An intelligent interaction system based on multi-dimensional evaluation analysis

Abstract

The application belongs to the technical field of artificial intelligence, and particularly relates to an intelligent interaction system based on multi-dimensional evaluation analysis, comprising: a man-machine interaction intelligent agent, which is used for extracting a target object from the dialogue content of a user; an index analysis intelligent agent, which is used for calling the relevant information of the target object from an accessed database and / or knowledge base, and analyzing the multi-level indexes of the target object according to the relevant information; and an evaluation calculation intelligent agent, which is used for dynamically weighting the indexes of the target object by using a subjective and objective combination method, and calculating the individual evaluation result of each index of the target object and the comprehensive evaluation result of all indexes according to the weight after weighting; wherein the man-machine interaction intelligent agent is further used for feeding back the individual evaluation result and the comprehensive evaluation result of the target object and the calculation process of the evaluation calculation intelligent agent to the user for checking and confirming. The application adopts a dynamic weighting mechanism based on the subjective and objective combination method to adaptively weight the multi-level indexes of the target object, so that the pertinence and accuracy of the evaluation result are improved.

Description

Technical Field

[0001] This invention belongs to the field of artificial intelligence technology, specifically relating to an intelligent interactive system based on multidimensional evaluation analysis. Background Technology

[0002] In the field of comprehensive analysis and evaluation targeting multiple objects and dimensions, users typically need to collect information from multiple data sources and perform quantitative scoring and comprehensive calculations for several target objects (such as scientific and technological projects, products, enterprises, equipment, and process solutions) based on a predefined multi-level indicator system. A typical approach is to establish a hierarchical indicator system. For example, multiple dimensions can be set as primary evaluation dimensions, such as equipment failure characteristics, maintainability, operational stability, and environmental adaptability. Each dimension can be further subdivided into several secondary factors and tertiary quantifiable indicators, forming a complete indicator tree structure. This evaluation method is widely used in scenarios such as scientific and technological project review, product reliability assessment, enterprise innovation capability assessment, and equipment health status assessment.

[0003] However, existing comprehensive evaluation systems generally suffer from the following significant shortcomings in practical applications: evaluation weights are fixed and cannot be adaptively adjusted based on the type, attributes, and indicator data characteristics of the target object, resulting in insufficient relevance and accuracy of the evaluation results; the evaluation process is opaque, with users unable to trace the assessment status of each indicator, the basis for determining the weights, and the calculation logic of the comprehensive evaluation results, leading to a lack of verifiability and credibility of the evaluation results; and the evaluation results lack an interactive confirmation mechanism, with users only passively receiving evaluation conclusions and unable to view and verify the various evaluation results and calculation processes, making it difficult for users to confirm the reliability of the evaluation conclusions. Summary of the Invention

[0004] In view of the shortcomings of the existing technology described above, the purpose of this invention is to propose an intelligent interactive system based on multi-dimensional evaluation analysis. This system employs a dynamic weighting mechanism based on a combination of subjective and objective methods to adaptively weight multi-level indicators of the target object, thereby improving the relevance and accuracy of the evaluation results. Simultaneously, the system discloses the evaluation results of each indicator, the comprehensive evaluation results, and the calculation process to the user, and provides an interactive mechanism for user viewing and confirmation, thereby enhancing the transparency and reliability of the evaluation process.

[0005] To achieve the above and other related objectives, this invention provides an intelligent interaction system based on multidimensional evaluation analysis, comprising: a human-computer interaction agent for extracting a target object from the user's dialogue content; an indicator analysis agent for retrieving relevant information of the target object from an accessed database and / or knowledge base, and analyzing multi-level indicators of the target object accordingly; and an evaluation calculation agent for dynamically assigning weights to the indicators of the target object using a combination of subjective and objective methods, and calculating the individual evaluation results of each indicator of the target object, as well as the comprehensive evaluation result of all indicators, based on the weighted weights; wherein, the human-computer interaction agent is also used to feed back the individual evaluation results and comprehensive evaluation results of the target object, as well as the calculation process of the evaluation calculation agent, to the user for viewing and confirmation.

[0006] According to a specific embodiment of the present invention, the evaluation and computational agent dynamically assigns weights to the higher-level indicators of the target object through a subjective method; wherein, the subjective method adopts the analytic hierarchy process (AHP).

[0007] According to a specific embodiment of the present invention, the evaluation computing agent dynamically assigns weights to the indicators at the higher level of the target object according to the following steps: constructing a corresponding hierarchical structure model based on the indicators at the higher level of the target object; wherein, each layer of the hierarchical structure model is constructed using indicators of the same level; for each layer of the hierarchical structure model, determining the relative importance based on the types of any two indicators to obtain the importance judgment matrix of that layer of the hierarchical structure model; and calculating the weights of each indicator from the importance judgment matrix using the eigenvalue decomposition method.

[0008] According to a specific embodiment of the present invention, the formula for calculating the index weight is as follows: , ,in, This is the importance judgment matrix. For feature vectors, For the preset feature values, Let be the weight of the i-th indicator.

[0009] According to a specific embodiment of the present invention, the evaluation computing agent dynamically assigns weights to the lower-level indicators of the target object using an objective method; wherein, the objective method employs the entropy weighting method.

[0010] According to a specific embodiment of the present invention, the evaluation computing agent dynamically assigns weights to the lower-level indicators of the target object according to the following steps: performing dimensionless processing on the data of each indicator to obtain standardized indicator data; calculating the information weight of each indicator based on the similar objects of the target object and the standardized indicator data; calculating the information entropy of each indicator based on the information weight, and determining the weight of the corresponding indicator accordingly.

