An internet of things industry development evaluation method and device

Abstract

This invention discloses a method and apparatus for evaluating the development of the Internet of Things (IoT) industry. The method includes: acquiring data sources related to development planning; collecting and organizing the data sources using text mining technology to generate a standardized text database; segmenting and removing stop words from the original text in the standardized text database; establishing a text feature set through statistical word frequency distribution and feature extraction to obtain text feature results; designing indicators based on the text feature results using an evaluation framework encompassing development goals, supporting tools, and complementary measures; constructing a multi-input-output table using the evaluation indicators in the multi-input-output table and a PMC index calculation model to obtain PMC index evaluation results; and evaluating the PMC index results using a consistency analysis model to formulate development improvement suggestions. This invention effectively solves the problems of strong subjectivity, low evaluation accuracy, and lack of a systematic evaluation framework in existing evaluation methods.

Description

Technical Field

[0001] This invention relates to the field of data analysis technology, and in particular to a method and apparatus for evaluating the development of the Internet of Things (IoT) industry. Background Technology

[0002] As one of the core technologies of the digital economy era, the Internet of Things (IoT) is an important support for realizing the digital transformation of industries. Since the 1990s, the development of IoT technology has undergone an evolution from sensing and identification to network interconnection, and then to intelligent analysis.

[0003] Currently, methods such as fuzzy comprehensive evaluation and analytic hierarchy process (AHP) are commonly used to evaluate industrial development. For example, some researchers use bibliometric methods to analyze the characteristics of development and evolution, while others use instrumental theory to explore structural issues.

[0004] The existing and relatively mature approach is to use text mining methods to identify features and extract information from text, and then use machine learning, natural language processing and other technologies to interpret the deep semantic meaning of the text, and then to quantitatively evaluate the results.

[0005] However, existing evaluation methods suffer from high subjectivity and low accuracy. Furthermore, quantitative evaluation studies on the development of my country's Internet of Things (IoT) industry are mostly limited to the macro level, with relatively insufficient regional-level assessments. In addition, existing research lacks a systematic evaluation framework, making it difficult to provide scientific decision-making support for optimization. Summary of the Invention

[0006] The purpose of this invention is to provide a method and apparatus for evaluating the development of the Internet of Things (IoT) industry, in order to solve the technical problems of existing evaluation methods, such as strong subjectivity, low evaluation accuracy, and lack of a systematic evaluation framework.

[0007] To achieve the above objectives, this invention provides a method for evaluating the development of the Internet of Things (IoT) industry, comprising:

[0008] Obtain data sources related to development planning, and use text mining technology to collect and organize the data sources to generate a standardized text database;

[0009] The original text in the standardized text database is segmented and stop words are removed. A text feature set is established by statistical word frequency distribution and feature extraction to obtain the text feature results.

[0010] Based on the text feature results, an evaluation framework encompassing development goals, support tools, and supporting measures is used to design indicators and construct a multi-input-output table.

[0011] Using the evaluation indicators in the aforementioned multi-input-output table, data is processed through a preset PMC index calculation model to obtain the PMC index evaluation result;

[0012] Based on the evaluation results of the PMC index, a system evaluation is conducted through a pre-set consistency analysis model to generate development and improvement suggestions.

[0013] Preferably, the step of using text mining technology to collect and organize the data source to generate a standardized text database specifically includes:

[0014] The data source is filtered using keyword indexing technology, and a multi-source data set is generated through automatic collection and processing.

[0015] The multi-source data set is subjected to format recognition and encoding standardization processing to establish a standardized data index;

[0016] Based on the standardized data index, data hierarchy is divided, and a multi-level data structure is generated through classification and organization.

[0017] The multi-level data structure is used to design text identifiers, and a unique identifier sequence is generated through automatic encoding.

[0018] The standardized text database is generated by integrating the data based on the unique identifier sequence and the multi-level data structure.

[0019] Preferably, the step of segmenting and removing stop words from the original text in the standardized text database specifically includes:

[0020] Obtain the original text from the standardized text database, and use domain knowledge to build a professional dictionary database to form a basic word segmentation dictionary database;

[0021] Based on the aforementioned word segmentation lexicon, text segmentation is performed to obtain initial word segmentation results;

[0022] The initial word segmentation results are subjected to syntactic structure analysis, and a basic set of word groups is obtained through part-of-speech tagging.

[0023] The aforementioned basic phrase set is used for synonym analysis and word form unification, and standard phrases are obtained through rule-based processing.

[0024] Stop word identification and filtering are performed based on the standard word groups to obtain a set of valid word groups.

[0025] Preferably, the step of designing indicators and constructing a multi-input-output table based on the text feature results and utilizing an evaluation framework encompassing development goals, supporting tools, and complementary measures includes:

[0026] The feature vectors of the text feature results are obtained, and the evaluation framework is used to design indicators to form an evaluation indicator framework.

[0027] Based on the aforementioned evaluation index framework, quantitative rules are designed, and index calculation standards are established through numerical mapping.

[0028] The qualitative features are numericalized using the aforementioned index calculation standards, and quantitative data is obtained through transformation rules.

[0029] Based on the quantified data, a matrix structure is designed, and a multidimensional data matrix is ​​obtained through dimensional organization.

[0030] The input-output mapping relationship is established using the multidimensional data matrix, and a multi-input-output table is obtained through modeling.

[0031] Preferably, the step of using the evaluation indicators in the multi-input-output table to process data through a preset PMC index calculation model to obtain the PMC index evaluation result specifically includes:

[0032] Coefficient analysis was performed using text type and issuing unit level information, and the text strength coefficient α value was calculated through hierarchical evaluation.

[0033] Based on the text intensity coefficient α value, the implementation period data is analyzed, and the implementation cycle parameter β value is obtained through time series rule processing.

[0034] The multi-input-output table is classified using the implementation cycle parameter β, and the input-output variable set is obtained through a variable identification algorithm.

[0035] Based on the input and output variable set, data standardization processing is performed, and a standardized index matrix is ​​obtained through dimensional transformation.

[0036] By using the standardized index matrix and coefficient parameters, the PMC index evaluation result containing the PMC index value is obtained through a preset PMC calculation model.

[0037] Preferably, after obtaining the PMC index evaluation result containing the PMC index value by combining the standardized index matrix with coefficient parameters through a preset PMC calculation model, the process includes:

[0038] The coordinate system is designed using the PMC index evaluation results, and a three-dimensional display space is constructed through dimension mapping.

[0039] Data positioning is performed based on the three-dimensional display space, and a scattered point distribution set is obtained through spatiotemporal mapping.

[0040] The scattered point distribution set is used for surface fitting, and the initial surface is obtained by cubic spline interpolation algorithm;

[0041] Smoothness optimization is performed based on the initial surface, and the optimized surface model is obtained by adjusting the parameters.

[0042] The optimized surface model is used for visualization processing, and the PMC surface diagram is obtained through color mapping and node annotation.

[0043] Preferably, the step of conducting a system evaluation based on the PMC index evaluation results using a preset consistency analysis model to generate development and improvement suggestions includes:

[0044] Based on the PMC index evaluation results, adjacent text data are extracted using the PMC surface plot, and the rate of change characteristics are obtained through index comparison.

[0045] Hierarchical analysis is performed based on the aforementioned rate of change characteristics, and a consistency score is obtained by calculating the matching degree.

[0046] The consistency score is used to perform a difference analysis, and the difference feature vector is obtained through index comparison;

[0047] Influencing factors are identified based on the aforementioned differential feature vectors, and an influence degree matrix is ​​obtained through correlation analysis.

[0048] Feature extraction is performed using the aforementioned influence degree matrix, and key influencing factors are obtained through attribution analysis.

[0049] Preferably, the step of conducting a system evaluation based on the PMC index evaluation results using a preset consistency analysis model to generate development and improvement suggestions includes:

[0050] Obtain the evaluation results of the PMC index and establish an adaptive mind mapping framework;

[0051] The adaptive mind mapping framework is used to break down complex problems into interconnected sub-problems;

[0052] Construct a directed acyclic graph structure, prioritize the subproblems, and analyze the subproblems according to the priority ranking to obtain the analysis results of the subproblems;

[0053] The analysis results of the aforementioned sub-problems are integrated to form the development and improvement recommendations.

[0054] Continuously update the data of the PMC index evaluation results and optimize the problem decomposition and analysis strategies.

[0055] Preferably, the step of conducting a system evaluation based on the PMC index evaluation results using a preset consistency analysis model to generate development and improvement suggestions includes:

[0056] Construct a policy language agent system and establish a latent space model for policy representation;

[0057] The strategy combination sampled from the latent space model is used to obtain a strategy combination, which is then applied to the evaluation and optimization process.

[0058] Establish a multi-level strategy feedback mechanism to evaluate the execution effect of the strategy combination and obtain the strategy execution evaluation result;

[0059] The latent space distribution of the latent space model is updated based on the evaluation results of the strategy execution, and an optimization strategy is generated.

[0060] By comprehensively optimizing evaluation, analysis, and implementation strategies, overall performance can be improved.

[0061] Preferably, the method further includes:

[0062] Obtain the evaluation indicators from the multi-input-output table and establish a mixed linear regression model;

[0063] A technique combining tensor decomposition and spectral methods is employed to ensure global convergence;

[0064] Key parameters of the PMC index are estimated using a gradient-based stochastic approximation method.

[0065] Using Bayesian nonparametric mixture models to handle nonlinear relationships in data;

[0066] A parameter learning algorithm is constructed to ensure the convergence of the hybrid linear regression model, and the PMC index evaluation result containing the PMC index value is obtained.

[0067] The present invention also provides an evaluation device for the development of the Internet of Things industry, comprising:

[0068] The data acquisition module acquires data sources related to the development plan, uses text mining technology to collect and organize the data sources, and generates a standardized text database.

[0069] The text processing module is used to perform word segmentation and stop word removal on the original text in the standardized text database, and to establish a text feature set by statistical word frequency distribution and feature extraction to obtain text feature results;

[0070] The indicator construction module is used to design indicators based on the text feature results and using an evaluation framework that includes development goals, support tools, and supporting measures, and to construct a multi-input-output table.

[0071] The index calculation module is used to process data using the evaluation indicators in the multi-input-output table and a preset PMC index calculation model to obtain the PMC index evaluation result.

[0072] The optimization suggestion module is used to conduct a system evaluation based on the evaluation results of the PMC index through a preset consistency analysis model, and generate development and improvement suggestions.

