Data service and processing optimization system based on cloud platform

Abstract

The invention relates to the technical field of data optimization processing, in particular to a data service and processing optimization system based on a cloud platform, and the system comprises a data quality evaluation module which carries out the real-time quality evaluation of an input data flow, generates a quality evaluation index comprising an integrity score, noise density and semantic consistency, and carries out the processing of the quality evaluation index; integrating the quality evaluation indexes to generate a window-level comprehensive credibility score; the processing strategy generation module is used for dividing the data flow into high-credibility core data and low-credibility edge data according to the comprehensive credibility score; and the cross-channel resource collaborative scheduling module is used for deploying parallel computing resources of a fixed time window in a deterministic computing channel and starting dynamic telescopic fuzzy computing unit allocation in a probabilistic processing channel. According to the method, low-quality data is effectively prevented from interfering with a high-priority computing task, meanwhile, it is ensured that core data is accurately processed, and the data credibility and computing stability of a cloud computing platform are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of data optimization processing, and particularly to a data service and processing optimization system based on a cloud platform. Background Art

[0002] With the rapid development of cloud computing and big data technologies, cloud platforms have become important infrastructures for data storage, processing, and analysis. In various application scenarios, cloud platforms need to efficiently manage and compute-schedule a large amount of heterogeneous data to meet business requirements. However, existing data processing systems still have technical bottlenecks when facing complex data environments.

[0003] In modern cloud computing environments, data sources are complex, including sensor data, log data, business transaction data, etc. These data often have problems such as missing values, inconsistencies, and noise interference. If effective data quality assessment is not performed before computing, low-quality data may enter the computing process, resulting in deviations in computing results and affecting the accuracy of decision-making. Traditional data quality management methods mainly rely on static rule definitions, are difficult to adapt to large-scale and dynamically changing data stream environments, and cannot achieve accurate measurement and adaptive optimization of data quality.

[0004] Most existing data computing modes of cloud platforms adopt a fixed resource allocation method, that is, computing resources are allocated to different tasks according to a preset strategy. However, in actual applications, due to the volatility of data quality, the fixed resource allocation method often leads to two extreme situations: one is the mixed processing of high-quality data and low-quality data, which causes computing resources to be consumed by low-quality data, thereby reducing the overall computing efficiency; the other is that computing resources fail to dynamically match data requirements, resulting in resource waste or computing bottlenecks. Traditional cloud computing architectures lack the ability to perceive data quality and cannot adaptively adjust computing strategies according to data quality, thus affecting the utilization rate and processing performance of computing resources. Summary of the Invention

[0005] The present invention provides a data service and processing optimization system based on a cloud platform.

[0006] The data service and processing optimization system based on a cloud platform includes:

[0007] A data quality assessment module: performing real-time quality assessment on the input data stream, generating quality assessment indicators including integrity score, noise density, and semantic consistency, and comprehensively fusing the quality assessment indicators to generate a window-level comprehensive credibility score;

[0008] A processing strategy generation module: dividing the data stream into high-trust core data and low-trust edge data according to the comprehensive credibility score, allocating a deterministic computing channel for the core data, and constructing a probabilistic processing channel for the edge data;

[0009] Cross-channel resource collaborative scheduling module: Deploy parallel computing resources with fixed time windows in the deterministic computing channel, enable dynamic scaling of fuzzy computing unit allocation in the probabilistic processing channel, and the two channels share a cache pool and dynamically migrate computing tasks according to changes in data quality.

[0010] Optionally, the data quality assessment module specifically includes:

[0011] Multi-dimensional computing unit: Slice the input data stream by time window, and synchronously execute the following within each time window:

[0012] a. Integrity score calculation: Statistically calculate the missing rate of required fields, and generate a dynamic integrity score in combination with the historical reliability weight of the data source;

[0013] b. Noise density detection: Use the sliding standard deviation algorithm to identify abnormal fluctuations in numerical data, and statistically calculate the noise density of text data through the proportion of irrelevant characters;

[0014] c. Semantic consistency verification: Detect the frequency of semantic conflicts between the data content and the preset business rules through a pre-trained business knowledge graph matching unit;

[0015] Comprehensive scoring unit: Integrate the results of integrity score, noise density, and semantic consistency to obtain a comprehensive credibility score

[0016] Optionally, the business knowledge graph matching unit includes:

[0017] Knowledge graph storage subunit: Store the semantic network related to business rules, including entities, relationships, and rule sets;

[0018] Semantic parsing and vectorization subunit: Convert data records into vector representations that can be compared and matched with the knowledge graph;

[0019] Rule matching engine: Based on graph traversal, detect whether data records violate business rules;

[0020] Conflict frequency calculation subunit: Statistically calculate the number of occurrences of semantic conflicts and output the semantic conflict frequency.