[0011] According to a specific embodiment of the present invention, the data of each indicator is dimensionless to obtain standardized indicator data, including: for positive indicators, the standardized indicator data is calculated according to the following formula: For the reverse indicator, its standardized indicator data is calculated using the following formula: ,in, The original data for the m-th positive indicator. The standardized data for the m-th positive indicator is... It is the minimum value among the m-th positive indicators. It is the maximum value among the m-th positive indicators. The original data for the nth contrarian indicator. The standardized data for the nth positive indicator. It is the minimum value among the nth positive indicators. It is the maximum value among the nth positive indicators.

[0012] According to a specific embodiment of the present invention, the formula for calculating the weight of indicator information is as follows: ,in, The standardized index data for the j-th index of the target object. The information weight of the j-th indicator of the target object. The standardized index data of the j-th index of the i-th class object of the target object. The number of objects of the same type as the target object.

[0013] According to a specific embodiment of the present invention, the formula for calculating the index information entropy is as follows: ,in, Let j be the information entropy of the j-th indicator of the target object. As a preset constant, The information weight of the j-th indicator of the i-th class of objects of the target object. The information weight of the j-th indicator of the target object. The number of objects of the same type as the target object.

[0014] According to a specific embodiment of the present invention, the formula for calculating the index weight is as follows: ,in, The weight of the j-th indicator of the target object, Let j be the information entropy of the j-th indicator of the target object. The number of metrics for the target object.

[0015] According to a specific embodiment of the present invention, the human-computer interaction intelligent agent is further configured to feed back the multi-level indicators of the target object and the corresponding analysis process to the user for viewing and confirmation, and obtain the user's confirmation instruction or modification content, or extract modification opinions from the user's dialogue content, so as to return them to the indicator analysis intelligent agent; wherein, the evaluation calculation intelligent agent is configured to dynamically assign weights to the indicators of the target object using a combination of subjective and objective methods after all the indicators of the target object have been confirmed by the user, and calculate the individual evaluation results of each indicator of the target object and the comprehensive evaluation results of all indicators according to the weighted weights.

[0016] According to a specific embodiment of the present invention, the indicator analysis agent is further configured to reanalyze the corresponding indicators based on the user's modifications or suggestions, so that the human-computer interaction agent can provide feedback on the modified indicators and the corresponding analysis process to the user for review and confirmation.

[0017] According to a specific embodiment of the present invention, for the lower-level indicators of the target object, the evaluation calculation agent is used to calculate the individual evaluation result of the indicator based on the data of the indicator and the corresponding weight. For the higher-level indicators of the target object, the evaluation calculation agent is used to calculate the individual evaluation result of the indicator based on the individual evaluation result of the lower-level indicators of the indicator and the corresponding weight.

[0018] According to a specific embodiment of the present invention, the human-computer interaction intelligent agent is further configured to extract questions about the target object from the user's new round of dialogue content and generate corresponding query instructions.

[0019] According to a specific embodiment of the present invention, it further includes: a joint analysis agent, used to use the individual evaluation results and comprehensive evaluation results of the target object, and the corresponding calculation process as context, and in conjunction with the query instruction initiated by the human-computer interaction agent, to query corresponding data information from the accessed database and / or knowledge base; wherein, the human-computer interaction agent is also used to generate a corresponding response based on the data information queried by the joint analysis agent, and provide feedback to the user for viewing and confirmation.

[0020] The beneficial effects of this invention are as follows: This invention adopts a dynamic weighting mechanism that combines subjective and objective factors, which realizes the adaptive adjustment of indicator weights according to the target object type and indicator data characteristics. This effectively overcomes the defects of fixed weight settings in traditional evaluation and significantly improves the pertinence and scientific nature of the evaluation results.

[0021] Meanwhile, this invention discloses the individual evaluation results, comprehensive evaluation results, and complete calculation process of each indicator to the user, making the entire evaluation process transparent and traceable. This eliminates the "black box" nature of traditional methods and greatly enhances the verifiability and credibility of the evaluation results. Furthermore, users can not only view and confirm the evaluation results and calculation process during the evaluation stage, but also confirm, modify, or propose modifications to the indicators during the indicator system analysis stage. This allows for re-analysis until the user confirms that there are no errors, ensuring the reliability and accuracy of subsequent evaluation calculations.

[0022] Furthermore, this invention uses completed evaluation data and calculation processes as context to support users in conducting in-depth inquiries and data backtracking on the evaluation results, achieving a seamless connection between the evaluation stage and the question-and-answer analysis stage, enabling users to fully interpret and utilize the evaluation results. Attached Figure Description

[0023] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with this application and, together with the description, serve to explain the principles of this application. It is obvious that the drawings described below are merely some embodiments of this application, and those skilled in the art can obtain other drawings based on these drawings without any inventive effort.

[0024] Figure 1 This is a schematic diagram of the structure of an intelligent interactive system based on multidimensional evaluation analysis provided in one embodiment of the present invention. Detailed Implementation

[0025] The embodiments of the present invention will be described below with reference to the accompanying drawings and preferred embodiments. Those skilled in the art can easily understand other advantages and effects of the present invention from the content disclosed in this specification. The present invention can also be implemented or applied through other different specific embodiments, and various details in this specification can also be modified or changed based on different viewpoints and applications without departing from the spirit of the present invention. It should be understood that the preferred embodiments are only for illustrating the present invention and not for limiting the scope of protection of the present invention.

[0026] It should be noted that the illustrations provided in the following embodiments are only schematic representations of the basic concept of the present invention. Therefore, the drawings only show the components related to the present invention and are not drawn according to the actual number, shape and size of the components in the actual implementation. In the actual implementation, the form, quantity and proportion of each component can be arbitrarily changed, and the layout of the components may also be more complex.