[0073] The beneficial effects of this invention are:

[0074] 1. This invention proposes a quantitative evaluation method for development based on a combination of text mining and PMC index model, which can effectively reduce the subjectivity of the evaluation process and improve the evaluation accuracy;

[0075] 2. This invention innovatively incorporates both macro and regional levels into the evaluation system, expanding the evaluation dimensions and making development assessment more comprehensive and systematic;

[0076] 3. This invention designs a multi-dimensional evaluation index system that includes development goals, support tools, supporting measures, and implementation effects, providing a more scientific evaluation framework;

[0077] 4. This invention establishes a dynamic feedback mechanism, enabling continuous optimization of the evaluation and ensuring the timeliness and accuracy of the evaluation results;

[0078] 5. This invention proposes a visualization method based on PMC surfaces, which intuitively displays the consistency level, making it easier for decision-makers to understand and use the evaluation results. Attached Figure Description

[0079] To more clearly illustrate the technical solutions in the embodiments of this application, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0080] Figure 1 A flowchart of the IoT industry development evaluation method provided in this embodiment of the invention;

[0081] Figure 2 A flowchart for PMC index calculation and consistency evaluation provided in this embodiment of the invention;

[0082] Figure 3 This is a structural block diagram of the IoT industry development evaluation device provided in an embodiment of the present invention. Detailed Implementation

[0083] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0084] Please see Figure 1 , Figure 1 A flowchart illustrating an evaluation method for the development of the Internet of Things (IoT) industry, provided in an embodiment of the present invention. The method includes:

[0085] Step 101: Obtain data sources related to development plans, and use text mining techniques to collect and organize the data sources to generate a standardized text database.

[0086] Specifically, in this step, first determine the scope of data sources required for the evaluation of the development of the Internet of Things industry, including various types of text materials such as development plan documents, industrial support documents, technical standard specifications, and industry reports. Then, use automated crawler technology to collect relevant texts from channels such as public websites, professional databases, and industry platforms, and perform format unification and structured processing on the collected text materials. During this process, verify the source and check the quality of the texts, and eliminate duplicate, invalid, or low-quality texts. Next, according to the professional knowledge system in the field of the Internet of Things, classify and code the texts, and establish a multi-dimensional index system, including dimensions such as release time dimension, technology field dimension, and development stage dimension. Finally, store the processed texts in a structured database to construct a text corpus for the development of the Internet of Things industry, providing a standardized data basis for subsequent analysis. This step ensures that the data sources in the evaluation process are comprehensive, authoritative, and timely, and is the basic link of the entire evaluation method. In actual operation, different data collection strategies will be set according to different sub-domains of the Internet of Things (such as intelligent manufacturing, smart agriculture, smart home, etc.) to ensure the professionalism and pertinence of the text database.

[0087] Step 102: Perform word segmentation and stop word removal on the original texts in the standardized text database, and establish a text feature set by statistical word frequency distribution and feature extraction to obtain text feature results.

[0088] In this step, first read the original documents in the text database, and perform Chinese word segmentation in combination with the professional word list in the field of the Internet of Things to accurately identify professional terms and compound concepts related to the Internet of Things. For special vocabulary in the field of the Internet of Things, such as "RFID", "M2M", "edge computing", etc., use special rules for identification and annotation to ensure the accuracy of word segmentation. Then, perform词性标注 (pos tagging) on the word segmentation results to identify different types of words such as nouns, verbs, adjectives, etc., and filter out words that have no substantial meaning for analysis, such as common function words like "的", "是", "和", etc., according to the predefined stop word list. Next, the system calculates the occurrence frequency and distribution characteristics of each word in the text, and uses the TF-IDF algorithm to evaluate the importance of each word in the text of the development of the Internet of Things industry, and identify the keyword vocabulary that can reflect the industry characteristics. In addition, semantic features of the text are also extracted, including multi-dimensional features such as theme features, development trends, and technical correlation degrees. Finally, these word frequency features and semantic features are integrated into a structured feature vector to form text feature results, providing data support for subsequent indicator design and evaluation system construction. In actual applications, the professional dictionary of the Internet of Things will be updated regularly to adapt to the rapidly evolving technical terms in this field and improve the accuracy of word segmentation and feature extraction.

[0089] Step 103: Based on the text feature results, design indicators using an evaluation framework that includes development goals, support tools, and supporting measures, and construct a multi-input-output table.

[0090] In this step, based on the aforementioned text feature analysis results, a multi-dimensional indicator system for evaluating the development of the Internet of Things (IoT) industry is constructed. First, key indicators reflecting the development goals of the IoT industry are extracted from the text features, including dimensions such as industry scale growth targets, technological innovation levels, and application penetration. Second, indicators related to support tools are extracted, covering aspects such as support intensity, investment scale, and market access conditions. Third, indicators related to supporting measures are extracted, including elements such as talent training systems, the completeness of standard construction, and ecosystem construction. These extracted qualitative indicators are then quantified using a combination of expert scoring and fuzzy comprehensive evaluation to transform the textual descriptions into measurable values. Next, a multi-dimensional indicator matrix structure is designed, classifying and organizing the indicators according to input and output indicators based on the upstream and downstream relationships and development stages of the IoT industry chain. Finally, logical mapping relationships between indicators are established, constructing a complete multi-input-output table. This table reflects both the internal structure and operating mechanism of the IoT industry development and provides a scientific data foundation for calculating the PMC index. In practice, this indicator system will be dynamically adjusted according to the evolution of IoT technology and the expansion of its applications to ensure the forward-looking nature and adaptability of the evaluation indicators.

[0091] Step 104: Using the evaluation indicators in the multi-input-output table, the data is processed through the preset PMC index calculation model to obtain the PMC index evaluation result.

[0092] This step is the core of the IoT industry development evaluation. Based on the evaluation indicators in the multi-input-output table, a systematic evaluation is conducted using the PMC index calculation model. First, the system determines the text strength coefficient α based on text type and issuing unit level, reflecting the guidance intensity and implementation rigidity of different texts in IoT industry development. Then, it analyzes the implementation period and timing of each text to determine the implementation cycle parameter β, reflecting the time span and execution cycle of different development documents. Next, the system standardizes the indicator data in the multi-input-output table to eliminate differences in the dimensions of different indicators, making them comparable. Then, the system calculates the indicator matching degree and strategy consistency between texts at different levels, forming a consistency coefficient γ, reflecting the degree of coordination in IoT industry development at different levels. Finally, based on the PMC index calculation formula, the system comprehensively considers text strength, implementation cycle, and consistency level to calculate the PMC index value for each evaluation object. As a comprehensive evaluation indicator, the PMC index reflects both the overall level of IoT industry development and the characteristics and interrelationships of different development stages, providing a quantitative basis for industry development decisions. Simultaneously, the system generates a PMC surface graph, visually displaying the trend of the index over time and at different levels, facilitating decision-makers to quickly grasp the development status. In practice, the PMC index model can set different parameter weights for different sub-sectors of the Internet of Things, reflecting the development characteristics and focus of each sub-sector.

[0093] The PMC index calculation model in this invention is a mathematical model based on multi-factor comprehensive evaluation, named the Public Text Modeling Consistency Analysis Framework. It is used to quantify the consistency level of the Internet of Things (IoT) industry development. The core idea of ​​this model is to weight and integrate three key elements—text strength, implementation cycle, and hierarchical coordination—to form a comprehensive evaluation index. Its mathematical expression is:

[0094] PMC = α·P + β·M + γ·C

[0095] Wherein, P represents the power factor, used to measure the guiding strength and implementation rigidity of development planning documents; M represents the maturity factor, reflecting the time span and execution cycle of development plans; and C represents the coordination factor, indicating the level of consistency between development plans at different levels. α, β, and γ are the weighting coefficients of the three factors, satisfying α + β + γ = 1.

[0096] In its implementation, the P-factor is calculated using the Analytic Hierarchy Process (AHP), which determines the relative importance weight of each text based on its text type, publishing unit level, and implementation rigidity. The system first constructs a judgment matrix A, with matrix elements a... ijThe importance ratio of text i to text j is represented by λmax. Then, the maximum eigenvalue λmax and the corresponding eigenvector w of the matrix are calculated. After normalization of the eigenvector, the weight coefficients of each text are obtained, and the P-factor value is calculated.

[0097] The M-factor is calculated using a time-weighted average method, taking into account the publication time, planned implementation duration, and coverage period of each text. The specific calculation formula is as follows:

[0098]

[0099] Among them, w i t represents the weight coefficient of the i-th text. i The time parameter representing the text is calculated based on the recentity of the text's publication time and the length of the planned implementation period.

[0100] The C-factor is calculated using a text similarity algorithm to assess the degree of content consistency between development plan texts at different levels. The system uses the TF-IDF vector space model to convert the text into a high-dimensional vector representation, and then calculates the cosine similarity between the vectors as a consistency index. For multiple texts, a similarity matrix is ​​also constructed, and the average similarity or minimum similarity is calculated as an overall consistency evaluation.

[0101] Ultimately, the PMC index is derived from a weighted combination of the three factors mentioned above, with a value ranging from 0 to 1. A value closer to 1 indicates a higher level of consistency in the development of the Internet of Things (IoT) industry. To improve the model's adaptability, the system dynamically adjusts the weight coefficients of the three factors based on the characteristics of different regions and development stages, ensuring the accuracy and relevance of the evaluation results. Furthermore, the PMC index calculation considers temporal continuity, incorporating historical data smoothing to reduce index fluctuations and reflect stable trends in industry development.

[0102] Step 105: Based on the evaluation results of the PMC index, conduct a system evaluation through a preset consistency analysis model to form development and improvement suggestions.

[0103] In the final step, the PMC index evaluation results are analyzed and systematically assessed to uncover problems and optimization opportunities in the development of the Internet of Things (IoT) industry. First, the system extracts key data points from the PMC surface graph, analyzes the time nodes and trends of index fluctuations, and identifies critical turning points in industry development. Then, it compares the differences in PMC indices at different levels, assesses the degree of synergy between macro-planning and regional implementation, and identifies inconsistencies between levels. Next, through difference analysis, the system precisely locates the specific indicators and influencing factors causing these inconsistencies, generating a problem diagnosis report. Then, using correlation analysis, it establishes a mapping relationship between PMC index fluctuations and external factors, identifying the key elements most significantly impacting the development of the IoT industry. Finally, based on the above analysis results, the system generates targeted development improvement suggestions from dimensions such as industry goal synergy, resource allocation optimization, support tool selection, and implementation path design. These suggestions are based on objective analysis of historical data and incorporate a deep understanding of the development patterns of the IoT industry, providing a scientific basis for the formulation and adjustment of industry planning. The system also provides a dynamic feedback mechanism, continuously updating the evaluation results and optimization suggestions as new data is input, ensuring the timeliness and accuracy of the evaluation. In practical applications, the system can generate differentiated development suggestions based on the industrial base and development stage of different regions, thereby improving the relevance and operability of the suggestions.

[0104] The consistency analysis model of this invention is a mathematical model specifically designed to evaluate the coordination of IoT industry development. It conducts in-depth analysis of PMC index results from both vertical and horizontal dimensions. Based on structural equation modeling (SEM) and multidimensional scaling (MDS) techniques, this model can identify the degree of coordination and the reasons for differences between different levels and regions.

