruction process reflects the quality and completeness of the latent representation.

[0035] Specifically, the feature learning module converts the features of MAC address, device model, transmission frequency, and sensor data into numerical values, performs feature standardization to obtain standardized feature values, arranges all standardized feature values ​​in order to form feature vectors, uses an encoder to encode the feature vectors to obtain latent representations, and then uses a decoder to decode the latent representations to obtain reconstructed features. The reconstruction accuracy is then evaluated based on the original features and the reconstructed features.

[0036] In one embodiment of the present invention, the scoring module adopts a multi-factor fusion scoring strategy, which comprehensively considers multiple dimensions of indicators such as the intrinsic quality of the data, the reliability of the source, and the timeliness. First, it calculates the basic credibility score based on the quality and stability of the latent representation, and then adjusts it according to the size of the reconstruction error. Data items with smaller reconstruction errors receive higher credibility scores, while data items with larger reconstruction errors are given lower scores or marked as suspicious data.

[0037] In one embodiment of the present invention, the information verification module is responsible for acquiring and processing external verification information, providing an independent verification data source for credibility assessment.

[0038] In one embodiment of the present invention, the analysis module maintains historical data archives for each device, recording the changing trends of the device's data characteristics, behavior patterns, and performance indicators at different time periods.

[0039] In one embodiment of the present invention, the formula for calculating the credibility score is as follows: ; ; in, Represents data items Credibility score, This represents the i-th IoT data item to be evaluated. Index representing the evaluation dimension This represents the total number of evaluation dimensions. This represents the weight coefficient of the j-th evaluation dimension. Represents data items The base credibility score on the j-th evaluation dimension, This represents the sensitivity parameter for cross-validation. Represents data items Cross-validation strength, This represents the threshold parameter for cross-validation. The normalized information entropy function, It is a data item Information entropy It is the theoretical maximum information entropy.

[0040] Understandable. The settings are based on the importance of different dimensions in the overall credibility assessment. This reflects the level of trust in a data item across a specific dimension. This represents a weighted overall credibility index; the larger this value, the higher the credibility index. The closer to 0, therefore The closer the value is to 1, the higher the confidence level. This ensures that the positive contributions from multiple dimensions can be superimposed non-linearly, while guaranteeing that the final score is within the [0,1] range. (Sensitivity parameter for cross-validation) This is used to adjust the influence of external verification information on the final score; it can be dynamically adjusted based on the reliability and completeness of the verification information. This reflects the quantity and quality of external verification information. The richer and more consistent the verification information, the larger this value. hour, A value greater than 0.5 indicates that the verification information has a positive effect on credibility; when hour, A value less than 0.5 indicates insufficient verification information, which has a negative impact on credibility. It reflects the richness and uncertainty level of the information content of the data item. Data items with higher information entropy usually contain more effective information and therefore obtain a higher credibility adjustment coefficient.

[0041] S3. Calculate the complexity of the structured query representation based on a nonlinear evaluation model to obtain a query complexity score.

[0042] In one embodiment of the present invention, the nonlinear evaluation model includes a query parsing module, a complexity calculation module, an analysis module, and a comprehensive scoring module, wherein, The query parsing module is used to parse the structured query representation and extract basic factors, wherein the basic factors include the number of entities, relation complexity, time span, and spatial range; The complexity calculation module is used to perform nonlinear transformations on the basic factors and calculate the basic complexity. The analysis module is used to construct a query semantic network and calculate the mutual information and degree centrality of semantic relations based on the query semantic network. The comprehensive scoring module is used to fuse the basic complexity and semantic network complexity to obtain the final complexity score.

[0043] This invention accurately quantifies the processing difficulty and resource requirements of queries through a nonlinear evaluation model, and more accurately predicts the query execution time and resource overhead.

[0044] In one embodiment of the present invention, the query parsing module receives a query representation in the form of a quintuple, performs entity identification and classification analysis on the query subject, counts basic information such as the number of IoT device types, data attribute types, and physical object categories involved in the query, and calculates the entity quantity factor. By analyzing the relationships between entities in the query, including semantic connections such as hierarchical relationships, dependency relationships, and aggregation relationships, the complexity and nesting depth of the relationships are evaluated to form a relationship complexity factor; The time constraint T contains information such as time range, time granularity, and temporal relationship; the time span and temporal complexity of the query are calculated; and a time span factor is generated. By analyzing spatial information such as geographical scope, spatial accuracy, and regional relationships within spatial constraints, the spatial coverage and spatial complexity are calculated to obtain the spatial range factor.