[0027] In the following description, numerous details are explored to provide a more thorough explanation of embodiments of the invention. However, it will be apparent to those skilled in the art that embodiments of the invention may be practiced without these specific details. In other embodiments, publicly known structures and devices are shown in block diagram form rather than in detail to avoid obscuring embodiments of the invention.

[0028] Please see Figure 1 The intelligent interactive system shown includes: a human-computer interaction agent 10, an index analysis agent 20, an evaluation calculation agent 30, and a joint analysis agent 40, and the agents can communicate with each other via API interfaces.

[0029] The human-computer interaction agent 10 can be built using a large language model to recognize user intent through semantic understanding and analyze the target object. For example, the human-computer interaction agent 10 can pre-load a pre-trained intent recognition model and a named entity model. When a user creates an evaluation and analysis task and the task begins execution, the human-computer interaction agent 10 receives the natural language input by the user. First, it uses the intent recognition model to determine whether the user's intent is an evaluation and analysis type intent. Then, it uses the named entity model to extract the name of the target object. For example, from "Please help me analyze the reliability performance of product model A", "product model A" can be extracted; from "How is the recent operating status of device model B?", "device model B" can be extracted; from "Analyze the fault performance of power modules in batch C", "power modules in batch C" can be extracted, and so on. There are no major restrictions on this.

[0030] The indicator analysis agent 20 can retrieve relevant information about the target object from the accessed knowledge base and / or database based on the target object extracted by the human-computer interaction agent 10. Specifically, the indicator analysis agent 20 is pre-configured with indicator templates for evaluation and analysis, and the templates can adopt a multi-level hierarchical structure. For example, the template can define first-level indicators, second-level indicators, and third-level indicators. The first-level indicators represent macro-level evaluation dimensions, the second-level indicators are subdivisions of the first-level indicators, and the third-level indicators are specific data items that can be quantified and measured. Second-level and third-level indicators can be combined into one category, and third-level indicators can be further subdivided without restriction. In practical applications, the third-level indicators can be configured with the following attributes: channel source marker, which identifies the data acquisition path of the indicator, including: channel A, such as knowledge base retrieval enhancement generation; channel B, such as external data import; channel C, such as web page retrieval enhancement generation; polarity marker, which identifies whether the indicator is a positive indicator (the larger the value, the better) or a negative indicator (the smaller the value, the better); weighting participation marker, which identifies whether the indicator participates in subsequent dynamic weighting; and pre-configured subjective weight, which can inherit its superior first-level and second-level indicators.

[0031] In one specific embodiment, assuming the target object is "Product Model A", the user expects to evaluate the reliability level of this product within a preset operating cycle. Accordingly, the indicator analysis agent 20 can load a corresponding multi-level indicator system from a preset indicator template based on the product type, application conditions, and reliability evaluation type of Product Model A. For this purpose, the comprehensive reliability evaluation result of Product Model A can be set as the target layer, and under the target layer, first-level indicators such as fault occurrence characteristics, maintainability, operational stability, and environmental adaptability can be set.

[0032] For example, regarding fault occurrence characteristics, secondary indicators such as fault frequency and fault severity can be set. Fault frequency can be characterized by tertiary indicators such as the number of faults per thousand hours or the fault rate, and fault severity can be characterized by tertiary indicators such as the proportion of critical faults. Regarding maintainability, secondary indicators such as repair efficiency and maintenance convenience can be set, and these are characterized by tertiary indicators such as average repair time, first-time repair success rate, and maintenance difficulty coefficient. Regarding operational stability, secondary indicators such as continuous operation capability and downtime impact can be set, and these are characterized by tertiary indicators such as mean time between failures (MTBF) and the proportion of unplanned downtime. Regarding environmental adaptability, secondary indicators such as environmental stress tolerance can be set, and these are characterized by tertiary indicators such as the pass rate of high-temperature cycling tests and the number of failures in vibration tests.

[0033] It is understood that the indicator analysis agent 20 can also determine each indicator and its corresponding hierarchical relationship according to the template provided by the user, and there is no limitation on this. Modifications and refinements made by those skilled in the art to the embodiments of the present invention without departing from the spirit of the present invention still fall within the protection scope of this application.

[0034] It should also be noted that the subordinate indicators of a certain higher-level indicator may involve specific parameter calculations. For example, the secondary indicators under the higher-level indicator "fault occurrence characteristics" include fault frequency and fault severity. Fault frequency can be further characterized by the tertiary indicator of fault count per thousand hours or fault rate. The corresponding fault count per thousand hours or fault rate requires the indicator analysis agent 20 to call the product's operation logs, fault work orders, maintenance records, test reports and other data during the evaluation period for analysis and calculation, thereby quantifying the indicator data.

[0035] Based on the above processing operations, the indicator analysis agent 20 can analyze and obtain multi-level indicators for the target object. Further, after determining the multi-level indicators for the target object, the indicator analysis agent 20 will transmit the complete indicator system and its analysis basis to the human-computer interaction agent 10. The human-computer interaction agent 10 will then sequentially feed back each indicator and its corresponding analysis process to the user in the form of structured cards, according to the hierarchical relationship of the indicators, for the user to view and confirm.

[0036] In response, users can directly edit the data (such as quantified parameters) or analysis logic of any indicator on the interactive interface, or users can express their intention to modify through natural language input, and the human-computer interaction intelligent agent 10 can extract the corresponding modification opinions through semantic parsing.

[0037] The human-computer interaction agent 10 will return the obtained modifications or suggestions to the indicator analysis agent 20. The indicator analysis agent 20 will then make adaptive adjustments based on the returned modification information and re-analyze the corresponding indicators. The modified indicators and analysis process will be fed back to the user for confirmation until the user confirms that all indicators are accurate.