[0021] The specific matching method is as follows:

[0022] First, perform semantic analysis on the input data and convert it into a structured vector representation for use in matching with business rules;

[0023] Extract the rule subgraph related to the current data from the knowledge graph, which includes entities (customers, transactions) and their relationships (transfer limits, credit ratings);

[0024] Calculate the matching degree between the calculated data content and the knowledge graph rules, and use the vector similarity method to determine whether the data conforms to the established business rules;

[0025] Count the number of times the data violates the rules. If the matching degree is lower than the set threshold, it is determined that there is a semantic conflict in the data.

[0026] Optionally, the processing strategy generation module includes a threshold division unit: when the comprehensive credibility score is, it is marked as high-trust core data and assigned to the deterministic calculation channel; when is, it is marked as low-trust edge data and assigned to the probabilistic processing channel.

[0027] Optionally, the comprehensive credibility score is normalized to [0, 1]. When is, one of the following operations is automatically triggered according to the real-time pressure of the data stream:

[0028] a. If the current load rate L of the deterministic channel core < η safe , promote to core data, η safe takes 70%;

[0029] b. Otherwise, demote it to edge data and inject it into the probabilistic channel.

[0030] The high-trust threshold θ high takes the value of 0.6, and the low-trust threshold θ low takes the value of 0.4.

[0031] Optionally, the cross-channel resource collaborative scheduling module specifically includes:

[0032] Fixed-time window deployment unit: According to the comprehensive credibility score of the core data, dynamically adjust the length of the calculation time window of this data. The calculation of the time window length considers the relationship between the comprehensive credibility score and the historical score mean, so that high-credibility data can obtain longer processing time. Adopt a linear mapping model to make the allocation of parallel computing resources proportional to the comprehensive credibility score of each data time window. Based on the total comprehensive credibility score of all core data time windows, calculate its proportion by combining the comprehensive credibility score of a single time window and then multiply by the total available computing resources to obtain the computing resources that can be allocated to this time window;

[0033] Fuzzy calculation unit: For edge data, the number of fuzzy calculation units is determined by the noise density within the current time window. Based on the noise density of text data, the noise coefficient of numerical data, combined with the noise density tolerance coefficient and the maximum number of fuzzy calculation units, calculate the actual number of fuzzy calculation units required;

[0034] Cross-channel task migration unit: including migration from the deterministic computing channel to the probabilistic processing channel and migration from the probabilistic processing channel to the deterministic computing channel.

[0035] Optionally, the computing accuracy of the fuzzy computing unit is adaptively adjusted according to the semantic conflict frequency. When the semantic conflict frequency is high, the computing accuracy decreases accordingly to reduce the computing overhead; when the semantic conflict frequency is low, the computing accuracy is improved.

[0036] Optionally, the conditions for migrating from the deterministic computing channel to the probabilistic processing channel include:

[0037] When the processing time of the core data time window exceeds its allocated time window and its comprehensive credibility score drops below the credibility threshold, the data of this time window is downgraded to the probabilistic processing channel to release the core computing resources. The amount of migrated computing resources is calculated by multiplying the original computing resources of the time window by the ratio of its current comprehensive credibility score to the high credibility threshold.

[0038] The conditions for migrating from the probabilistic processing channel to the deterministic computing channel include:

[0039] For edge data, after fuzzy computing, if the updated comprehensive credibility score exceeds the adjustment range of the high credibility threshold, it is migrated to the deterministic computing channel and computing resources are allocated to it. The allocation method of computing resources is proportional to the comprehensive credibility score, so that high-credibility edge data can obtain more computing resources.

[0040] Optionally, the linear mapping model is expressed as:

[0041] Where K is the total number of current core data time windows, R total is the total available computing resources, is the computing resources allocated to the kth core data window, is the comprehensive credibility score of the ith time window, is the comprehensive credibility score of the jth time window.