[0105] Vertical consistency analysis employs structural equation modeling to construct a hierarchical transmission path model for the development of the Internet of Things (IoT) industry. This model includes three latent variables: macro, meso, and micro levels. Each latent variable has multiple observed variables, corresponding to the specific components of the PMC index. Maximum likelihood estimation is used to calculate model parameters and evaluate path coefficients and goodness of fit between levels. Path coefficients reflect the strength of influence between upper and lower levels, while goodness of fit indicates the degree of matching between the overall model and actual data. Based on these parameters, the system identifies weak links in hierarchical transmission, such as the failure of macro-level planning to effectively translate into regional implementation, or the disconnect between regional planning and enterprise actions.

[0106] The horizontal consistency analysis employs a multidimensional scaling algorithm and clustering analysis to map the PMC indices and their components of different regions into a two-dimensional space, intuitively displaying the similarities and differences between regions. The system first constructs a dissimilarity matrix D between regions, where the matrix element dij represents the weighted Euclidean distance between region i and region j across each dimension of the PMC index. Then, through an iterative optimization algorithm, it finds a point set P in the two-dimensional space that preserves the original dissimilarity relationship between points as much as possible, i.e.:

[0107]

[0108] Where ||p i - p j || represents the Euclidean distance between points i and j in two-dimensional space. The optimized two-dimensional point set forms a regional consistency scatter plot; the closer the distance between points, the higher the consistency of regional development. The system further applies K-means or hierarchical clustering algorithms to group these points and identify regional clusters with similar development characteristics.

[0109] Furthermore, the model incorporates a consistency decomposition algorithm to break down overall consistency discrepancies into specific indicator dimensions, precisely identifying the key factors leading to inconsistencies. This algorithm employs analysis of variance (ANOVA) and relative importance analysis to calculate the contribution rate of each indicator to the overall discrepancy, thereby identifying the dominant factors influencing the level of consistency.

[0110] The consistency analysis model not only provides static evaluation results but also performs dynamic evolution analysis. By constructing time series models, it predicts future trends in consistency levels. The system uses an Autoregressive Integrated Moving Average (ARIMA) model or a Long Short-Term Memory (LSTM) network model to process historical data of the PMC index, capturing its cyclical fluctuations and long-term trends to provide early warnings and guidance for future development.

[0111] The above steps will be explained in detail below with reference to specific embodiments.

[0112] The steps for constructing a text corpus for the development of the Internet of Things (IoT) industry provided in this embodiment of the invention include:

[0113] Step 201: Use keyword indexing technology to filter the data source and generate a multi-source data set through automatic collection and processing.

[0114] Specifically, in this embodiment, firstly, using keywords such as "Internet of Things" (IoT), relevant texts from 2006 to 2023 are collected through channels such as legal databases and official portals at various levels. Specifically, the collection must adhere to three principles: authoritativeness, relevance, and representativeness. Authoritativeness means selecting officially released documents; relevance means the content of the documents is directly related to the development of the IoT industry; and representativeness means the documents have strong guiding significance.

[0115] Step 202: Perform format recognition and encoding standardization processing on the multi-source data set to establish a standardized data index.

[0116] Specifically, the collected texts undergo initial screening, removing documents such as letters and approvals, as well as documents whose content does not focus on the development of the Internet of Things. Furthermore, the text formats need to be standardized, including removing special characters and unifying encoding formats, to prepare for subsequent analysis.

[0117] Step 203: Based on the standardized data index, perform data hierarchy division and generate a multi-level data structure through classification and organization.

[0118] Specifically, the selected texts are categorized and organized at both macro and regional levels, and a multi-level text structure is established based on text type (such as planning, implementation, etc.).

[0119] Step 204: Use the multi-level data structure to design text identifiers and generate unique identifier sequences through automatic encoding.

[0120] Step 205: Integrate the data based on the unique identifier sequence and the multi-level data structure to generate the standardized text database.

[0121] Specifically, each text is assigned a unique identifier to facilitate subsequent tracing and analysis.

[0122] The steps for text preprocessing and feature extraction using software provided in this embodiment of the invention include:

[0123] Step 301: Obtain the original text from the standardized text database, construct a professional dictionary database using domain knowledge, and form a basic word segmentation dictionary database.

[0124] Step 302: Perform text segmentation based on the aforementioned word segmentation lexicon to obtain the initial word segmentation results.

[0125] Specifically, in this embodiment, ROST CM6 software is used to perform Chinese word segmentation and part-of-speech tagging on the text. Special attention is paid to handling IoT-related technical terms and compound words to ensure the accuracy of word segmentation. Simultaneously, a specialized IoT dictionary is established to assist the word segmentation process.

[0126] Step 303: Perform syntactic structure analysis on the initial word segmentation results, and obtain a basic word set through part-of-speech tagging.

[0127] Step 304: Perform synonym analysis and word form unification using the basic phrase set, and obtain standard phrases through rule-based processing.

[0128] Step 305: Based on the standard phrases, stop word identification and filtering are performed to obtain a set of valid phrases.

[0129] Specifically, stop word filtering is performed on the word segmentation results to remove function words and modal particles that have no substantial meaning for the analysis. In addition, data cleaning is required to process synonyms and near-synonyms, unify word forms, and improve the accuracy of subsequent analysis.

[0130] Next, the Term Frequency-Inverse Document Frequency (TF-IDF) algorithm is used to extract key feature words from the text, and the weight of each feature word is calculated. Through term frequency statistical analysis, high-frequency words and key concepts in the text are identified, revealing the key areas of interest.

[0131] The steps for constructing a multi-input-output evaluation system for the PMC index model provided in this embodiment of the invention include:

[0132] Step 401: Obtain the feature vector of the text feature result, and use the evaluation framework to design indicators to form an evaluation indicator framework.

[0133] Specifically, based on the results of feature word analysis, an evaluation index system is constructed from the dimensions of development goals, supporting tools, supporting measures, and implementation effects. Development goals include industry scale and technological innovation; supporting tools include fiscal and tax support and market access; and supporting measures include talent cultivation and standards development.

[0134] Step 402: Design quantitative rules based on the evaluation index framework and establish index calculation standards through numerical mapping.

[0135] Specifically, quantitative standards for each evaluation indicator were designed, employing a combination of quantitative and qualitative methods. For quantitative indicators, numerical measurements were used directly; for qualitative indicators, expert scoring was used for quantification to ensure the objectivity of the evaluation.

[0136] Step 403: Use the aforementioned index calculation standard to quantify the qualitative features and obtain quantitative data through conversion rules.

[0137] Step 404: Design a matrix structure based on the quantized data, and obtain a multidimensional data matrix through dimensional organization.

[0138] Step 405: Establish input-output mapping relationship using the multidimensional data matrix, and obtain a multi-input-output table through modeling.

[0139] Specifically, the quantified evaluation index data is organized into a multi-input-output matrix. The rows of the matrix represent different texts, the columns represent each evaluation index, and the matrix elements represent the corresponding quantified values.

[0140] Please see Figure 2 , Figure 2 A flowchart for PMC index calculation and consistency evaluation provided in this embodiment of the invention. The flowchart includes:

[0141] Step 501: Perform coefficient analysis using text type and issuing unit level information, and calculate the text strength coefficient α value through hierarchical evaluation.

[0142] Specifically, the key parameters in the PMC index model need to be defined. Specifically, for each text, its text strength coefficient α needs to be identified. This coefficient reflects the degree of mandatory nature of the text and is determined based on the text type and the issuing department level. For example, the α value is 1.0 for important documents, 0.8 for general documents, and 0.6 for routine documents.

[0143] Step 502: Analyze the implementation period data based on the text intensity coefficient α value, and obtain the implementation cycle parameter β value through time sequence rule processing.

[0144] Specifically, it is necessary to determine the text implementation period parameter β, which represents the time required for the text to produce its expected effects. This is determined by analyzing the implementation period explicitly stated in the text or by referring to the implementation experience of similar texts. For example, planning texts typically have a β value of 5 years, while implementation detail texts typically have a β value of 2-3 years.

[0145] Step 503: Use the implementation cycle parameter β value to classify the indicators of the multi-input-output table, and obtain the input-output variable set through the variable identification algorithm.

[0146] Specifically, the evaluation indicators are categorized into input-type and output-type variables. Input-type variables include the level of support, the scale of financial investment, and the completeness of supporting measures; higher values ​​for these variables indicate greater investment. Output-type variables include the industry scale growth rate, the degree of improvement in technological innovation capabilities, and the number of market entities cultivated; higher values ​​for these variables indicate better results.

[0147] Step 504: Perform data standardization processing based on the input and output variable set, and obtain a standardized index matrix through dimensional transformation.

[0148] Specifically, the indicator data in the multi-input-output matrix are standardized to eliminate the dimensional differences between different indicators. The standardization method uses the range method to uniformly transform the values ​​of each indicator to the [0,1] interval.

[0149] Step 505: Using the standardized index matrix and coefficient parameters, obtain the PMC index evaluation result containing the PMC index value through the preset PMC calculation model.

[0150] Specifically, consistency levels are calculated. For time-series texts, the continuity of adjustments is assessed by calculating the rate of change of each indicator value between adjacent texts. Simultaneously, considering the synergy between texts at different levels, the matching degree of macro and regional texts across various indicator dimensions is calculated.

[0151] Finally, according to the PMC index calculation formula:

[0152]

[0153] Where γ i denoted as PMC, where n is the number of texts, and represents the text consistency coefficient. The closer the PMC index is to 1, the higher the level of consistency.

[0154] Next, the calculation results of the PMC index can be visualized:

[0155] Step 506: Use the PMC index evaluation results to design a coordinate system and construct a three-dimensional display space through dimension mapping.

[0156] The PMC index evaluation results obtained from the aforementioned calculations need to be visualized. First, a coordinate system suitable for displaying the spatial distribution characteristics of the PMC index must be constructed. Specifically, the system will design a three-dimensional spatial coordinate system. The X-axis typically represents the time dimension, scaled according to the time series of the IoT industry development (e.g., quarterly or annually), from the earliest evaluation time to the latest evaluation time. The Y-axis represents the regional or hierarchical dimension, which can be divided according to different regional types (e.g., developed regions, emerging regions, basic regions, etc.) or management levels (e.g., macro level, regional level, enterprise level, etc.). The Z-axis represents the PMC index value, typically ranging from 0 to 1, with values ​​closer to 1 indicating a higher level of consistency. When constructing the coordinate system, the system will consider the data distribution characteristics, appropriately setting the axial scale and scale to ensure the clarity and interpretability of the final display. Simultaneously, the system will also set appropriate reference planes and grid lines in the coordinate system to help observers better understand the positional relationships of the data in three-dimensional space. After the coordinate system design is completed, the system will convert the PMC index evaluation results into spatial location information in the coordinate system through dimension mapping rules, in order to prepare for subsequent data visualization processing.

