CN-121997265-A - Multi-source information decision fusion method and system for multi-granularity multi-view rough set

CN121997265ACN 121997265 ACN121997265 ACN 121997265ACN-121997265-A

Abstract

The invention relates to the technical field of multi-source information processing, in particular to a multi-source information decision fusion method and a multi-source information decision fusion system of a multi-granularity multi-view rough set. And constructing a multi-granularity hierarchical structure based on the view subset, calculating rough set approximation precision and attribute dependence under different granularity levels, and generating a local decision rule set. And carrying out weighted aggregation on the local decision rule set by utilizing a multi-source information fusion algorithm to construct a decision evaluation parameter based on consistency measurement. According to the invention, the problem of detail loss of the information sheet under a single view angle and under a single granularity is effectively solved by calculating the reverse driving decision cost of a multi-view and multi-granularity cooperative mechanism, and the accuracy and the robustness of decision analysis under a complex uncertain environment are remarkably improved by feedback adjustment based on dynamic weights, so that the deep mining and efficient fusion of multi-source information are realized.

Inventors

LI MINSONG
MIAO DECHENG
HUANG XIONGHUA
LIU XIAOLIANG
Zeng Puhua

Assignees

韶关学院

Dates

Publication Date: 20260508
Application Date: 20260128

Claims (10)

1. The multi-source information decision fusion method of the multi-granularity multi-view rough set is characterized by comprising the following steps of: S1, acquiring an original multi-source data set of a target object, performing data cleaning and normalization processing on the original multi-source data set, constructing a multi-source information decision table containing conditional attributes and decision attributes, and dividing the multi-source information decision table into a plurality of independent view attribute subsets according to data source characteristics; s2, constructing a multi-level multi-granularity space structure by setting different granularity criteria for each view attribute subset, and calculating a binary relation matrix among objects under each granularity level to generate a multi-view multi-granularity knowledge base; S3, calculating a lower approximation set and an upper approximation set under each granularity level by utilizing a rough set theory based on the multi-view multi-granularity knowledge base, calculating attribute importance indexes according to the lower approximation set and the upper approximation set, and further extracting local decision feature vectors of each view under different granularities; And S4, constructing a weighted fusion model based on attribute importance, inputting the local decision feature vectors under all views and granularity into the weighted fusion model for consistency aggregation calculation, generating a global decision score as a decision evaluation parameter of the target object, and mapping the decision evaluation parameter into the priority, the selection probability or the sorting weight of the object, thereby determining the final classification result and the decision response strategy of the target object.
2. The multi-source information decision fusion method of the multi-granularity multi-view rough set according to claim 1, wherein the process of constructing the multi-source information decision table specifically comprises: Detecting the missing values in the original multi-source data set, and filling the missing values by using the average value of the similar samples; mapping the numerical data to a standard interval by adopting a maximum and minimum normalization method; And aligning the conditional attribute with the decision attribute to form a two-dimensional data matrix serving as the multi-source information decision table.
3. The multi-source information decision fusion method of the multi-granularity multi-view rough set according to claim 1, wherein the process of constructing the multi-granularity spatial structure of the multi-hierarchy specifically comprises: Setting a group of incremental neighborhood radius threshold sequences; And for each view attribute subset, judging the neighborhood similarity among samples by utilizing each threshold value in the neighborhood radius threshold value sequence, and classifying the samples meeting the similarity condition into the same information grain, so as to form a plurality of granularity levels from thin to thick.
4. The multi-source information decision fusion method of the multi-granularity multi-view coarse set according to claim 3, wherein the process of calculating the attribute importance index specifically comprises: calculating the dependency value of the decision attribute relative to the current view attribute subset at a specific granularity level; calculating the dependency variation of the current view attribute subset after eliminating a certain attribute; and taking the dependency variation as an attribute importance value of the attribute under the current granularity level.
5. The multi-source information decision fusion method of multi-granularity multi-view coarse set according to claim 4, wherein the process of extracting the local decision feature vector specifically comprises: screening out core attributes with the attribute importance value higher than a preset importance threshold value; constructing a reduced local decision rule based on the core attribute; substituting the object to be decided into the local decision rule, outputting a local prediction classification label of the object under the current view and the current granularity, and taking the local prediction classification label as an element of the local decision feature vector.
6. The multi-source information decision fusion method of the multi-granularity multi-view rough set according to claim 5, wherein the process of constructing the weighted fusion model based on the attribute importance degree specifically comprises the following steps: Calculating entropy weight values of different granularity levels under each view by using an information entropy algorithm; Distributing a fusion weight coefficient for each local decision feature vector according to the entropy weight value; and constructing a weighted voting matrix as a calculation core of the weighted fusion model.
7. The multi-source information decision fusion method of the multi-granularity multi-view rough set according to claim 1, wherein the process of generating the global decision score specifically comprises: converting the local prediction classification labels of each granularity of each view into corresponding independent heat coding vectors; Carrying out weighted summation on the single thermal coding vector and the corresponding fusion weight coefficient to obtain a fusion probability vector; and extracting the maximum value in the fusion probability vector as the global decision score.
8. The multi-source information decision fusion method of a multi-granularity multi-view coarse set of claim 3 further comprising a multi-granularity adaptive adjustment step: And (3) monitoring the change of external user demand parameters in real time, dynamically resetting a decision reference set when a new demand instruction is received or when the global decision score is lower than a preset confidence threshold value, automatically adjusting the numerical value of the neighborhood radius threshold sequence, and re-executing the steps S2 to S4 to realize the self-adaptive update of the fusion weight and the decision result along with the change of the demand.
9. The multi-source information decision fusion method of a multi-granularity multi-view coarse set of claim 8 further comprising a collision detection step: calculating conflict rates among the local prediction classification labels under different views; when the conflict rate exceeds a security threshold, the view attribute subset with the highest conflict rate is rejected, and the aggregation calculation in S4 is performed based on only the remaining view attribute subset.
10. A multi-source information decision fusion system for a multi-granularity multi-view coarse set, wherein the system is configured to implement the multi-source information decision fusion method for a multi-granularity multi-view coarse set according to any one of claims 1 to 9, the system comprising: The data preprocessing and view dividing module is configured to acquire an original multi-source data set, execute cleaning normalization processing, construct a multi-source information decision table and divide the multi-source information decision table into view attribute subsets; the multi-granularity space construction module is configured to construct a multi-level multi-granularity space structure based on the neighborhood radius threshold sequence and generate a binary relation matrix; The rough set approximation calculation and feature extraction module is configured to calculate a lower approximation set and attribute importance indexes and extract local decision feature vectors; And the multi-source weighted decision fusion module is configured to aggregate the local decision feature vectors by using a weighted fusion model and output a final classification result.