[0042] Optionally, the number U of fuzzy computing units fuzzy is bound to the noise density:

[0043] Where is the noise density of text data, is the noise coefficient of numerical data, ρ noise is the noise density tolerance coefficient, U max is the maximum number of fuzzy computing units.

[0044] The beneficial effects of the present invention:

[0045] In the present invention, through a data quality assessment module, real-time quality analysis is performed on the input data stream, and a comprehensive evaluation is carried out in three dimensions of integrity score, noise density, and semantic consistency to generate a window-level comprehensive credibility score. This scoring system can accurately identify data quality differences, and combined with a dynamic dual-threshold division strategy, the data stream is divided into high-credibility core data and low-credibility edge data in real time. Through this mechanism, the system can effectively avoid low-quality data from interfering with high-priority computing tasks, while ensuring that the core data is accurately processed, improving the data credibility and computing stability of the cloud computing platform.

[0046] In the present invention, the cross-channel resource collaborative scheduling module realizes the intelligent collaboration between the deterministic computing channel and the probabilistic processing channel. The core data time window adopts a dynamically adjusted fixed time window processing strategy, automatically expanding or shrinking the time window according to the credibility score to adapt to the computing requirements of data with different credibility levels. At the same time, a credibility-resource linear mapping model is adopted to ensure that the allocation of computing resources matches the data quality, thereby optimizing the computing priority of high-credibility data. For edge data, a noise density-driven dynamic scaling mechanism of the fuzzy computing unit is adopted, so that data streams with higher noise can obtain adaptive computing resources, and the computing accuracy is dynamically adjusted through the feedback of the semantic conflict frequency, thereby reducing the computing cost and improving the utilization rate of computing resources.

[0047] In the present invention, the intelligent migration of cross-channel tasks is realized through a shared cache pool, effectively improving the load balancing of computing resources. When the core computing channel has too high a load due to a decrease in data quality or processing timeout, the system can automatically trigger data degradation, migrating low-quality data from the core channel to the probabilistic processing channel to release high-priority computing resources. At the same time, when the credibility score of edge data reaches the adjustment range of the high-credibility threshold after fuzzy computing, it is automatically upgraded to the core channel, and computing resources are allocated according to the credibility ratio. This two-way migration mechanism ensures that the system can dynamically adjust the attribution of computing tasks according to data quality, improve the accuracy of data computing, and optimize the overall allocation and scheduling ability of computing resources. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] In order to more clearly illustrate the technical solutions in the present invention or the prior art, the following will briefly introduce the drawings required for the description of the embodiments or the prior art. Obviously, the drawings in the following description are only those of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can also be obtained based on these drawings.

[0049] Figure 1 It is a schematic diagram of the system function module of the embodiment of the present invention;

[0050] Figure 2Schematic diagram of the cross-channel resource collaborative scheduling module according to an embodiment of the present invention. Detailed implementation manners

[0051] The present invention will be described in detail below with reference to the accompanying drawings and specific embodiments. At the same time, it should be noted here that in order to make the embodiments more detailed, the following embodiments are the best and preferred embodiments. For some well-known technologies, those skilled in the art can also adopt other alternative ways for implementation; moreover, the accompanying drawings are only for more specifically describing the embodiments, and are not intended to specifically limit the present invention.

[0052] It should be noted that in the specification, references to "an embodiment", "embodiments", "exemplary embodiments", "some embodiments", etc. indicate that the described embodiments may include specific features, structures, or characteristics, but not necessarily every embodiment includes such specific features, structures, or characteristics. Additionally, when combining embodiments to describe specific features, structures, or characteristics, implementing such features, structures, or characteristics in combination with other embodiments (whether explicitly described or not) should be within the knowledge of those skilled in the relevant art.

[0053] Generally, terms can be understood at least in part from their use in context. For example, at least in part depending on the context, the term "one or more" as used herein can be used to describe any feature, structure, or characteristic in a singular sense, or can be used to describe a combination of features, structures, or characteristics in a plural sense. Additionally, the term "based on" can be understood as not necessarily intended to convey a set of exclusive factors, but rather, at least in part depending on the context, can allow for the existence of other factors that may not be explicitly described.