[0157] Step 507: Perform data positioning based on the three-dimensional display space, and obtain a scattered distribution set through spatiotemporal mapping.

[0158] Specifically, a three-dimensional coordinate system is established, where the X-axis represents the time dimension, the Y-axis represents the text hierarchy dimension (macroscopic, regional), and the Z-axis represents the PMC index value. The calculated PMC index values ​​are then plotted as scatter points in the coordinate system according to the time series and text hierarchy.

[0159] After constructing the coordinate system, the system maps the PMC index data of each evaluation object to a specific location in three-dimensional space. For each PMC index evaluation record, the system determines the X-coordinate value based on its time attribute, the Y-coordinate value based on its region or hierarchical attribute, and the Z-coordinate value based on its PMC index value, thus forming a data point in three-dimensional space. Once all evaluation records have been mapped, the system obtains a three-dimensional scatter plot consisting of multiple data points. The distribution of these scatter plots intuitively reflects the changes in the consistency level of the IoT industry development in the temporal and spatial dimensions. The system performs preliminary analysis on these scatter plots, identifying dense and sparse areas of distribution and discovering potential anomalies and clustering phenomena. To improve the visualization of the scatter plot distribution set, the system also assigns different colors to the scatter plots based on different PMC index value ranges; for example, high index value ranges (0.8-1.0) can be represented in red, medium index value ranges (0.5-0.8) in yellow, and low index value ranges (0-0.5) in blue. Furthermore, the system can further enhance the expressiveness of the data by adjusting the size, shape, and other visual attributes of the scatter points. Through this visualization method of scatter point distribution, decision-makers can quickly grasp the overall trend and local characteristics of the IoT industry's development.

[0160] Step 508: Use the scatter point distribution set to perform surface fitting, and obtain the initial surface through cubic spline interpolation algorithm.

[0161] While scatter plots can visually represent the discrete distribution of the PMC index, they fail to directly reflect the continuous trend of data changes. To more comprehensively demonstrate the variation patterns of the PMC index, the system needs to construct a continuous and smooth surface model based on the scatter plot. In this step, the system uses cubic spline interpolation to fit the scatter data to a surface. The core idea of ​​cubic spline interpolation is to construct a cubic polynomial function between adjacent data points, ensuring that the fitted surface not only passes through all original data points but also possesses good smoothness and continuity. Specifically, the system first divides the XY plane in three-dimensional space into a regular grid, then applies cubic spline interpolation to the data points within each grid to calculate the Z-value of each point. For grid regions with missing data, the system uses the interpolation results from neighboring grids for reasonable inference. By processing all grid regions one by one, the system finally obtains a continuous surface covering the entire XY plane. The advantage of this method is that it accurately preserves the characteristics of the original data while filling in data gaps and presenting the overall trend of data changes. After the initial fitting is completed, the system will calculate the error between the fitted surface and the original data points, evaluate the fitting quality, and provide a reference for subsequent surface optimization.

[0162] Step 509: Optimize the smoothness based on the initial surface, and obtain the optimized surface model by adjusting the parameters.

[0163] Specifically, cubic spline interpolation is used to fit discrete PMC index values ​​to a surface, generating a continuous PMC surface. By adjusting the surface smoothness parameter, it is ensured that the surface reflects both the overall consistent trend and local characteristics.

[0164] The initially fitted surface may exhibit local unevenness or overfitting, requiring further optimization. In this step, the system optimizes the smoothness of the initial surface by adjusting key parameters of the cubic spline interpolation algorithm. The main optimization methods include: First, adjusting the tension parameter to control the surface's "tightness"—a higher tension value makes the surface flatter, while a lower tension value makes the surface closer to the original data points; second, introducing a smoothing factor to reduce local fluctuations in the surface—a larger smoothing factor results in a smoother surface, but may increase deviation from the original data points; third, for potentially outlier data points, the system applies a local weighting strategy to reduce the impact of outliers on the overall surface shape. The system seeks the optimal balance between maintaining surface morphology and ensuring fitting accuracy, based on the actual needs of evaluating the development of the Internet of Things (IoT) industry. The optimization process is iterative; after each iteration, the system calculates the fitting error and smoothness index. The optimization process ends when both indices reach satisfactory levels or the preset number of iterations is reached. The resulting optimized surface model accurately reflects the overall distribution and trend of the PMC index, while also possessing good visual appeal and interpretability, providing decision-makers with clearer data insights.

[0165] The optimized surface model in this invention is a mathematical model for PMC index visualization and trend analysis. It employs an improved Bézier surface algorithm and adaptive smoothing technology to achieve high-precision and aesthetically pleasing surface fitting results. The key innovation of this model lies in combining traditional cubic spline interpolation with machine learning regularization methods, overcoming the overfitting and fluctuation problems that standard interpolation algorithms are prone to when dealing with irregular data distributions.

[0166] The mathematical expression for optimizing the surface model is:

[0167]

[0168] Among them, B i,n (x) and B j,m (y) are the nth and mth order B-spline basis functions in the x and y directions, respectively, P ij These are control point meshes that control the shape and features of the surface. The objective function for the model optimization process is:

[0169]

[0170] The first term is the data fitting error term, which represents the degree of deviation between the surface and the original data points; the second term, R(S), is the regularization term, which is used to control the smoothness and complexity of the surface; λ is the smoothing parameter, which controls the balance between fitting accuracy and smoothness.

[0171] In its implementation, an adaptive grid subdivision strategy is employed, dynamically adjusting the density of the control point grid based on the data distribution density and gradient. In regions with dense and rapidly changing data, the system increases the number of control points to improve fitting accuracy; conversely, in regions with sparse data or gradual changes, the number of control points is reduced to maintain the smoothness of the surface. This adaptive strategy significantly improves the fitting performance, especially for the uneven data distribution common in the development of the Internet of Things (IoT) industry.

[0172] To further optimize the surface quality, the system introduces a physical model-based surface optimization algorithm, treating the surface as a thin plate with elastic potential energy, and achieving natural smoothness by minimizing bending energy. Its potential energy function is:

[0173]

[0174] The optimal surface shape is obtained by solving the minimization problem using a variational method. Furthermore, the model integrates a surface feature preservation algorithm, which can retain important feature points such as peaks and valleys in the data while maintaining overall smoothness, thus avoiding information loss due to over-smoothing.

[0175] Optimized surface models are not only used for static visualization but also support dynamic evolutionary analysis. The system generates spatiotemporal evolution animations of the IoT industry development PMC index by constructing time-series 3D surface models, intuitively displaying the speed, direction, and acceleration characteristics of index changes. This dynamic visualization method helps decision-makers better understand the historical trajectory and future trends of industry development, providing intuitive references for strategic planning.

[0176] Step 510: Visualize the optimized surface model by color mapping and node annotation to obtain the PMC surface diagram.

[0177] Specifically, by setting different color zones, the numerical distribution of the PMC index can be visually displayed. For example, the PMC index can be represented by red, yellow, and blue colors for high, medium, and low zones, respectively, making it easier for decision-makers to quickly identify the strength of consistency. At the same time, key time events are marked on the surface graph to help analyze the impact of adjustments on the level of consistency.

[0178] After optimizing the surface model, the system performs final visualization processing to generate an intuitive and easy-to-read PMC surface diagram. First, the system applies a color mapping scheme to the optimized surface, assigning a corresponding color based on the Z-value (i.e., PMC index value) of each point on the surface. Common color mapping schemes include heatmap color schemes, such as displaying high index areas in red, medium index areas in yellow, and low index areas in blue, visually demonstrating the trend of index changes through color gradients. Next, the system marks key nodes and characteristic regions on the surface, such as peak points, trough points, stable areas, and rapidly changing areas of the PMC index, and adds corresponding text descriptions to help decision-makers understand the practical significance of these characteristic points. Simultaneously, the system adds clear scale markings and dimension descriptions to the coordinate axes to ensure the readability of the surface diagram. To enhance the 3D effect, the system applies appropriate lighting models and shadow effects to improve the three-dimensionality of the surface. Furthermore, the system provides interactive browsing functionality, allowing decision-makers to observe the PMC surface from different angles and distances, and even slice the surface to view the PMC index distribution at specific times or in specific regions. The resulting PMC surface plot not only possesses excellent visual appeal, but more importantly, it can intuitively demonstrate the consistent spatiotemporal distribution characteristics and changing patterns of the IoT industry, providing strong data support for decision-makers. In practical applications, this surface plot can be updated regularly to reflect the latest evaluation results, forming a dynamic evaluation mechanism.

[0179] Finally, development and improvement recommendations are formulated based on the PMC index evaluation results, and the specific implementation process is as follows:

[0180] Step 601: Based on the evaluation results of the PMC index, extract adjacent text data using the PMC surface plot, and obtain the rate of change characteristics through index comparison.

[0181] Specifically, by comparing the PMC index values ​​at adjacent time points, the rate of change is calculated, and periods of rapid rise or fall in the index are identified as key areas for analysis.

[0182] Step 602: Perform hierarchical analysis based on the change rate characteristics, and obtain a consistency score by calculating the matching degree.

[0183] Specifically, the matching degree of PMC indices for macro and regional level texts is calculated to assess the degree of synergy between texts at different levels. The matching degree can be quantified by calculating the correlation coefficient or the degree of difference between the PMC indices of two levels.

[0184] Step 603: Perform a difference analysis using the consistency score and obtain the difference feature vector through index comparison.

[0185] Specifically, for regions and time periods with low matching degrees, further analysis is conducted on the specific indicators and factors causing the inconsistencies. The difference values ​​of each indicator are used to form a difference feature vector, reflecting the specific manifestations of the inconsistencies.

[0186] Step 604: Identify influencing factors based on the difference feature vectors, and obtain the degree of influence matrix through correlation analysis.

[0187] Specifically, a correlation model between the PMC index and various influencing factors can be established using methods such as multiple regression analysis or structural equation modeling. Through model analysis, the degree of influence of each influencing factor on the PMC index is calculated, forming an influence degree matrix.

[0188] Step 605: Use the influence degree matrix to extract features and obtain key influencing factors through attribution analysis.

[0189] Specifically, based on the influence matrix, the key factors that have the greatest impact on the PMC index are identified. These key factors are the focus of optimization recommendations.