[0045] In one embodiment of the present invention, the complexity calculation module receives the basic factors extracted by the query parsing module and performs nonlinear transformation and complexity quantification.

[0046] In this process, a nonlinear response function is used to transform each fundamental factor, and a power function is used to perform a nonlinear mapping on the fundamental factors.

[0047] In one embodiment of the present invention, the analysis module constructs a directed graph structure query semantic network based on the semantic information of entities and relationships in the query, wherein nodes represent conceptual entities involved in the query and edges represent semantic relationships between entities; By analyzing the topological characteristics of semantic networks, including graph theory metrics such as node degree distribution, path length, clustering coefficient, and centrality index, the semantic complexity of queries is quantified. Calculate the mutual information between each edge and the query target, and evaluate the importance and information contribution of each semantic relation to query processing.

[0048] In one embodiment of the present invention, the comprehensive scoring module combines multiple complexity factors through a nonlinear function to ensure that the complexity of each dimension is reasonably reflected in the final score, and uses a product form to combine the basic factors to avoid the low complexity of one dimension masking the high complexity of other dimensions.

[0049] In one embodiment of the present invention, the formula for calculating the query complexity score is as follows: in, This represents the query complexity score. Indices representing the index of the fundamental factors. This represents the adjustment parameter of the k-th basic factor. This represents the standardized value of the k-th fundamental factor. This represents the nonlinear response parameter of the k-th fundamental factor. Indicates the index of a semantic network edge. This indicates the number of edges in the query semantic network. Indicates the first l Degree centrality of edges in semantic networks Indicates the first l The mutual information between the edge and the query.

[0050] Understandably, the query complexity score reflects the difficulty of processing a query across multiple dimensions, typically ranging from 1 to 100. A higher score indicates greater query complexity, requiring more computational resources and processing time. The value of k, from 1 to 4, corresponds to the entity number factor, relation complexity factor, time span factor, and spatial range factor, respectively. (Entity number factor) The relationship complexity factor is calculated by dividing the number of different entity types involved in the query by the reference baseline. It reflects the richness of the entities involved in the query. Due to the diversity of entity relationships and the nesting levels in queries, complex relationship structures significantly increase the difficulty of query processing. (Time span factor) Considering the length of the time range and the required time precision, long-term and high-precision time queries require processing more historical data. Spatial range factor. Based on the geographic coverage and spatial precision calculation of the query, large-scale, high-precision spatial queries will consume more storage and computing resources.

[0051] These parameters are used to control the impact of each fundamental factor on the overall complexity. They are set based on the characteristics of different types of IoT applications and historical statistical data. Controlling the nonlinear variation characteristics of each factor, when This leads to a superlinear increase in complexity, reflecting the nonlinear characteristics of resource requirements in IoT query processing.

[0052] Using a logarithmic function to handle semantic network complexity effectively addresses the impact of the exponential growth in the number of semantic relations on complexity, preventing excessive influence on the overall score when the semantic network size is too large. Simultaneously, the logarithmic function maintains sensitivity to changes in semantic complexity, ensuring that the increase in semantic relations and changes in their importance are reasonably reflected in the final score.

[0053] S4. Determine the query execution strategy based on the query complexity score and the IoT resource status. The IoT resource status includes computing resource status, storage resource status, and network resource status. The computing resource status includes edge node computing power, memory resource status, and GPU acceleration resources. The storage resource status includes local storage capacity, data cache status, and data backup status. The network resource status includes network connection quality, communication protocol support, and network topology status.

[0054] In one embodiment of the present invention, the query execution strategy includes a local processing strategy, an edge collaborative processing strategy, and a cloud processing strategy, wherein... The local processing strategy is used to complete query processing within a single edge node; The edge collaborative processing strategy is used to collaboratively complete query processing across multiple edge nodes; The cloud processing strategy is used to report queries to the cloud service layer for processing.

[0055] Specifically, step S4 includes: The computing resource status is obtained based on the CPU utilization, memory usage, remaining storage space, and network bandwidth usage of local edge nodes. The availability, completeness, and timeliness of the data are evaluated locally through structured queries. Based on the processing capabilities of local edge nodes and historical query statistics, a first complexity threshold and a second complexity threshold are set, where the first complexity threshold corresponds to the upper limit of queries that local resources can process independently, and the second complexity threshold corresponds to the upper limit of queries that edge collaborative processing can handle. When the query complexity score is lower than the first complexity threshold and the local data completeness meets the query requirements, the local processing strategy is selected. When the complexity score is greater than the first complexity threshold but less than the second complexity threshold, or when the local data is incomplete, the preset neighbor node status table is queried to obtain the processing capabilities, data types, and network connection quality information of the surrounding edge nodes, and the edge collaborative processing strategy is executed. When the complexity score is higher than the second complexity threshold or the combination of adjacent edge nodes still cannot meet the query requirements, the cloud processing strategy is selected.