[0054] As Figure 1 - Figure 2 shown, a data service and processing optimization system based on a cloud platform includes:

[0055] A data quality assessment module: performing real-time quality assessment on the input data stream, generating quality assessment indicators including integrity score, noise density, and semantic consistency, and comprehensively fusing the quality assessment indicators to generate a window-level comprehensive credibility score;

[0056] A processing strategy generation module: dividing the data stream into core data with high credibility and edge data with low credibility according to the comprehensive credibility score, allocating a deterministic computing channel for the core data, and constructing a probabilistic processing channel for the edge data;

[0057] A cross-channel resource collaborative scheduling module: deploying parallel computing resources with a fixed time window in the deterministic computing channel, enabling dynamic scaling of fuzzy computing unit allocation in the probabilistic processing channel, sharing a cache pool between the two channels, and dynamically migrating computing tasks according to changes in data quality.

[0058] The data quality assessment module specifically includes:

[0059] 1. Slice the input data stream according to time windows with a window length of T seconds, and synchronously execute operations within each window:

[0060] 1.1 Calculate the integrity score: For N data records in the i-th time window, calculate the dynamic integrity score where represents the total number of missing required fields within the window, and K req is the number of required fields determined by preset business rules. represents the historical reliability weight of the data source, calculated as: H is the number of historical windows, is the number of error records of the data source in the i - t window;

[0061] 1.2 Detect noise density: Calculate the sliding standard deviation noise coefficient for numerical data

[0062] where L is the total number of numerical fields within the current window, and σ s is the sliding standard deviation of the data of the s-th field within the window (the sliding window size is 5), and s is the sliding average of the s-th field;

[0063] Calculate the noise density of irrelevant characters for text data where is the number of characters in the j-th record that are not in the preset character set (excluding ASCII printable characters), is the total number of characters in the j-th record;

[0064] 1.3 Verify semantic consistency: Calculate the frequency of semantic conflicts through knowledge graph matching

[0065] where R is the total number of preset business rules, and KG r is the knowledge graph subgraph of the r-th business rule, is the j-th data record in the i-th time window, indicates that when the data record has a semantic conflict with KG r is 1, otherwise 0.

[0066] 2. Comprehensive scoring unit: Integrate the results of integrity score, noise density, and semantic consistency to obtain a comprehensive credibility score

[0067] where α, β, γ are dynamic weight coefficients, satisfying α + β + γ = 1, α is 0.4, β is 0.3, and γ is 0.3.

[0068] The business knowledge graph matching unit includes:

[0069] A knowledge graph storage subunit: storing a semantic network related to business rules, including entities, relationships, and rule sets;

[0070] A semantic parsing and vectorization subunit: converting data records into vector representations that can be compared and matched with the knowledge graph;

[0071] A rule matching engine: based on graph traversal, detecting whether data records violate business rules;

[0072] A conflict frequency calculation subunit: counting the occurrence times of semantic conflicts and outputting the semantic conflict frequency.

[0073] The specific matching method is as follows:

[0074] First, perform semantic analysis on the input data and convert it into a structured vector representation for matching with business rules;

[0075] Extract a rule subgraph related to the current data from the knowledge graph, which contains entities (customers, transactions) and their relationships (transfer limits, credit ratings);

[0076] Calculate the matching degree between the data content and the knowledge graph rules, and use the vector similarity method to determine whether the data conforms to the established business rules;

[0077] Count the number of times the data violates the rules. If the matching degree is lower than the set threshold, it is determined that there is a semantic conflict in the data.

[0078] The processing strategy generation module includes a threshold division unit: when the comprehensive credibility score is the case, mark it as high-trust core data and allocate it to the deterministic calculation channel; when is the case, mark it as low-trust edge data and allocate it to the probabilistic processing channel.

[0079] The comprehensive credibility score is normalized to [0, 1]. When is the case, automatically trigger one of the following operations according to the real-time pressure of the data stream:

[0080] a. If the current load rate L of the deterministic channel core <η safe , promote to core data, η safe take 70%;

[0081] b. Otherwise, downgrade it to edge data and inject it into the probabilistic channel.

[0082] The high-trust threshold θ highThe value is 0.6, the low credibility threshold θ low The value is 0.4.