[0190] Based on the PMC index evaluation results, adjacent text data is extracted using the PMC surface graph, and the rate of change characteristics are obtained through index comparison. This step first requires a detailed interpretation of the PMC surface graph to identify key feature points and regions within the surface. The system automatically extracts adjacent data points of the PMC index on the time axis and calculates the rate of change and trend of the index. Specifically, the system focuses on time periods when the PMC index shows significant increases or decreases, calculating the rate of change for these time periods and expressing it as a percentage or slope. For time periods with large rates of change (such as intervals with rates of change exceeding ±10%), the system marks them as key observation periods for focused analysis. Simultaneously, the system compares the consistency of PMC index trends across different levels to determine whether macro-level development directions are synchronized with regional implementation. In this way, the system can capture important turning points and trends in the development of the Internet of Things (IoT) industry, providing a foundation for subsequent analysis. In practical applications, the system combines background knowledge of industry development to conduct a preliminary interpretation of these rate of change characteristics, such as identifying cyclical change characteristics that may be related to technological innovation cycles and industry investment cycles.

[0191] Based on the aforementioned rate of change characteristics, a hierarchical analysis is performed, and a consistency score is obtained through matching degree calculation. In this step, the system further analyzes the level of development consistency between different levels. Specifically, the system calculates the matching degree of the PMC index at the macro and regional levels, which can be quantified using correlation coefficients or difference indices. The correlation coefficient is calculated using the Pearson correlation coefficient method, with a value range of [-1, 1]. The closer to 1, the more consistent the development trends of the two levels. The difference index calculates the absolute or relative difference between the PMC indices of the two levels; the closer to 0, the closer the development levels of the two levels. The system calculates the matching degree index under different time windows (e.g., short-term, medium-term, long-term) to form consistency scores at multiple time scales. Furthermore, the system analyzes the level of consistency between different regions, assessing the balance and coordination of development between regions. These consistency scores provide a quantitative basis for identifying collaborative issues in the development of the IoT industry, helping decision-makers accurately grasp the collaborative status between different levels. In the specific implementation process, the system sets a threshold standard for the consistency score; for example, a score below 0.5 is considered low consistency, requiring focused attention and improvement.

[0192] The consistency score is used for discrepancy analysis, and a discrepancy feature vector is obtained through index comparison. Once regions or time periods with low consistency scores are identified, the system further analyzes the specific reasons for the discrepancies. In this step, the system compares and analyzes specific evaluation indicators at different levels, calculates the difference values ​​of each indicator, and forms a discrepancy feature vector. Each element of the discrepancy feature vector represents the degree of difference of a specific indicator, which can intuitively reflect in which aspects there are significant differences at different levels. For example, the system may find a large gap between macro and regional levels in terms of investment in technological innovation; or in terms of the construction of standards systems, the pace of progress is inconsistent between different regions. The system will sort the discrepancy feature vectors and identify the discrepancy indicators that contribute the most (e.g., the difference value exceeds twice the average difference). These indicators are often the key factors leading to a decline in the overall consistency level. In this way, the system can select several key indicators that need to be focused on from numerous indicators, providing precise direction for subsequent improvement suggestions. In practical applications, the system will also combine professional knowledge of the development of the Internet of Things industry to reasonably explain these differences, such as some differences being normal phenomena caused by differences in regional industrial foundations or development stages.

[0193] Based on the aforementioned difference feature vectors, influencing factors are identified, and an influence degree matrix is ​​obtained through correlation analysis. After identifying the key difference indicators, the system needs to further analyze the underlying reasons for these differences. In this step, the system constructs a correlation model between the PMC index and various influencing factors, employing statistical methods such as multiple regression analysis, path analysis, or structural equation modeling to explore the influence path and degree of each factor on the PMC index. The system first establishes a set of candidate influencing factors from two dimensions: internal factors of IoT development (such as technological innovation capabilities, industrial chain completeness, and human resources) and external factors (such as changes in market demand, capital investment, and the competitive environment). Then, through data mining and model calculation, the system determines the correlation between these factors and changes in the PMC index, calculates the influence coefficient of each factor, and forms an influence degree matrix. The rows of this matrix represent different influencing factors, the columns represent different evaluation indicators or different regions, and the matrix elements represent the degree of influence of the corresponding influencing factor on a specific indicator or region. By analyzing the influence degree matrix, the system can identify the key factors that have the most significant impact on the consistency of IoT industry development, providing a scientific basis for developing targeted improvement measures. In the specific implementation process, the system will use multiple analysis methods for cross-validation to improve the reliability of the results. For example, it will use correlation analysis and causal inference techniques at the same time to ensure that the identified influence relationships have practical significance rather than just statistical correlation.

[0194] Feature extraction is performed using the aforementioned influence degree matrix, and key influencing factors are identified through attribution analysis. This step deepens and refines the influencing factor analysis; the system further extracts and summarizes key influencing factors based on the influence degree matrix. In practice, the system first performs factor clustering analysis, grouping similar influencing factors into the same category to reduce complexity. Then, the system calculates the comprehensive influence of each category of factors, identifying the key factors with the greatest impact on the PMC index. For these key factors, the system conducts attribution analysis to deeply explore the specific mechanisms and paths by which each factor influences the development of the IoT industry. Attribution analysis is conducted from both qualitative and quantitative perspectives: quantitative analysis mainly calculates the contribution rate and sensitivity coefficient of each factor to changes in the PMC index; qualitative analysis combines theoretical knowledge and practical experience in the development of the IoT industry to explain the mechanisms of action of each factor. Through this comprehensive analysis, the system can generate a structured list of key influencing factors, including the importance ranking of each factor, explanation of the influence mechanism, and correlation path analysis. These key influencing factors become the direct basis for subsequent development and improvement recommendations, ensuring the relevance and effectiveness of the recommendations. In practical applications, the system will generate differentiated key influencing factor analysis results for IoT industries in different regions and at different stages of development, thereby improving the accuracy of the analysis.

[0195] Based on key influencing factors, the system generates targeted suggestions for improving the development of the Internet of Things (IoT) industry, drawing upon the aforementioned analysis results. These suggestions cover multiple dimensions, including recommendations for adjusting development goals, optimizing resource allocation, achieving breakthroughs in key areas, and improving collaborative mechanisms. In the development goal dimension, the system proposes a more coordinated approach to setting development goals based on consistency analysis across different levels, ensuring that the macro-development direction aligns with regional realities. In the resource allocation dimension, based on influencing factor analysis, the system recommends prioritizing limited resources in key areas that have the greatest impact on improving consistency, such as talent cultivation, technology research and development, or standards development. In the key areas dimension, the system focuses on identified bottlenecks and proposes breakthrough solutions, such as strengthening standards development and promotion to address the issue of inconsistent technical standards. In the collaborative mechanism dimension, based on the analysis of differences between levels and regions, the system proposes specific measures to improve information communication and strengthen collaborative actions. These suggestions are not only based on data analysis results but also incorporate professional knowledge and best practices in IoT industry development, demonstrating strong practicality and foresight. The system categorizes suggestions based on priority and implementation difficulty, creating three levels: short-term action suggestions, medium-term improvement suggestions, and long-term development suggestions, providing decision-makers with clear action guidelines. In practical application, the system continuously optimizes the suggestions based on implementation feedback, forming a closed-loop improvement mechanism.

[0196] A dynamic feedback mechanism is established to continuously optimize the evaluation results. To ensure the timeliness and accuracy of the evaluation methods and results, a complete dynamic feedback mechanism is established. First, the system regularly collects new IoT industry development data and updates the text database and feature analysis results. Second, the system recalculates the PMC index based on the new data and updates the PMC surface diagram and consistency evaluation results. Third, the system tracks the implementation and effects of previously proposed improvement suggestions and evaluates the effectiveness of the suggestions. Finally, the system continuously adjusts the evaluation algorithm parameters and suggestion generation strategy based on implementation feedback to improve the system's adaptability and accuracy. This dynamic feedback mechanism makes the evaluation of IoT industry development a continuous optimization process, rather than a one-time static assessment. As data accumulation increases and algorithm optimization deepens, the system's evaluation accuracy and suggestion quality will continuously improve, providing long-term and stable decision support for the healthy development of the IoT industry. In addition, the system also supports customized analysis functions, allowing decision-makers to set evaluation priorities and parameters according to specific needs, obtaining evaluation results and improvement suggestions that better meet actual needs. This flexibility further enhances the system's practical value, enabling it to adapt to the evaluation needs of the IoT industry in different regions and at different stages of development.

[0197] In one embodiment, the process of conducting a system evaluation based on the PMC index evaluation results using a pre-defined consistency analysis model to generate development and improvement recommendations may further include:

[0198] Step 701: Obtain the evaluation results of the PMC index and establish an adaptive mind mapping framework.

[0199] Specifically, the Adaptive Graph of Thought (AGoT) framework is an innovative extension of the traditional Chain of Thought (CoT) and Tree of Thought (ToT) methods. Unlike the linear CoT and the branching ToT, AGoT employs a dynamic directed acyclic graph (DAG) structure, which can adaptively adjust the reasoning path according to the complexity of the problem and the needs of the analysis.

[0200] In the analysis of IoT industry development, the AGoT framework uses Graph Neural Networks (GNN) technology to construct the relationships between development elements. Each development element is represented as a node in the graph, and the dependencies between elements are represented as edges. Through an attention mechanism, the system can dynamically adjust the weights of connections between nodes, highlighting the impact of key development elements.

[0201] In this step, complete evaluation results are first obtained from the PMC index evaluation database, including PMC index values ​​and their constituent elements at each time point and region. The Adaptive Graph of Thought (AGoT) framework is a key innovation of this invention. It breaks through the limitations of traditional linear thought chains and tree-like thought structures, adopting a more flexible dynamic directed acyclic graph structure. In specific implementation, the system constructs an initial knowledge graph based on the PMC index evaluation results. This graph uses various elements of the IoT industry development as nodes and the relationships between elements as edges, forming a network structure. The system uses graph neural network technology to perform representation learning on this structure, enabling each node to obtain a vector representation containing rich semantic information. Unlike traditional knowledge graphs, the AGoT framework is adaptive, dynamically adjusting the graph structure and reasoning path according to the complexity of the problem and the analysis requirements. For example, when analyzing IoT technology standard consistency issues, the system automatically strengthens the weights of nodes and edges related to standards; while when analyzing industry chain collaboration issues, it adjusts to highlight the connections between upstream and downstream enterprises. This adaptive mechanism is achieved through an attention mechanism. The system calculates the importance weights of different nodes and edges based on the current analysis task, forming an optimized mind map for a specific problem.

[0202] Step 702: Use the adaptive mind mapping framework to decompose the complex problem into interrelated subproblems.

[0203] Specifically, in this process, the AGoT framework uses recursive neural networks and the Transformer architecture to automatically decompose the problem. The system first represents the text as a high-dimensional vector and extracts semantic features using pre-trained language models such as BERT or T5.

[0204] Then, the complex problem is broken down into a set of sub-problems using the Question Decomposition Module (QDM). QDM is based on a reinforcement learning strategy, using the final analysis quality as a reward signal to continuously optimize the decomposition strategy.

[0205] The decomposed subproblems form a DAG structure, and the connection weights between nodes are dynamically adjusted through an Edge Attention Network.