[0038] To optimize the analysis path and improve user review efficiency, the indicator analysis agent 20 supports at least the following two analysis strategies: One approach involves first analyzing the higher-level metrics of the target object, and then having the human-computer interaction agent 10 provide all higher-level metrics and the analysis process to the user at once. Once the user confirms the accuracy of any higher-level metric, the agent then analyzes the specific lower-level metrics for that confirmed higher-level metric and sends the lower-level metrics to the user for confirmation in sequence according to their relationships.

[0039] Another approach is to start from a specific dimension, simultaneously analyzing the higher-level indicators and all subordinate indicators of that dimension. The human-computer interaction agent 10 then provides the user with all indicators for that dimension at once, allowing the user to clearly understand the complete analysis path and logic for the current dimension. Correspondingly, the indicator analysis agent 20 can also readjust the analysis logic for the current dimension based on the user's feedback.

[0040] It is evident that, through the aforementioned interactive confirmation mechanism, the indicator analysis agent 20 ensures that the final confirmed multi-level indicators have high accuracy and reliability, thereby providing a solid data foundation for the dynamic weighting of the subsequent evaluation calculation agent 30.

[0041] In one specific embodiment, assuming that the human-computer interaction agent 10 displays to the user the hierarchical indicators (such as "fault occurrence characteristics - fault frequency - number of faults per thousand hours") and their analysis basis for a certain dimension of product A, it will also display the initial configuration of the evaluation of the indicators in that dimension, such as "the ratio between the number of faults and the cumulative operating hours within the statistical period". At this time, the user proposes a modification: "Product A has a period of debugging operation data within the evaluation period. Please remove the operating hours and corresponding fault records during the debugging phase, and recalculate the number of faults per thousand hours according to the comparable caliber of normal service conditions."

[0042] In response, the human-computer interaction agent 10 can extract the specific indicator to be modified, the object of modification, and the modification rules from the modification suggestions, forming structured modification information such as "indicator identified as number of failures per thousand hours," "object of modification as statistical range of operating hours and statistical range of failure records," and "modification rules as removing data from the debugging phase and adopting comparable standards under normal service conditions," and return this information to the indicator analysis agent 20. Accordingly, the indicator analysis agent 20 can adjust the configuration of the indicator evaluation in this dimension (such as data filtering conditions, operating condition filtering rules, and calculation rules), and re-analyze the operating logs, failure work orders, and maintenance records of product model A within the corresponding evaluation period. The human-computer interaction agent 10 then displays the re-analyzed upper and lower level indicators, along with their modified indicator definitions, data ranges, calculation logic, and analysis basis, to the user.

[0043] The evaluation calculation agent 30 is used to dynamically assign weights to the indicators of the target object through a combination of subjective and objective methods after the multi-level indicators of the target object have been confirmed by the user. Based on the weighted weights, it calculates the individual evaluation results of each indicator of the target object and the comprehensive evaluation results of all indicators.

[0044] Specifically, in constructing an evaluation system, the choice of weighting method directly affects the scientific validity and accuracy of the evaluation results. To ensure the objectivity and scientific rigor of the evaluation system, this embodiment uses a combined subjective and objective weighting method to dynamically assign weights to the indicators. Specifically, subjective weighting is applied to indicators at the higher level of the target object (such as primary and secondary indicators), employing the Analytic Hierarchy Process (AHP) as the weighting method.

[0045] The Analytic Hierarchy Process (AHP) is a method for solving complex, multi-factor evaluation problems by structuring and hierarchizing the problem and combining qualitative and quantitative approaches. In AHP, evaluators or evaluation systems decompose complex evaluation problems into multiple levels, including the objective level, criteria level, and alternative level. By comparing pairs of factors at each level, the relative importance of different factors is assessed, and the weight of each factor is ultimately calculated.

[0046] To address this, the evaluation computational agent 30 can first construct a hierarchical model corresponding to the target object. Understandably, the target layer represents the evaluation objectives of the target object, such as the comprehensive reliability evaluation result of a product; the criterion layer represents the primary indicators affecting the reliability evaluation result, such as fault occurrence characteristics, maintainability, operational stability, and environmental adaptability; and the sub-criterion layer represents the secondary indicators refined from the primary indicators. Through the construction of this hierarchical model, each evaluation dimension and its corresponding specific indicators can be clearly identified.

[0047] Based on the current hierarchical model, the evaluation computational agent 30 utilizes past expert experience to perform pairwise comparisons of factors at each layer to determine their relative importance. Specifically, for each layer of the hierarchical model, relative importance is determined based on the type of any two indicators. For example, a 1-5 scale method can be used for pairwise comparisons, with the scale used to characterize the relative importance between indicators. This ultimately yields the importance judgment matrix for that layer. .

[0048] Furthermore, the evaluation of computational agents 30 can be based on the importance judgment matrix. The weights of each indicator are calculated using the following formula: , , in, For feature vectors, For the preset feature values, Let be the weight of the i-th indicator.

[0049] In one specific embodiment, taking the reliability evaluation of product model A as an example, a hierarchical structure model can be used to represent the hierarchical relationship between product reliability evaluation objectives and indicators at various levels. Specifically, for any higher-level indicator in the hierarchical structure model... It will be directly subordinate to this higher-level indicator. The three subordinate indicators at the same level are denoted as follows: , and Furthermore, the three subordinate indicators at the same level are grouped together for comparison. In response, the evaluation computation agent 30 can use the indicators... , and The relative importance between each pair of elements is used to construct the following importance judgment matrix:

[0050] Among them, the importance judgment matrix The first row and first column correspond to the indicators The second row and second column correspond to the indicators The third row and third column correspond to the indicators The hierarchical structure model is used to determine the indicators participating in this comparison, the level to which the indicators belong, and the common superior indicators of the indicators, while the importance judgment matrix is ​​used to record the relative importance of each indicator within the same comparison group.