[0056] Understandably, CPU utilization monitoring uses a sliding window averaging method, which calculates the recent average load level and load change trend through continuous sampling; memory utilization assessment includes the use of physical memory and virtual memory; and monitoring of remaining storage space includes not only overall available space statistics, but also analysis of storage device read / write performance, data distribution characteristics, and access hotspot patterns.

[0057] By analyzing the data types, data sources, and data ranges involved in the query, and checking whether the local storage contains all the data items required for the query, a data availability assessment is obtained; by checking the integrity, accuracy, consistency, and relevance of the data, a data integrity assessment is obtained; and by analyzing the degree of matching between the update time, data age, and timeliness requirements of the local data, a data timeliness assessment is obtained.

[0058] This invention achieves optimized allocation and load balancing of computing resources by intelligently selecting execution strategies based on query complexity scores and real-time resource status. It avoids resource waste or performance bottlenecks that may be caused by fixed strategy selection, ensuring that queries of different complexities can obtain appropriate processing resources and execution environments, thereby improving information query efficiency and stability.

[0059] S5. Perform an information query according to the query execution strategy to obtain the initial query results; S6. The query results are fused based on the credibility score to obtain the final information query results.

[0060] This invention achieves adaptive processing and quality optimization of information queries in the Internet of Things (IoT) environment through multimodal query input, improved data credibility assessment algorithm, nonlinear complexity assessment model, intelligent execution strategy decision-making, and credibility fusion processing. It can dynamically select the optimal execution strategy based on query complexity and resource status, and switch between local processing, edge collaborative processing, and cloud processing. This effectively balances the relationship between query response time, resource utilization efficiency, and result quality, and significantly improves the overall performance and user experience of the IoT information query system.

[0061] Specifically, step S6 includes: The initial query results are deduplicated to obtain a set of deduplicated data items; The data in the deduplicated data item set is matched with the structured query representation to obtain a relevance score for each data item; Calculate the similarity and difference between data items in the deduplicated data item set to obtain the information evaluation result of each data item; Based on the credibility score, data integrity index, and measurement accuracy index of each data item, a weighting algorithm is used to calculate the data quality weight of each data item. The timeliness discount factor for each data item is calculated based on the data generation time and query time. The contribution of each data item to the query result is calculated based on the relevance score, information evaluation result, data quality weight and timeliness discount factor. Based on the credibility score and contribution of each data item, a weighted fusion process is used to obtain the final information query results.

[0062] Understandably, deduplication is based on precise matching of key identifiers of data items to identify identical duplicate data items, resulting in a set of deduplicated data items. The similarity between data items is calculated to determine the matching degree of each query element of the data item.

[0063] This invention also provides an information query system based on the Internet of Things, implementing the information query method described above, including: The query input module is used to receive multimodal query input and convert the multimodal query input into a structured query representation; The credibility assessment module is used to assess the credibility of IoT data using an improved data credibility assessment algorithm to obtain a credibility score for each data item. The complexity calculation module is used to calculate the complexity of the structured query representation based on a nonlinear evaluation model to obtain a query complexity score. The execution strategy determination module is used to determine the query execution strategy based on the query complexity score and the IoT resource status. The information query module is used to perform information queries according to the query execution strategy and obtain initial query results; The fusion module is used to fuse the query results based on the credibility score to obtain the final information query results.