[0206] Based on the established adaptive mind mapping framework, the complex evaluation problem in the development of the Internet of Things (IoT) industry can be decomposed into a series of interconnected but more specific sub-problems. This decomposition process is implemented using a combination of recursive neural networks and the Transformer architecture. First, the system semantically represents the problem points (such as areas or indicators with low consistency levels) found in the PMC evaluation results, transforming them into representations in a high-dimensional vector space. Then, the Question Decomposition Module (QDM) further breaks down the complex problem into a set of sub-problems. For example, when facing the complex problem of "insufficient collaboration in the IoT industry chain," the system automatically decomposes it into multiple sub-problems such as "upstream chip supply stability," "midstream module standardization," and "downstream application scenario diversity." QDM employs a reinforcement learning strategy, using the final analysis quality as a reward signal to continuously optimize the decomposition strategy, ensuring that the decomposed sub-problems are relatively independent and analyzable while collectively covering the core elements of the original problem. During the decomposition process, the system preserves the logical relationships between sub-problems, forming a directed acyclic graph (DAG) representing the dependencies between problems. This structure expresses the sequential dependencies between sub-problems while avoiding analytical loops caused by circular dependencies. Through this intelligent problem decomposition, the system can transform the massive and complex evaluation problem of the development of the Internet of Things industry into a manageable network of sub-problems, laying the foundation for subsequent accurate analysis.

[0207] Step 703: Construct a directed acyclic graph structure, prioritize the subproblems, and analyze the subproblems according to the priority ranking to obtain the analysis results of the subproblems.

[0208] Specifically, the AGoT framework combines the Multi-Armed Bandit (MAB) algorithm with the Thompson sampling method to dynamically adjust the analysis priority of subproblems. For each subproblem node pi, the system maintains its expected analytical value vi and uncertainty σi.

[0209] Each time a node is selected for analysis, the system samples from the Beta distribution Beta(vi + 1, n - vi + 1), where n is the total number of attempts, and selects the node with the highest sample value for analysis. This method balances the relationship between "exploration" (trying new analysis paths) and "utilization" (deeply analyzing known important paths).

[0210] After decomposing the problem, the system needs to rationally arrange the analysis order of the sub-problems to ensure the efficiency and accuracy of the analysis process. The directed acyclic graph (DAG) structure constructed by the system not only represents the dependencies between sub-problems but also provides a basic framework for priority ranking. In the priority ranking stage, the system adopts a strategy combining the Multi-Armed Bandit (MAB) algorithm and the Thompson sampling method to dynamically adjust the priority order of analysis tasks. Specifically, for each sub-problem node pi, the system maintains its expected analysis value vi (representing the contribution of solving the sub-problem to the overall analysis) and uncertainty σi (representing the confidence level of the value estimate of the sub-problem). When selecting the next sub-problem to be analyzed, the system samples from the Beta distribution Beta(vi+1, n-vi+1), where n is the total number of attempts, and then selects the sub-problem with the highest sample value for analysis. This method cleverly balances the relationship between "exploration" (attempting to analyze new sub-problem paths) and "utilization" (deeply analyzing known important sub-problems), maximizing the analysis effect with limited computing resources. Following a predetermined priority order, the system analyzes each sub-problem one by one, with each sub-problem being processed using a dedicated analysis module. For example, for sub-problems related to technical consistency, the system invokes the standard analysis engine; for sub-problems related to regional collaboration, it invokes the regional comparison analysis engine. After each sub-problem is analyzed, the system stores the results in a result pool and updates the node states and value estimates of related dependent nodes in the directed acyclic graph, providing a reference for subsequent analyses.

[0211] Step 704: Integrate the analysis results of the sub-problems to form the development and improvement suggestions.

[0212] Specifically, the AGoT framework integrates the analysis results of various sub-problems to form systematic development optimization suggestions. By analyzing the dependency graphs between development elements, the system can identify key nodes and weak links in the development system, thereby proposing more targeted improvement measures.

[0213] After analyzing all sub-problems, the scattered results need to be integrated into systematic development and improvement recommendations. This integration process is not a simple result patching, but rather involves semantic understanding and reconstruction of the results using deep learning models. The system employs Graph Attention Networks (GAT) to process the sub-problem analysis results. This network can automatically adjust the weights and combinations of different results based on the dependencies and content relevance between sub-problems. The integration process is divided into three levels: First, problem-level integration, where the system integrates directly related sub-problem results according to the structure of a directed acyclic graph and logical dependency paths; second, domain-level integration, where the system cross-domains the analysis results from different technical fields or industry segments, identifying cross-domain related problems and solutions; and finally, strategic-level integration, where the system maps the analysis results from all domains to the strategic dimension of IoT industry development, forming global improvement recommendations. During the integration process, the system automatically detects consistency and conflicts among the results of each sub-problem, reconciling conflicting content to ensure the internal consistency of the final recommendations. Simultaneously, the system supplements the theoretical basis and implementation paths of key recommendations based on a professional knowledge base for IoT industry development, improving the operability of the recommendations. The final development and improvement recommendations provide guidance on macro-level development directions as well as detailed measures for specific areas. They consider both short-term improvement priorities and long-term development strategies, providing comprehensive decision-making support for the development of the Internet of Things (IoT) industry.

[0214] Step 705: Continuously update the data of the PMC index evaluation results and optimize the problem decomposition and analysis strategies.

[0215] Specifically, in establishing the dynamic feedback mechanism, the AGoT framework continuously learns from new evaluation results and constantly optimizes its problem decomposition and analysis strategies. By accumulating analytical experience, the system gradually improves the accuracy of problem diagnosis and the feasibility of optimization suggestions.

[0216] To ensure the timeliness and accuracy of the system's analysis, this invention establishes a complete dynamic update mechanism. The system regularly collects the latest IoT industry development data, updates the PMC index evaluation results, and adjusts the structure and parameters of the adaptive mind map based on the new evaluation results. In the data update phase, the system employs an incremental learning method, requiring only the recalculation of the graph portion related to the new data, avoiding the computational overhead of full graph reconstruction. Simultaneously, the system records performance indicators for each problem decomposition and analysis process, such as analysis accuracy and suggestion adoption rate, and uses these indicators as reward signals for reinforcement learning, continuously optimizing problem decomposition strategies and analysis methods. As data accumulation and analytical experience increase, the system's AGoT framework becomes increasingly intelligent, enabling it to more accurately capture key issues and potential opportunities in the development of the IoT industry. Furthermore, the system introduces a self-learning module, which learns improvement directions from feedback from industry experts on the analysis results, continuously refining its knowledge base and analytical models. This continuous optimization mechanism allows the system to adapt to the rapid changes in IoT technology and industry, maintaining the forward-looking and practical nature of its analysis. In practical applications, the system supports differentiated update frequency settings, adjusting the data update cycle according to the development speed of different regions and fields, ensuring efficient utilization of analytical resources.

[0217] In another embodiment, the process of conducting a system evaluation based on the PMC index evaluation results using a preset consistency analysis model to generate development and improvement recommendations may further include:

[0218] Construct a policy language agent system and establish a latent space model for policy representation.

[0219] Specifically, this module first establishes a strategy language agent system capable of understanding and generating development-related strategic language. Specifically, the system takes PMC index evaluation results and Adaptive GoT (AGoT) analysis results as input to construct a latent space for strategy representation. This latent space contains various possible strategies in the evaluation, optimization, and implementation processes.

[0220] In the intelligent analysis framework for evaluating the development of the Internet of Things (IoT) industry, the policy language proxy system is a crucial link connecting analysis results with specific action recommendations. This system first constructs a natural language processing engine specifically designed to express and understand IoT industry development strategies, capable of transforming abstract analytical conclusions into concrete and feasible policy statements. The system uses a pre-trained language model (such as BERT or T5) as its foundation, fine-tuning it on IoT literature and case data to enable the model to understand and generate IoT terminology and policy language. Based on this, the system establishes a latent space model for policy representation, mapping various possible IoT development strategies to a high-dimensional continuous vector space. Unlike traditional discrete policy enumeration, the latent space model can represent an infinite number of policy variations and capture semantic similarities and combinatorial patterns between policies. This latent space is implemented using variational autoencoder (VAE) technology; the encoder converts the policy statements in natural language into latent vectors, and the decoder reconstructs these latent vectors into specific policy statements. The system also introduces a conditional control mechanism, allowing for targeted constraints on the distribution of the latent space based on specific environmental conditions (such as regional characteristics and development stages), thereby generating strategies more suitable for specific scenarios. This latent space representation method not only greatly expands the policy space that the system can explore, but also provides continuously differentiable optimization objectives for subsequent policy optimization, making the policy generation and improvement process more flexible and efficient.

[0221] Among them, the latent space model is a hybrid generative model based on variational autoencoders (VAEs) and generative adversarial networks (GANs), specifically designed for representing and generating development strategies for the Internet of Things (IoT) industry. Unlike traditional discrete policy representations, this model maps policies to a continuous high-dimensional latent space, enabling it to capture semantic relationships and innovative possibilities between policies.

[0222] The core architecture of the latent space model consists of three parts: encoder E, decoder D, and discriminator C. Encoder E maps the input policy s to a latent space vector z, i.e., z = E(s); decoder D reconstructs the latent space vector z into a policy representation s', i.e., s' = D(z); and discriminator C is responsible for distinguishing between generated and real policies, improving generation quality. The overall optimization objective function is:

[0223] L = L recon + λ1·L KL + λ2·L adv

[0224] Where L recon It is the reconstruction loss, ensuring the fidelity of the encoding-decoding process; L KLIt is the KL divergence loss, which makes the latent distribution approximate the pre-defined prior distribution (usually the standard normal distribution); L adv This is to counteract losses and improve the realism and diversity of the generation strategy. λ1 and λ2 are the weighting coefficients that balance the various loss terms.

[0225] To enhance the model's understanding of IoT domain knowledge, the latent space model introduces a domain knowledge constraint mechanism. The system constructs a knowledge graph G for IoT industry development, containing core concepts, relationships, and rules. During encoding, for an input policy s, the system calculates its consistency score cG(s) with the knowledge graph and incorporates this score as an additional feature into the latent representation z. During decoding, a graph guidance layer is added to ensure that the generated policies conform to domain knowledge specifications. This mechanism significantly improves the professionalism and feasibility of the model's generated policies.

[0226] Another innovation of the latent space model is the condition generation mechanism. The model supports multiple condition controls, including: (1) regional condition cr, which enables the generation strategy to adapt to the industrial development characteristics of different regions; (2) stage condition cs, which generates differentiated strategies for different development stages of the Internet of Things; and (3) target condition co, which adjusts the strategy direction according to specific development goals. Condition control is achieved by introducing condition embedding in the encoder and decoder, and the condition vector is concatenated with the original feature vector to participate in the calculation.