[0051] The eigenvector is related to the importance judgment matrix. a certain characteristic value The corresponding non-zero ordered vector can be represented as: , And the eigenvectors and their corresponding eigenvalues ​​satisfy: , That is, the importance judgment matrix Acting on the feature vector Afterwards, the direction of the eigenvector remains unchanged, and it is only according to the eigenvalues. The represented multiplier is scaled. Components in the eigenvector. , and According to the row and column order of the importance judgment matrix, respectively, they are compared with the indicators. , and One-to-one correspondence is used to characterize the three indicators relative to their common superior indicator. The relative importance of.

[0052] Therefore, the feature vector is not directly generated by the hierarchical structure model itself, but is calculated by the importance judgment matrix corresponding to the comparison group after determining the comparison group of the same level indicators according to the hierarchical structure model.

[0053] Specifically, the evaluation computational agent uses 30 pairs of importance judgment matrices. Perform eigenvalue decomposition to find the eigenvalues ​​that satisfy the following equation: ,in, It is an identity matrix. The evaluation computation agent 30 can select the largest eigenvalue from multiple calculated eigenvalues. And by solving: , to obtain the largest eigenvalue The corresponding feature vector.

[0054] Based on the above, it is possible The corresponding feature vector can be represented as: Since the same eigenvector multiplied by any non-zero constant still represents the same relative proportion, the evaluation calculation agent 30 can further normalize the eigenvector to obtain: , Therefore, the indicators , and Compared to common superior indicators The weights are 4 / 7, 2 / 7, and 1 / 7, respectively.

[0055] For indicators at lower levels of the target object (such as third-level indicators), objective weighting is adopted, specifically using the entropy weighting method.

[0056] The entropy weight method determines the weight of each indicator by measuring its information entropy (uncertainty). It can objectively reflect the importance of each indicator in the comprehensive evaluation. Its basic idea is that the greater the degree of variation of an indicator, the greater its contribution to the differentiation of the comprehensive evaluation results, and therefore it should be assigned a higher weight.

[0057] To address this, the evaluation computational agent 30 first performs dimensionless processing on the data for each indicator, for example, using a 0-1 standardization method. Here, it can be understood that indicators include positive and negative indicators, evaluating the characteristics of the target object from the positive and negative perspectives, respectively. For example, when evaluating the reliability of product model A, positive indicators may include mean time between failures (MTBF), first-time repair success rate, and environmental test pass rate. The larger the indicator data (quantified parameter value), the better the performance of product model A in the corresponding aspect. Negative indicators may include failure rate, mean repair time, maintenance difficulty coefficient, and percentage of unplanned downtime. The smaller the indicator data, the better the performance of product model A in the corresponding aspect.

[0058] Therefore, for positive indicators, 0-1 standardization can be performed using the following formula: , For contrarian indicators, 0-1 standardization can be performed using the following formula: , in, The original data for the m-th positive indicator. The standardized data for the m-th positive indicator is... It is the minimum value among the m-th positive indicators. It is the maximum value among the m-th positive indicators. The original data for the nth contrarian indicator. The standardized data for the nth positive indicator. It is the minimum value among the nth positive indicators. This represents the maximum value among the nth positive indicators. It should be noted that the minimum and maximum values ​​mentioned above refer to the minimum and maximum values ​​of the original data for the corresponding indicators of the target object and its counterparts within the same evaluation period and under the same statistical caliber, within the same evaluation task.

[0059] For example, in a product reliability evaluation task, the mean time between failures (MTBF) of product models A, B, C, and D are 600 hours, 450 hours, 700 hours, and 550 hours, respectively. The minimum value of this positive indicator is 450 hours, and the maximum value is 700 hours. The standardized indicator data for product model A is (600-450) / (700-450) = 0.60.

[0060] For example, if the average repair times for product type A, product type B, product type C, and product type D are 4 hours, 2 hours, 7 hours, and 5 hours respectively, then the minimum value of this reverse indicator is 2 hours and the maximum value is 7 hours. The standardized indicator data corresponding to product type A is (7-4) / (7-2) = 0.60.

[0061] Therefore, indicators with different units of measurement and different evaluation directions can be converted into a unified numerical range.

[0062] Furthermore, after standardizing each indicator, the corresponding information weight can be calculated based on the standardized data. , which is the proportion of the evaluation result of the j-th indicator of the target object among the same indicators of similar objects, and the specific calculation formula is as follows: , in, The information weight of the j-th indicator of the target object. For the j-th indicator of the target object, the standardized indicator data is... The standardized indicator data for the j-th indicator of the i-th class of objects in the target object. The number of objects of the same type as the target object.

[0063] Then, based on the information weight of each indicator, the information entropy of each indicator can be calculated. The specific calculation formula is as follows: , in, Let the information entropy of the j-th indicator of the target object be . This is a preset constant, generally taken as... , The information weight of the j-th indicator of the i-th class of objects in the target object.

[0064] Finally, after calculating the information entropy of the indicator, its corresponding weight can be determined. as follows: , in, Let j be the weight of the j-th indicator of the target object. The number of metrics for the target object.

[0065] Based on the above, the evaluation and computational agent 30 can calculate the corresponding individual evaluation results according to the weights assigned to each indicator of the target object. Specifically, the calculation logic of the indicator evaluation results follows the principle of "bottom-up, hierarchical aggregation".

[0066] First, determine the individual evaluation results for the lowest-level indicators. For example, calculate the individual evaluation results for all third-level indicators. Since third-level indicators can be specific quantifiable data items, the corresponding individual evaluation results can be calculated based on the indicator data and indicator weights. For example, if the evaluation result is... .