[0064] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. An information retrieval method based on the Internet of Things, characterized in that, include: S1. Receive multimodal query input and convert the multimodal query input into a structured query representation; The structured query representation adopts a five-tuple structure, including query subject, time constraint, space constraint, context information, and data credibility requirements; S2. An improved data credibility assessment algorithm is used to assess the credibility of IoT data and obtain a credibility score for each data item. Step S2 specifically includes: The feature extraction module is used to extract features from IoT data to obtain device hardware fingerprint features, network behavior features, and data content features; The device hardware fingerprint features, network behavior features, and data content features are reconstructed by using a feature learning module to encode and decode feature vectors, thereby obtaining the data potential representation and reconstruction error. The scoring module calculates an initial confidence score based on the latent representation and reconstruction error of the data; Obtain external validation information and use the information validation module to calculate the cross-validation strength; Acquire historical IoT data and use the analysis module to analyze the historical data to obtain a time series consistency score; The credibility score is calculated based on the initial credibility score, cross-validation strength, and temporal consistency score. S3. The complexity of the structured query representation is calculated based on a nonlinear evaluation model to obtain a query complexity score. The nonlinear evaluation model includes an analysis module, a query parsing module, a complexity calculation module, and a comprehensive scoring module. The analysis module constructs a query semantic network and calculates the mutual information and degree centrality of semantic relationships based on the query semantic network. The query parsing module parses the structured query representation and extracts basic factors, including the number of entities, relationship complexity, time span, and spatial range. The complexity calculation module performs a nonlinear transformation on the basic factors and calculates the basic complexity. The comprehensive scoring module fuses the basic complexity with the semantic network complexity to obtain the final complexity score. S4. Determine the query execution strategy based on the query complexity score and the IoT resource status; the query execution strategy includes a local processing strategy, an edge collaborative processing strategy, and a cloud processing strategy. Step S4 specifically includes: The computing resource status is obtained based on the CPU utilization, memory usage, remaining storage space, and network bandwidth usage of local edge nodes. The availability, completeness, and timeliness of the data are evaluated locally through structured queries. Based on the processing capabilities of local edge nodes and historical query statistics, a first complexity threshold and a second complexity threshold are set, where the first complexity threshold corresponds to the upper limit of queries that local resources can process independently, and the second complexity threshold corresponds to the upper limit of queries that edge collaborative processing can handle. When the query complexity score is lower than the first complexity threshold and the local data completeness meets the query requirements, the local processing strategy is selected. When the complexity score is greater than the first complexity threshold but less than the second complexity threshold, or when the local data is incomplete, the preset neighbor node status table is queried to obtain the processing capabilities, data types, and network connection quality information of the surrounding edge nodes, and the edge collaborative processing strategy is executed. When the complexity score is higher than the second complexity threshold or the combination of adjacent edge nodes still cannot meet the query requirements, the cloud processing strategy is selected. S5. Perform an information query according to the query execution strategy to obtain the initial query results; S6. Based on the credibility score, the query results are fused to obtain the final information query results. Step S6 specifically includes: The initial query results are deduplicated to obtain a set of deduplicated data items; The data in the deduplicated data item set is matched with the structured query representation to obtain a relevance score for each data item; Calculate the similarity and difference between data items in the deduplicated data item set to obtain the information evaluation result of each data item; Based on the credibility score, data integrity index, and measurement accuracy index of each data item, a weighting algorithm is used to calculate the data quality weight of each data item. The timeliness discount factor for each data item is calculated based on the data generation time and query time. The contribution of each data item to the query result is calculated based on the relevance score, information evaluation result, data quality weight and timeliness discount factor. Based on the credibility score and contribution of each data item, a weighted fusion process is used to obtain the final information query results.

2. The information query method based on the Internet of Things according to claim 1, characterized in that, The improved data credibility assessment algorithm includes a feature extraction module, a feature learning module, a scoring module, an information verification module, and an analysis module.

3. The information query method based on the Internet of Things according to claim 1, characterized in that, The formula for calculating the credibility score is as follows: ； in, Represents data items Credibility score, This represents the i-th IoT data item to be evaluated. Index representing the evaluation dimension This represents the total number of evaluation dimensions. This represents the weight coefficient of the j-th evaluation dimension. Represents data items The base credibility score on the j-th evaluation dimension, This represents the sensitivity parameter for cross-validation. Represents data items Cross-validation strength, This represents the threshold parameter for cross-validation. This represents the normalized information entropy function.

4. The information query method based on the Internet of Things according to claim 1, characterized in that, The local processing strategy is used to complete query processing within a single edge node; The edge collaborative processing strategy is used to collaboratively complete query processing across multiple edge nodes; The cloud processing strategy is used to report queries to the cloud service layer for processing.

5. An information query system based on the Internet of Things, characterized in that, The information query method as described in any one of claims 1-4 includes: The query input module is used to receive multimodal query input and convert the multimodal query input into a structured query representation; The credibility assessment module is used to assess the credibility of IoT data using an improved data credibility assessment algorithm to obtain a credibility score for each data item. The complexity calculation module is used to calculate the complexity of the structured query representation based on a nonlinear evaluation model to obtain a query complexity score. The execution strategy determination module is used to determine the query execution strategy based on the query complexity score and the IoT resource status. The information query module is used to perform information queries according to the query execution strategy and obtain initial query results; The fusion module is used to fuse the query results based on the credibility score to obtain the final information query results.

Images