[0227] To support policy exploration and innovation, the model employs a set of latent space operation tools, including: vector interpolation (generating transitional policies between two known policies), vector arithmetic (such as A - B + C, replacing a characteristic of policy A with the corresponding characteristic of policy C), region clustering (discovering clusters of similar policies), and novelty search (finding innovative policies in low-density regions). These tools enable decision-makers to systematically explore the policy space and discover innovative solutions that are difficult to access through conventional thinking.

[0228] Through this complex latent space representation, the system can handle the high-dimensional, multi-factor strategy generation problem in the development of the Internet of Things industry, providing richer and more flexible strategy choices for decision-making.

[0229] The strategy combinations sampled from the latent space model are used to obtain strategy combinations, which are then applied to the evaluation and optimization process.

[0230] Specifically, this module employs an iterative optimization algorithm to continuously improve the performance of the policy language agent. In each iteration, the system first samples possible policy combinations in the latent space and applies these policies to the evaluation and optimization process. By observing the effects of policy execution (such as changes in the PMC index), the system learns to update the policy distribution, gradually converging towards a better policy space.

[0231] With the latent space model of the policy representation, the system needs to intelligently sample from this high-dimensional space to generate effective policy combinations for evaluation and optimization in the development of the Internet of Things (IoT) industry. The system employs a particle-based policy search (PPS) algorithm for policy sampling. This algorithm maintains a set of particles {z} i In this system, each particle represents a policy candidate point in the potential space. The system first initializes multiple particles (typically 100-500) from a prior distribution (usually a standard normal distribution), then uses a decoder to convert each particle into a corresponding policy representation. To improve sampling efficiency, the system employs importance sampling, adjusting the sampling distribution based on historical policy evaluation results to concentrate sampling more on areas with potentially high returns. The system also implements a hierarchical sampling strategy, first sampling at the policy category level (e.g., selecting policy directions such as technological innovation, talent cultivation, or standard promotion), and then refining the sampling at the specific measures level. This hierarchical structure ensures that the final policy combination has both a clear strategic direction and actionable details. The sampled policy combinations undergo initial screening to remove obviously unreasonable or conflicting policies, and are then applied to the evaluation and optimization process for the development of the Internet of Things (IoT) industry. During application, the system simulates the implementation effects of these policies in different scenarios, assesses their potential impact on the PMC index, and further adjusts the policy combination based on the evaluation results. Through this iterative sampling-evaluation-adjustment process, the system can gradually find the optimal policy combination scheme.

[0232] A multi-level strategy feedback mechanism is established to evaluate the execution effect of the strategy combination and obtain the strategy execution evaluation result.

[0233] Specifically, this module constructs a multi-layered policy feedback mechanism. The system compares the results of policy execution with the expected goals, identifying successful and unsuccessful policy patterns. This information is used to update the distribution of the latent space, enabling the system to generate more effective policies in subsequent iterations. In particular, the system focuses on policy combinations with significantly improved PMC indices, using them as high-quality learning samples.

[0234] Generating strategy combinations is only the first step; more crucially, it's about evaluating the effectiveness of these strategies to provide a basis for subsequent optimization. The system establishes a multi-layered strategy feedback mechanism to evaluate strategy effectiveness from different dimensions and time scales. In terms of evaluation dimensions, the system sets up three main levels: first, a technical level evaluation, primarily examining the strategy's improvement effect on core IoT technical indicators, such as the speed of technological innovation, the degree of standardization, and platform compatibility; second, an industry level evaluation, focusing on the strategy's impact on the IoT industry structure and operational efficiency, such as supply chain synergy, enterprise innovation activity, and market application penetration; and finally, an ecosystem level evaluation, assessing the strategy's impact on the health of the entire IoT ecosystem, such as diversity, stability, and sustainability. In terms of time scale, the system simultaneously conducts short-term effect assessments (within one year), medium-term effect predictions (1-3 years), and long-term impact projections (over 3 years), comprehensively grasping the immediate benefits and lasting value of the strategies. The evaluation process employs a combination of methods, including data-driven statistical analysis, simulation predictions based on IoT industry development theoretical models, and qualitative judgments supported by an expert knowledge base. The system generates detailed evaluation reports for each strategy combination, including quantified performance metrics, visualized impact path analysis, and identification of key success factors. These evaluation results not only guide the adjustment and improvement of current strategies but are also stored as valuable experiential data in the system's knowledge base, providing a reference for future strategy generation.

[0235] The latent space distribution of the latent space model is updated based on the evaluation results of the strategy execution, and an optimization strategy is generated.

[0236] Specifically, the system employs a particle-based policy search (PPS) algorithm for policy sampling. This algorithm maintains a set of particles {z}. i In this system, each particle represents a policy candidate point in the latent space. The system first initializes N particles (typically N = 100-500) from a prior distribution. Then, each particle is converted into its corresponding policy representation using a decoder D(z).

[0237] The strategy evaluation uses a multi-objective evaluation function, comprehensively considering factors such as the degree of improvement in the PMC index, implementation costs, and time efficiency.

[0238] R(z) = w1·ΔPMC(z) + w2·(1 / Cost(z)) + w3·(1 / Time(z))

[0239] Where w1, w2, and w3 are weighting coefficients, satisfying w1 + w2 + w3 = 1.

[0240] Based on the evaluation results of policy execution, the system needs to continuously update and optimize the policy generation model to generate more effective development policies. The core of this step is adjusting the distribution parameters of the latent space model to ensure that regions containing efficient policies receive a higher sampling probability. The system employs a gradient-based policy optimization method, using the policy evaluation score as the objective function. It calculates the gradient direction of the latent space parameters using the backpropagation algorithm and then updates the distribution parameters along the gradient direction. To prevent premature convergence to local optima, the system introduces a temperature annealing mechanism. A high exploration temperature is maintained in the early stages of optimization, and the temperature is gradually reduced as the number of iterations increases, enhancing the utilization of regions containing efficient policies. Simultaneously, the system also employs ensemble learning, maintaining multiple independent latent space models, each focusing on different types of policy generation. The outputs of each model are then fused through weighted ensemble, improving the diversity and robustness of policy generation. The updated latent space model generates a new generation of optimized policies. These policies retain the core characteristics of the original efficient policies while incorporating new innovative elements, forming a more complete policy system. The system regularly evaluates the performance metrics of the latent space model, such as policy diversity, generation quality, and optimization convergence speed, and adjusts the model architecture and training parameters based on these metrics to ensure that the model is always in the best state.

[0241] By comprehensively optimizing evaluation, analysis, and implementation strategies, overall performance can be improved.

[0242] Specifically, this module develops a cross-step strategy optimization mechanism. It can simultaneously optimize evaluation, analysis, and implementation strategies, thereby improving overall performance. By sharing learned strategy knowledge across different steps, the system can generate more coordinated and efficient decision recommendations.

[0243] In the complete closed loop of IoT industry development evaluation and optimization, it is necessary not only to optimize specific development strategies, but also to comprehensively optimize evaluation methods, analysis processes, and implementation mechanisms to form a systematic improvement plan. The system adopts a meta-learning framework, treating evaluation strategies, analysis strategies, and implementation strategies as three interrelated learning tasks, improving overall performance through shared feature representations and collaborative optimization. Regarding evaluation strategy optimization, the system dynamically adjusts the parameter weights of the PMC index calculation model based on new data and feedback, optimizes the structural design of the indicator system, and improves the accuracy and sensitivity of the evaluation. Regarding analysis strategy optimization, the system continuously improves the problem decomposition algorithm and priority ranking mechanism of the AGoT framework, optimizes the dependency structure between sub-problems, and improves the efficiency and depth of analysis. Regarding implementation strategy optimization, the system improves the expressive power and contextual understanding capabilities of the policy language agent, improves policy sampling and evaluation methods, and enhances the operability and adaptability of the generated policies. These three optimizations are not carried out in isolation; the system designs a cross-policy knowledge transfer mechanism, enabling the optimization results of each stage to learn from and promote each other. For example, new indicator correlation patterns discovered in the evaluation strategy can guide the problem decomposition of the analysis strategy; key influencing factors identified in the analysis strategy can guide the focus of the implementation strategy; and the feedback effect of the implementation strategy can verify the effectiveness of the evaluation strategy. Through this cyclical improvement mechanism, the system can continuously improve itself during continuous use, adapt to the new characteristics and needs of the Internet of Things industry development, and provide increasingly accurate development evaluations and suggestions.

[0244] This architecture maintains the independence of the original steps while enabling intelligent collaboration among them through policy language proxies, thereby significantly improving the overall performance of the system.

[0245] In yet another embodiment, the method further includes:

[0246] Obtain the evaluation indicators from the multi-input-output table and establish a mixed linear regression model.

[0247] Specifically, mixed linear regression (MLR) models are an effective method for handling data with heterogeneity and multimodal distributions. In the data analysis of IoT development, different levels and periods of development often exhibit different characteristics, and simple linear models are difficult to accurately capture this heterogeneity. The MLR model treats the data as a mixture of K different linear models, where the probability of each data point belonging to a specific model is determined by its features.

[0248] Formal representation:

[0249] y = <w k , x>+ε

[0250] k ~ Multinomial(π1,...,π k )

[0251] Where y is the PMC index, x is the development characteristic vector, and w k Let ε be the parameter vector of the k-th model, ε be the noise term, and π be the parameter vector of the k-th model. k represents the mixed weights of the k-th model.

[0252] A technique combining tensor decomposition and spectral methods is employed to ensure global convergence.

[0253] Specifically, traditional hybrid linear regression methods often employ the Expectation-Maximization (EM) algorithm, but it is prone to getting trapped in local optima under typical data conditions. This step uses a combination of tensor decomposition and spectral methods to ensure global convergence under certain conditions.

[0254] Specifically, the algorithm first constructs the third-order moment tensor of the development features:

[0255]

[0256] in This represents the tensor product operation.

[0257] Then, the parameters of each component are estimated using tensor decomposition:

[0258]

[0259] To ensure global convergence, the algorithm introduces a sampling mechanism based on stochastic gradient Langevin dynamics (SGLD):

[0260]

[0261] Where η t Let Ση be a step-size sequence. t =∞ and Ση t ²<∞,ξ t It is standard Gaussian noise.

[0262] This mechanism can escape local optima in non-convex optimization problems, theoretically guaranteeing convergence to the global optimal solution.

[0263] Key parameters of the PMC index are estimated using a gradient-based stochastic approximation method.

[0264] Specifically, in the calculation of the PMC index, key parameters include the text strength coefficient α, the implementation period parameter β, and the consistency coefficient γ. This step employs a gradient-based stochastic approximation method, adaptively adjusting these parameters through stochastic approximation techniques.

[0265] This method is an extension of the Robbins-Monro algorithm, which expresses the parameter update as follows:

[0266]

[0267] Where θ represents the parameter vector (α, β, γ), Q is the performance metric function, and ξ t It is a random perturbation.