[0067] Secondly, the evaluation results of higher-level indicators can be determined based on the evaluation results of all their subordinate indicators. For example, the evaluation result of a secondary indicator is the weighted sum of the individual evaluation results of all its tertiary indicators, where the weight of the weighted sum is the weight of each tertiary indicator after being assigned weights.

[0068] Similarly, the individual evaluation results of a primary indicator can be calculated based on the individual evaluation results of all its secondary indicators, thus forming a complete tree-like calculation structure.

[0069] It is also understandable that the overall evaluation result of all indicators is the weighted sum of the individual evaluation results of all indicators.

[0070] Finally, the human-computer interaction intelligent agent 10 feeds back the individual evaluation results and comprehensive evaluation results of the target object, as well as the specific calculation process, to the user for review and confirmation.

[0071] Understandably, the above calculation results are persistently stored in the form of structured data and written into a detailed data table of a preset template, supporting subsequent multi-dimensional export and visualization.

[0072] Furthermore, after viewing the evaluation results of the target object (including individual evaluation results and comprehensive evaluation results of the indicators), users may have further query needs. Accordingly, the human-computer interaction intelligent agent 10 can extract questions about the target object from the user's new round of dialogue and generate corresponding query instructions.

[0073] The joint analysis agent 40 can construct structured contextual data, which includes multi-level indicators of the target object, corresponding individual evaluation results, comprehensive evaluation results, and the analysis process of the indicators and the calculation process of the evaluation results. Then, it can query the corresponding data information from the accessed database and / or knowledge base by combining the query command initiated by the human-computer interaction agent 10, so that the human-computer interaction agent 10 can generate a corresponding response based on the queried data information and provide feedback to the user for viewing and confirmation.

[0074] The database and knowledge base can be external open-source data information, such as web pages, public intelligence databases, etc., or a corresponding knowledge base can be built for the system to support the needs of batch import and storage of internal data.

[0075] Specifically, the knowledge base construction process includes the following steps: Document parsing and text extraction: The document parsing engine converts raw documents of different formats into plain text content. For image documents such as scanned PDFs, visual AI services can be used for intelligent parsing.

[0076] Text slicing: Dividing text into semantically complete blocks according to a preset slicing strategy. The slicing strategy includes fixed-length slices and semantic boundary slices, and maintains appropriate overlap between slices to ensure contextual coherence.

[0077] Vectorization encoding: A vector encoding model (such as the Embedding model) is called to generate a corresponding high-dimensional semantic vector representation for each text slice, and the vector index is written to the vector database.

[0078] Metadata binding: The vector index of each text slice is associated with and stored with the following metadata: analysis object identifier (identifying the target object to which the document belongs), data source identifier (identifying the document type and source), and time period parameter (identifying the year or period corresponding to the document). Metadata binding enables precise filtering by target object and time dimension during subsequent retrieval, significantly reducing the recall rate of irrelevant slices.

[0079] Object storage: The original document file is stored in the object storage service, and the system retains the original document text to support subsequent accurate backtracking.

[0080] It is evident that once all the information data of the objects required for the analysis has been imported, the knowledge base stores all the original materials required for the evaluation analysis, providing a data foundation for the subsequent information retrieval by the indicator analysis agent 20, the evaluation calculation agent 30, and the joint analysis agent 40.

[0081] In one specific embodiment, after the product reliability evaluation is completed, the user further inputs "Why is the standardized mean time between failures (MTBF) result for product model A 0.60?". Accordingly, the human-computer interaction agent 10 can recognize this question as a request to trace the calculation process of the existing evaluation results and generate a query instruction that includes at least the target object identifier, indicator identifier, evaluation period, and query type.

[0082] In response, the Joint Analytical Agent 40 will use the indicator data, similar object data, indicator directions, standardized formulas, and user correction records from this evaluation task as context, and retrieve the following data information from the database and / or knowledge base according to the target object identifier, indicator identifier, and evaluation period: The original data for the mean time between failures (MTBF) of product model A is 600 hours, and the corresponding data for product models B, C, and D are 450 hours, 700 hours, and 550 hours, respectively.

[0083] Since this indicator is a positive indicator, the positive indicator standardization formula should be used. The human-computer interaction agent 10 can then generate the response: "0.60 is the standardized indicator data, not the actual mean time between failures (MTBF) of product model A. In this evaluation sample, the minimum value of this indicator is 450 hours, and the maximum value is 700 hours. Substituting the actual value of 600 hours for product model A into the positive indicator standardization formula yields a standardized result of 0.60." Simultaneously, the data source, scope of similar objects, statistical caliber, user correction records, and specific calculation steps are provided to the user for review and confirmation.

[0084] For example, a user might input, "Based on the results of this product reliability evaluation, please generate a reliability diagnostic report for product model A, focusing on analyzing the advantages, disadvantages, major potential failures, and maintenance improvement suggestions of product model A compared to products model B, C, and D." Correspondingly, the human-computer interaction agent 10 can recognize this as a comprehensive analysis request for the overall evaluation results and generate a query command that includes at least the target object identifier, document type, scope of comparison objects, analysis content, and output requirements.

[0085] In response, the joint analytical agent 40 uses the evaluation results of each level of indicators of product A, the comprehensive evaluation results, the indicator weights, the calculation process, and the comparison results of product A with product B, product C, and product D as context to query relevant data information from the database and / or knowledge base.

[0086] The human-computer interaction intelligent agent 10 generates a corresponding reliability diagnostic report based on the data information queried by the joint analysis intelligent agent 40. The reliability diagnostic report includes a summary of evaluation conclusions, main advantages, main disadvantages, comparison with similar products, key potential failures, items to be further verified, and maintenance and improvement suggestions. It also links each analysis conclusion with the corresponding indicator evaluation results, raw data, and / or calculation basis for users to view, confirm, and modify.