Description

Multi-source information decision fusion method and system for multi-granularity multi-view rough set Technical Field The invention relates to the technical field of multi-source information processing, in particular to a multi-source information decision fusion method and system for a multi-granularity multi-view rough set. Background The multi-source information processing refers to the technical field of collecting, processing and comprehensively analyzing heterogeneous data from different sensors, databases or networks through a computer technology, and aims to eliminate redundancy and contradiction among the data and extract high-value decision basis. With the development of the Internet of things and big data technology, the field is widely applied to the scenes of medical diagnosis, financial wind control, industrial fault detection and the like, and becomes a core means for solving the decision problem of a complex system. The traditional rough set decision method is to utilize a single granularity division standard or a single data observation view angle to carry out approximate classification and rule extraction on incomplete or uncertain information, and mainly relies on a static attribute reduction algorithm to carry out data dimension reduction processing. However, existing conventional rough set methods have significant drawbacks in facing highly complex multi-source heterogeneous data. Because it can only view data from a single view or single resolution, it often results in missing local key features and is not effective in capturing deep structural relationships within the data. Under the condition of large data noise or nonlinear attribute relationship, the single-layer processing mode is easy to generate a single-sided decision rule, so that the final classification accuracy is low, the robustness is insufficient, and the actual requirement of a high-precision decision scene is difficult to meet. Disclosure of Invention The invention aims to solve the defects in the prior art, and provides a multi-source information decision fusion method and a system for a multi-granularity multi-view rough set. In order to achieve the above purpose, the invention adopts the following technical scheme that the multi-source information decision fusion method of the multi-granularity multi-view rough set comprises the following steps: S1, acquiring an original multi-source data set of a target object, performing data cleaning and normalization processing on the original multi-source data set, constructing a multi-source information decision table containing conditional attributes and decision attributes, and dividing the multi-source information decision table into a plurality of independent view attribute subsets according to data source characteristics; s2, constructing a multi-level multi-granularity space structure by setting different granularity criteria for each view attribute subset, and calculating a binary relation matrix among objects under each granularity level to generate a multi-view multi-granularity knowledge base; S3, calculating a lower approximation set and an upper approximation set under each granularity level by utilizing a rough set theory based on the multi-view multi-granularity knowledge base, calculating attribute importance indexes according to the lower approximation set and the upper approximation set, and further extracting local decision feature vectors of each view under different granularities; And S4, constructing a weighted fusion model based on attribute importance, inputting the local decision feature vectors under all views and granularity into the weighted fusion model for consistency aggregation calculation, generating a global decision score, and determining a final classification result of the target object according to the global decision score. The process of constructing the multi-source information decision table specifically comprises the steps of detecting missing values in the original multi-source data set, and filling the missing values by using the average value of similar samples; mapping the numerical data to a standard interval by adopting a maximum and minimum normalization method; And aligning the conditional attribute with the decision attribute to form a two-dimensional data matrix serving as the multi-source information decision table. As a further scheme of the invention, the process for constructing the multi-level multi-granularity space structure specifically comprises the steps of setting a group of incremental neighborhood radius threshold sequences; And for each view attribute subset, judging the neighborhood similarity among samples by utilizing each threshold value in the neighborhood radius threshold value sequence, and classifying the samples meeting the similarity condition into the same information grain, so as to form a plurality of granularity levels from thin to thick. The process of calculating the attribute importance index spe