[0268] To address the uncertainty in parameter estimation, the system introduces an adaptive noise injection mechanism:

[0269]

[0270] in Let be the initial step size, and a and b be hyperparameters controlling the decay rate.

[0271] Using Bayesian nonparametric mixture models to handle nonlinear relationships in data.

[0272] Specifically, to flexibly handle potential nonlinear relationships in the development data, this step introduces a Bayesian nonparametric mixture model framework. This framework uses a Dirichlet Process (DP) to automatically determine the optimal number of models, K.

[0273]

[0274] Where G0 is the base distribution, usually chosen as a normal-inverse gamma distribution, and α is the concentration parameter.

[0275] The model inference uses a combination of Gibbs sampling and variational inference, which ensures both inference accuracy and computational efficiency.

[0276] A parameter learning algorithm is constructed to ensure the convergence of the hybrid linear regression model, and the PMC index evaluation result containing the PMC index value is obtained.

[0277] Specifically, this step uses the Kiefer-Wolfowitz stochastic approximation method to estimate key parameters in the PMC index calculation. Compared with the fixed parameters in the original technical solution, the parameters of the adaptive estimation are more flexible: the text strength coefficient α is adjusted from a fixed value {1.0, 0.8, 0.6} to a range that allows continuous variation [0.1, 1.0]; the implementation period parameter β is expanded from a fixed {2,3, 5} to a more refined continuous estimate; and the consistency coefficient γ is also changed from empirically determined to a data-driven adaptive estimate.

[0278] Alternatively, the IoT industry development evaluation method provided by this invention can be implemented by an IoT industry development evaluation device. Figure 3 An exemplary embodiment of the IoT industry development evaluation device of the present invention is shown. This device can be implemented through software, hardware, or a combination of both.

[0279] Please see Figure 3 , Figure 3 This is a structural block diagram of an IoT industry development evaluation device provided in an embodiment of the present invention. The device includes:

[0280] The data acquisition module 801 acquires data sources related to the development plan, and uses text mining technology to collect and organize the data sources to generate a standardized text database.

[0281] The text processing module 802 is used to perform word segmentation and stop word removal on the original text in the standardized text database, and to establish a text feature set by statistical word frequency distribution and feature extraction to obtain text feature results.

[0282] The indicator construction module 803 is used to design indicators based on the text feature results and using an evaluation framework that includes development goals, support tools, and supporting measures, and to construct a multi-input-output table.

[0283] The index calculation module 804 is used to process data using the evaluation indicators in the multi-input-output table and a preset PMC index calculation model to obtain the PMC index evaluation result.

[0284] The optimization suggestion module 805 is used to conduct a system evaluation based on the evaluation results of the PMC index through a preset consistency analysis model, and form development and improvement suggestions.

[0285] The modules described above can implement the functions of the corresponding steps in the method of the present invention.

[0286] In summary, the IoT industry development evaluation method and apparatus provided in this embodiment of the invention solves the problems of strong subjectivity, low evaluation accuracy, and lack of systematic evaluation framework in the evaluation of IoT industry development by constructing an evaluation framework based on the PMC index. It can provide more comprehensive, objective, and scientific evaluation and improvement suggestions for the development of IoT industry.

[0287] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0288] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0289] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0290] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0291] In a typical configuration, a computing device includes one or more processors (CPUs), input / output interfaces, network interfaces, and memory. Memory may include non-persistent storage in computer-readable media, such as random access memory (RAM) and / or non-volatile memory, such as read-only memory (ROM) or flash RAM. Memory is an example of computer-readable media.

[0292] Computer-readable media, including both permanent and non-permanent, removable and non-removable media, can store information using any method or technology. Information can be computer-readable instructions, data structures, modules of programs, or other data. Examples of computer storage media include, but are not limited to, phase-change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, CD-ROM, digital versatile optical disc (DVD) or other optical storage, magnetic tape, magnetic tape, disk storage or other magnetic storage devices, or any other non-transferable medium that can be used to store information accessible by a computing device. As defined herein, computer-readable media does not include transient computer-readable media, such as modulated data signals and carrier waves.

[0293] It should also be noted that the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0294] The foregoing has provided a detailed description of the IoT industry development evaluation method and apparatus provided by the embodiments of the present invention. Specific examples have been used to illustrate the principles and implementation methods of the present invention. The descriptions of the embodiments above are merely for the purpose of helping to understand the method and core ideas of the present invention. It should be noted that those skilled in the art can make various improvements and modifications to the present invention without departing from its principles, and these improvements and modifications also fall within the protection scope of the claims of the present invention.

Claims

1. A method for evaluating the development of the Internet of Things (IoT) industry, characterized in that, include: Obtain data sources related to development planning, and use text mining technology to collect and organize the data sources to generate a standardized text database; The original text in the standardized text database is segmented and stop words are removed. A text feature set is established by statistical word frequency distribution and feature extraction to obtain the text feature results. Based on the text feature results, an evaluation framework encompassing development goals, support tools, and supporting measures is used to design indicators and construct a multi-input-output table. Using the evaluation indicators in the aforementioned multi-input-output table, data processing is performed through a preset PMC index calculation model to obtain the PMC index evaluation results, including: Coefficient analysis is performed using text type and issuing unit level information, and the text intensity coefficient α is calculated through hierarchical evaluation. Based on the text intensity coefficient α, implementation period data is analyzed, and the implementation cycle parameter β is obtained through time-series rule processing. The implementation cycle parameter β is used to classify indicators in a multi-input-output table, and input-output variable sets are obtained through a variable identification algorithm. Data standardization is performed based on the input-output variable sets, and a standardized indicator matrix is ​​obtained through dimension conversion. Using the standardized indicator matrix and coefficient parameters, a PMC index evaluation result containing PMC index values ​​is obtained through a preset PMC calculation model. Based on the PMC index evaluation results, a system evaluation is conducted using a pre-defined consistency analysis model to generate development and improvement recommendations, including: Based on the PMC index evaluation results, adjacent text data are extracted using a PMC surface plot, and the rate of change feature is obtained through index comparison; hierarchical analysis is performed based on the rate of change feature, and a consistency score is obtained through matching degree calculation; difference analysis is performed using the consistency score, and a difference feature vector is obtained through index comparison; influencing factors are identified based on the difference feature vector, and an influence degree matrix is ​​obtained through correlation analysis; feature extraction is performed using the influence degree matrix, and key influencing factors are obtained through attribution analysis. Also includes: Obtain the evaluation indicators from the multi-input-output table and establish a mixed linear regression model; A technique combining tensor decomposition and spectral methods is employed to ensure global convergence. Key parameters of the PMC index are estimated using a gradient-based stochastic approximation method. Using Bayesian nonparametric mixture models to handle nonlinear relationships in data; A parameter learning algorithm is constructed to ensure the convergence of the hybrid linear regression model, and the evaluation result of the PMC index containing the PMC index value is obtained.

2. The method according to claim 1, characterized in that, The process of collecting and organizing the data source using text mining technology to generate a standardized text database specifically includes: The data source is filtered using keyword indexing technology, and a multi-source data set is generated through automatic collection and processing. The multi-source data set is subjected to format recognition and encoding standardization processing to establish a standardized data index; Based on the standardized data index, data hierarchy is divided, and a multi-level data structure is generated through classification and organization. The multi-level data structure is used to design text identifiers, and a unique identifier sequence is generated through automatic encoding. The standardized text database is generated by integrating the data based on the unique identifier sequence and the multi-level data structure.

3. The method according to claim 1, characterized in that, The process of segmenting and removing stop words from the original text in the standardized text database specifically includes: Obtain the original text from the standardized text database, and use domain knowledge to build a professional dictionary database to form a basic word segmentation dictionary database; Based on the aforementioned word segmentation lexicon, text segmentation is performed to obtain initial word segmentation results; The initial word segmentation results are subjected to syntactic structure analysis, and a basic set of word groups is obtained through part-of-speech tagging. The aforementioned basic phrase set is used for synonym analysis and word form unification, and standard phrases are obtained through rule-based processing. Stop word identification and filtering are performed based on the standard word groups to obtain a set of valid word groups.

4. The method according to claim 1, characterized in that, Based on the text feature results, an evaluation framework encompassing development goals, supporting tools, and complementary measures is used to design indicators and construct a multi-input-output table, including: The feature vectors of the text feature results are obtained, and the evaluation framework is used to design indicators to form an evaluation indicator framework. Based on the aforementioned evaluation index framework, quantitative rules are designed, and index calculation standards are established through numerical mapping. The qualitative features are numericalized using the aforementioned index calculation standards, and quantitative data is obtained through transformation rules. Based on the quantified data, a matrix structure is designed, and a multidimensional data matrix is ​​obtained through dimensional organization. The input-output mapping relationship is established using the multidimensional data matrix, and a multi-input-output table is obtained through modeling.

5. The method according to claim 1, characterized in that, After obtaining the PMC index evaluation result containing the PMC index value by combining the standardized index matrix with coefficient parameters through a preset PMC calculation model, the process includes: The coordinate system is designed using the PMC index evaluation results, and a three-dimensional display space is constructed through dimension mapping. Data positioning is performed based on the three-dimensional display space, and a scattered point distribution set is obtained through spatiotemporal mapping. The scattered point distribution set is used for surface fitting, and the initial surface is obtained by cubic spline interpolation algorithm; Smoothness optimization is performed based on the initial surface, and the optimized surface model is obtained by adjusting the parameters. The optimized surface model is used for visualization processing, and the PMC surface diagram is obtained through color mapping and node annotation.

6. The method according to claim 1, characterized in that, Based on the evaluation results of the PMC index, a system assessment is conducted using a pre-set consistency analysis model to generate development and improvement recommendations, including: Obtain the evaluation results of the PMC index and establish an adaptive mind mapping framework; The adaptive mind mapping framework is used to break down complex problems into interconnected sub-problems; Construct a directed acyclic graph structure, prioritize the subproblems, and analyze the subproblems according to the priority ranking to obtain the analysis results of the subproblems; The analysis results of the aforementioned sub-problems are integrated to form the development and improvement recommendations. Continuously update the data of the PMC index evaluation results and optimize the problem decomposition and analysis strategies.

7. The method according to claim 1, characterized in that, Based on the evaluation results of the PMC index, a system assessment is conducted using a pre-set consistency analysis model to generate development and improvement recommendations, including: Construct a policy language agent system and establish a latent space model for policy representation; The strategy combination sampled from the latent space model is used to obtain a strategy combination, which is then applied to the evaluation and optimization process. Establish a multi-level strategy feedback mechanism to evaluate the execution effect of the strategy combination and obtain the strategy execution evaluation result; The latent space distribution of the latent space model is updated based on the evaluation results of the strategy execution, and an optimization strategy is generated. By comprehensively optimizing evaluation, analysis, and implementation strategies, overall performance can be improved.

Images