[0083] The cross-channel resource collaborative scheduling module specifically includes:

[0084] 1. Fixed time window deployment of the deterministic computing channel:

[0085] The time window length T core And the comprehensive credibility score of the core data Is positively correlated:

[0086] Among them, T base Is the reference time window, μ Q Is the mean value of the historical credibility score, used to balance the dynamic adjustment of the calculation window length;

[0087] The parallel computing resource allocation adopts a credibility-resource linear mapping model:

[0088] Among them, K is the total number of current core data windows, R total Is the total amount of available computing resources, Is the computing resource allocated to the kth core data window;

[0089] 2. Dynamic scaling mechanism of the probabilistic processing channel:

[0090] The number of fuzzy computing units U fuzzy Is dynamically bound to the noise density:

[0091] Among them, Is the noise density of the text data, Is the noise coefficient of the numerical data, ρ noise Is the noise density tolerance coefficient, U max Is the maximum number of fuzzy computing units;

[0092] The fuzzy computing accuracy P fuzzy Is adaptively adjusted according to the semantic conflict frequency:

[0093] Among them, Represents the semantic conflict frequency in the ith window, N is the total number of data records in the window, λ conflict Is the conflict frequency decay factor (preset value).

[0094] 3. Cross-channel task migration of the shared cache pool:

[0095] Migration from the core channel to the edge channel: When the core data window processing times out (t process > T core ) and the credibility Descend to θ high When it reaches -2Δ, demote the current window data to the probabilistic channel, θ high -2Δ is the confidence threshold, releasing computing resources, expressed as:

[0096] where t process is the data processing time of the current window, and R release represents the amount of computing resources released from the core channel;

[0097] Migration from the edge channel to the core channel: When the edge data meets the following after fuzzy calculation:

[0098]

[0099] Immediately migrate it to the core channel and allocate initial resources:

[0100] where Δ is the upgrade buffer threshold (to prevent frequent switching of edge data), R avg is the current average resource occupancy of the core channel, and R migrate is the final computing resources migrated to the core channel.

[0101] The present invention covers any substitutions, modifications, equivalent methods, and solutions made within the spirit and scope of the present invention. For the public to have a thorough understanding of the present invention, specific details are described in detail in the following preferred embodiments of the present invention. However, those skilled in the art can fully understand the present invention without these detailed descriptions. Additionally, well-known methods, processes, procedures, components, and circuits are not described in detail to avoid unnecessary confusion to the essence of the present invention.

[0102] The above are only the preferred embodiments of the present invention. It should be noted that for those of ordinary skill in the art, without departing from the principle of the present invention, several improvements and refinements can be made, and these improvements and refinements should also be regarded as the protection scope of the present invention.

Claims

1. A data service and processing optimization system based on a cloud platform, characterized in that, including: Data Quality Assessment Module: Conduct real-time quality assessment on the input data stream, generate quality assessment metrics including integrity score, noise density, and semantic consistency, and fuse various quality assessment metrics to generate a window-level comprehensive credibility score; Processing Strategy Generation Module: According to the comprehensive credibility score, divide the data stream into core data with high credibility and edge data with low credibility, allocate a deterministic computing channel for the core data, and construct a probabilistic processing channel for the edge data; Cross-channel Resource Cooperative Scheduling Module: Deploy parallel computing resources with a fixed time window in the deterministic computing channel, enable dynamic scaling of fuzzy computing units in the probabilistic processing channel, share a cache pool between the two channels, and dynamically migrate computing tasks according to changes in data quality.

2. The data service and processing optimization system based on the cloud platform according to claim 1, characterized in that The Data Quality Assessment Module specifically includes: Multi-dimensional Computing Unit: Slice the input data stream by time window, and synchronously execute the following within each time window: a. Integrity Score Calculation: Statistically calculate the missing rate of required fields, and generate a dynamic integrity score in combination with the historical reliability weight of the data source; b. Noise Density Detection: Use a sliding standard deviation algorithm to identify abnormal fluctuations in numerical data, and statistically calculate the noise density of text data by the proportion of irrelevant characters; c. Semantic Consistency Verification: Detect the frequency of semantic conflicts between the data content and the preset business rules through a pre-trained business knowledge graph matching unit; Comprehensive scoring unit: Integrate the results of integrity scoring, noise density, and semantic consistency to obtain a comprehensive credibility score 3. The data service and processing optimization system based on a cloud platform according to claim 2, characterized in that The business knowledge graph matching unit includes: Knowledge Graph Storage Sub-unit: Store the semantic network related to business rules, including entities, relationships, and rule sets; Semantic Parsing and Vectorization Sub-unit: Convert data records into vector representations that can be compared and matched with the knowledge graph; Rule Matching Engine: Based on graph traversal, detect whether data records violate business rules; Conflict Frequency Calculation Sub-unit: Statistically calculate the number of occurrences of semantic conflicts and output the semantic conflict frequency.