[0087] It should be noted that the human-computer interaction intelligent agent 10 can be integrated into the indicator analysis intelligent agent 20, the evaluation calculation intelligent agent 30, and the joint analysis intelligent agent 40, thereby skipping the interaction process between intelligent agents. No restrictions are placed on this. Modifications and refinements made by those skilled in the art to the embodiments of the present invention without departing from the spirit of the present invention still fall within the protection scope of this application.

[0088] In summary, by constructing a system architecture in which a human-computer interaction intelligent agent, an indicator analysis intelligent agent, an evaluation calculation intelligent agent, and a joint analysis intelligent agent work collaboratively, this invention achieves the following technical effects: First, a dynamic weighting mechanism combining subjective and objective approaches is adopted. For macro-level indicators, the analytic hierarchy process is used to incorporate expert experience, while for micro-level indicators, the entropy weighting method is used to objectively assign weights based on the characteristics of the data itself. This enables the indicator weights to be adaptively adjusted according to the type of target object and the characteristics of the indicator data, effectively overcoming the defects of fixed weight settings in traditional evaluations and significantly improving the pertinence and scientific nature of the evaluation results. Secondly, the individual evaluation results, comprehensive evaluation results, and complete calculation process (including the basis for weight determination and evaluation logic) of each indicator are all made public to users, making the entire evaluation process transparent and traceable. This eliminates the "black box" status of traditional methods and greatly enhances the verifiability and credibility of the evaluation results. Furthermore, users can not only view and confirm the evaluation results and calculation process during the evaluation phase, but also confirm, modify, or propose modifications to the indicators during the indicator system analysis phase. This allows for re-analysis until the user confirms that everything is correct, ensuring the reliability and accuracy of subsequent evaluation calculations.

[0089] Furthermore, using the completed evaluation data and calculation process as context, it supports users to conduct in-depth inquiries and data backtracking on the evaluation results, achieving a seamless connection between the evaluation stage and the question-and-answer analysis stage, enabling users to fully interpret and utilize the evaluation results.

[0090] The above embodiments are merely illustrative of the principles and effects of the present invention and are not intended to limit the invention. Any person skilled in the art can modify or alter the above embodiments without departing from the spirit and scope of the present invention. Therefore, all equivalent modifications or alterations made by those skilled in the art without departing from the spirit and technical concept disclosed in the present invention should still be covered by the claims of the present invention.

[0091] Throughout this specification, the terms "an embodiment," "embodiment," or "specific embodiment" refer to a particular feature, structure, or characteristic described in connection with an embodiment that is included in at least one embodiment of the invention, but not necessarily in all embodiments. Therefore, the various representations of the phrases "in one embodiment," "in an embodiment," or "in a specific embodiment" in different places throughout the specification do not necessarily refer to the same embodiment. Furthermore, a particular feature, structure, or characteristic of any specific embodiment of the invention can be combined with one or more other embodiments in any suitable manner. It should be understood that other variations and modifications of the embodiments of the invention described and illustrated herein may be based on the teachings herein and will be considered part of the spirit and scope of the invention.

[0092] Furthermore, unless otherwise expressly stated, any arrows in the accompanying drawings should be considered illustrative only and not limiting. Additionally, unless otherwise stated, the term "or" as used herein is generally intended to mean "and / or". Where a term is anticipated to provide a separation or combination capability that is unclear, a combination of components or steps will also be considered as indicated.

[0093] As used herein and throughout the claims below, unless otherwise specified, “a” and “the” include the plural references. Similarly, as used herein and throughout the claims below, unless otherwise specified, “in” means “in” and “on”.

[0094] The above description of the embodiments shown in this invention (including the content set forth in the abstract of the specification) is not intended to be an exhaustive enumeration or to limit the invention to the precise forms disclosed herein. Although specific embodiments and examples of the invention have been described herein for illustrative purposes only, various equivalent modifications are possible within the spirit and scope of the invention, as will be recognized and understood by those skilled in the art. As indicated, these modifications can be made to the invention in accordance with the above description of the embodiments described herein, and such modifications will be within the spirit and scope of the invention.

[0095] This document has generally described the systems and methods in detail to aid in understanding the invention. Furthermore, various specific details have been set forth to provide a general understanding of embodiments of the invention. However, those skilled in the art will recognize that embodiments of the invention can be practiced without one or more specific details, or using other means, systems, accessories, methods, components, materials, parts, etc. In other instances, publicly known structures, materials, and / or operations have not been specifically shown or described in detail to avoid obscuring aspects of embodiments of the invention.

[0096] Therefore, although the invention has been described herein with reference to specific embodiments thereof, freedom of modification, various changes and substitutions are also within the scope of the foregoing disclosure, and it should be understood that in some cases, certain features of the invention may be adopted without departing from the scope and spirit of the invention and without corresponding use of other features. Thus, many modifications can be made to adapt a particular environment or material to the essential scope and spirit of the invention. The invention is not intended to be limited to the specific terminology used in the following claims and / or the specific embodiments disclosed as the best mode for carrying out the invention, but the invention will include any and all embodiments and equivalents falling within the scope of the appended claims. Therefore, the scope of the invention will be defined only by the appended claims.

Claims

1. An intelligent interactive system based on multidimensional evaluation analysis, characterized in that, include: Human-computer interaction intelligent agents are used to extract target objects from the user's dialogue content; The indicator analysis agent is used to retrieve relevant information of the target object from the accessed database and / or knowledge base, and analyze the multi-level indicators of the target object based on it; The evaluation computational agent is used to dynamically assign weights to the indicators of the target object using a combination of subjective and objective methods, and to calculate the individual evaluation results of each indicator of the target object and the comprehensive evaluation results of all indicators based on the weighted weights. The human-computer interaction agent is also used to feed back the individual evaluation results and comprehensive evaluation results of the target object, as well as the calculation process of the evaluation calculation agent, to the user for viewing and confirmation.