4. The data service and processing optimization system based on the cloud platform according to claim 1, characterized in that, The processing strategy generation module includes a threshold division unit: when the comprehensive credibility score is less than, it is marked as high-trust core data and assigned to the deterministic calculation channel; when is less than, it is marked as low-trust edge data and assigned to the probabilistic processing channel.

5. The data service and processing optimization system based on a cloud platform according to claim 4, wherein The comprehensive credibility score is normalized to [0, 1]. When occurs, one of the following operations is automatically triggered according to the real-time pressure of the data stream: a. If the current deterministic channel load rate L core <η safe , promote to core data, η safe take 70%; b. Otherwise, degrade it to edge data and inject it into the probabilistic channel.

6. The data service and processing optimization system based on a cloud platform according to claim 2, characterized in that The Cross-channel Resource Cooperative Scheduling Module specifically includes: Fixed Time Window Deployment Unit: Dynamically adjust the length of the computing time window of the data according to the comprehensive credibility score of the core data. The calculation of the time window length considers the relationship between the comprehensive credibility score and the historical score average, so that high-credibility data can obtain longer processing time. Adopt a linear mapping model to make the allocation of parallel computing resources proportional to the comprehensive credibility score of each data time window. Based on the sum of the comprehensive credibility scores of all core data time windows, calculate its proportion by combining the comprehensive credibility score of a single time window and then multiply it by the total available computing resources to obtain the computing resources that can be allocated for this time window; Fuzzy Computing Unit: For edge data, the number of fuzzy computing units is determined by the noise density within the current time window. Based on the noise density of text data, the noise coefficient of numerical data, combined with the noise density tolerance coefficient and the maximum number of fuzzy computing units, calculate the actual number of fuzzy computing units required; Cross-channel Task Migration Unit: Includes migration from the deterministic computing channel to the probabilistic processing channel and migration from the probabilistic processing channel to the deterministic computing channel.

7. The data service and processing optimization system based on the cloud platform according to claim 6, wherein, The calculation accuracy of the fuzzy calculation unit is adaptively adjusted according to the semantic conflict frequency. When the semantic conflict frequency is high, the calculation accuracy decreases accordingly to reduce the calculation overhead; when the semantic conflict frequency is low, the calculation accuracy is increased.

8. The data service and processing optimization system based on a cloud platform according to claim 6, characterized in that The conditions for the migration of the deterministic calculation channel to the probabilistic processing channel include: When the processing time of the core data time window exceeds its allocated time window and its comprehensive credibility score drops below the credibility threshold, the data of this time window is degraded to the probabilistic processing channel to release the core computing resources. The amount of migrated computing resources is calculated by multiplying the original computing resources of the time window by the ratio of its current comprehensive credibility score to the high credibility threshold. The conditions for the migration of the probabilistic processing channel to the deterministic calculation channel include: For edge data, after fuzzy calculation, if the updated comprehensive credibility score exceeds the adjustment range of the high credibility threshold, it is migrated to the deterministic calculation channel and computing resources are allocated to it. The allocation method of the computing resources is proportional to the comprehensive credibility score, so that the edge data with high credibility obtains more computing resources.

9. The data service and processing optimization system based on a cloud platform according to claim 6, wherein The linear mapping model is expressed as: where K is the total number of current core data time windows, and R total is the total available computing resources, is the computing resources allocated to the k-th core data window, is the comprehensive credibility score of the i-th time window, is the comprehensive credibility score of the j-th time window.

10. The data service and processing optimization system based on a cloud platform according to claim 6, characterized in that, The number U of fuzzy computing units fuzzy is bound to the noise density: Among them, is the noise density of text data, is the noise coefficient of numerical data, ρ noise is the noise density tolerance coefficient, U max is the maximum number of fuzzy computing units.

Images