2. The intelligent interactive system based on multidimensional evaluation analysis according to claim 1, characterized in that, The evaluation and computational intelligent agent dynamically assigns weights to the higher-level indicators of the target object through subjective methods. The subjective method mentioned above employs the analytic hierarchy process (AHP).

3. The intelligent interactive system based on multidimensional evaluation analysis according to claim 2, characterized in that, The evaluation and computational agent dynamically assigns weights to the higher-level indicators of the target object according to the following steps: A hierarchical structure model is constructed based on the indicators at the higher level of the target object; wherein each layer of the hierarchical structure model is constructed using indicators at the same level. For each layer of the hierarchical model, the relative importance is determined based on the types of any two indicators to obtain the importance judgment matrix for that layer of the hierarchical model. The weights of each indicator are calculated from the importance judgment matrix using the eigenvalue decomposition method.

4. The intelligent interactive system based on multidimensional evaluation analysis according to claim 3, characterized in that, The formula for calculating the indicator weight is as follows: ， ， in, This is the importance judgment matrix. For feature vectors, For the preset feature values, Let be the weight of the i-th indicator.

5. The intelligent interactive system based on multidimensional evaluation analysis according to claim 1, characterized in that, The evaluation and computational intelligent agent dynamically assigns weights to the lower-level indicators of the target object using objective methods; The objective method mentioned above employs the entropy weight method.

6. The intelligent interactive system based on multidimensional evaluation analysis according to claim 5, characterized in that, The evaluation and computational agent dynamically assigns weights to the lower-level indicators of the target object according to the following steps; The data for each indicator is dimensionless to obtain standardized indicator data; Based on similar objects of the target object, the information weight of each indicator is calculated according to the standardized indicator data; The information entropy of each indicator is calculated based on the information weight, and the weight of the corresponding indicator is determined accordingly.

7. The intelligent interactive system based on multidimensional evaluation analysis according to claim 6, characterized in that, The data for each indicator is dimensionless to obtain standardized indicator data, including: For positive indicators, their standardized data are calculated using the following formula: ， For the contrarian indicator, its standardized indicator data is calculated using the following formula: ， in, The original data for the m-th positive indicator. The standardized data for the m-th positive indicator is... It is the minimum value among the m-th positive indicators. It is the maximum value among the m-th positive indicators. The original data for the nth contrarian indicator. The standardized data for the nth positive indicator. It is the minimum value among the nth positive indicators. It is the maximum value among the nth positive indicators.

8. The intelligent interactive system based on multidimensional evaluation analysis according to claim 6, characterized in that, The formula for calculating the weight of indicator information is as follows: ， in, The standardized index data for the j-th index of the target object. The information weight of the j-th indicator of the target object. The standardized index data for the j-th index of the i-th class object of the target object. The number of objects of the same type as the target object.

9. The intelligent interactive system based on multidimensional evaluation analysis according to claim 6, characterized in that, The formula for calculating the information entropy of an indicator is as follows: ， in, Let j be the information entropy of the j-th indicator of the target object. As a preset constant, The information weight of the j-th indicator of the i-th class of objects of the target object. The information weight of the j-th indicator of the target object. The number of objects of the same type as the target object.

10. The intelligent interactive system based on multidimensional evaluation analysis according to claim 6, characterized in that, The formula for calculating the indicator weight is as follows: ， in, The weight of the j-th indicator of the target object, Let j be the information entropy of the j-th indicator of the target object. The number of metrics for the target object.

11. The intelligent interactive system based on multidimensional evaluation analysis according to claim 1, characterized in that, The human-computer interaction intelligent agent is also used to feed back the multi-level indicators of the target object and the corresponding analysis process to the user for viewing and confirmation, and to obtain the user's confirmation instructions or modification content, or to extract modification opinions from the user's dialogue content and return them to the indicator analysis intelligent agent. The evaluation calculation agent is used to dynamically assign weights to the indicators of the target object using a combination of subjective and objective methods after all indicators of the target object have been confirmed by the user, and to calculate the individual evaluation results of each indicator of the target object and the comprehensive evaluation results of all indicators based on the weighted weights.

12. The intelligent interactive system based on multidimensional evaluation analysis according to claim 11, characterized in that, The indicator analysis agent is also used to reanalyze the corresponding indicators based on the user's modifications or suggestions, so that the human-computer interaction agent can provide the user with feedback on the modified indicators and the corresponding analysis process for confirmation.

13. The intelligent interactive system based on multidimensional evaluation analysis according to claim 1, characterized in that, For the indicators at the lower level of the target object, the evaluation calculation agent is used to calculate the individual evaluation result of the indicator based on the indicator data and corresponding weights; For the higher-level indicators of the target object, the evaluation calculation agent is used to calculate the individual evaluation result of the indicator based on the individual evaluation results and corresponding weights of the lower-level indicators of the indicator.

14. The intelligent interactive system based on multidimensional evaluation analysis according to claim 1, characterized in that, The human-computer interaction intelligent agent is also used to extract questions about the target object from the user's new round of dialogue and generate corresponding query instructions.

15. The intelligent interactive system based on multidimensional evaluation analysis according to claim 14, characterized in that, Also includes: The joint analysis agent is used to use the individual evaluation results and comprehensive evaluation results of the target object, as well as the corresponding calculation process, as context, and combine them with the query instructions initiated by the human-computer interaction agent to query the corresponding data information from the accessed database and / or knowledge base; The human-computer interaction agent is also used to generate corresponding responses based on the data information queried by the joint analysis agent, and to provide feedback to the user for viewing and confirmation.

Images