The Research On Incremental Attribute Reduciton Algorithm Decision System With Weakly Labeling

With the rapid development of information technology,a large amount of data will be produced in various application fields which has the characteristics of high dimensionality.If the learning model is directly trained on data with high dimensionality,it will not only require excessive memories,but also leads to the large computational time complexity.Meanwhile,in the real-world applications,it takes much work and cost to label all instances,so weakly labeled data is ubiquitous.If only the unlabeled data or the labeled data is used,some will cause the loss of the supervised information and poor classification results.Moreover,in the real-world,the data generally presents the status of dynamic updating.The traditional data mining algorithm cannot effectively use the results obtained before the data updates,thereby it needs to recompute based on the latest status of data,which will lead to a lot of repetitive work,and the efficiency of algorithm is poor.In summary,this paper focuses on the above-mentioned background to develop the research work with respect to feature selection.Firstly,in the third chapter of this paper,based on the granular computing theory,the concept of discernibility pairs of incomplete data is given from the perspective of discernibility,and the evaluation method of the relative significance of attributes is given.Then,an attribute reduction algorithm for weakly labeled in decision systems is designed.This algorithm continuously reduces the search space during the process of iteration,and improves the efficiency of attribute reduction.Furthermore,the data is often updated in real time.In order to update the reduction rules quickly,the dynamic updating mechanism of attribute reduction is further analyzed according to the dynamic changes of the instances.On this basis,an incremental attribute reduction algorithm under semi-supervised is designed.Finally,an example is given to illustrate the process of attribute reduction in incomplete data with weak labels and the updating process of attribute reduction results after the instance changes dynamically,and the feasibility and effectiveness of the proposed algorithm are verified by extensive experiments.Second,in the fourth chapter of this paper,an incremental attribute reduction and updating mechanism is proposed in view of the changes of attributes in weakly labeled data.On the basis of the previous work,the influence of attribute change on reduction results is analyzed in detail,and the corresponding incremental attribute reduction algorithm is proposed.After the change of attributes,the result of attribute reduction can be updated quickly on the basis of the original result of attribute reduction,which reduces a lot of repeated work and improves the updating efficiency of attribute reduction result.Finally,an example is given to illustrate the process of the algorithm in detail,and the high efficiency of the algorithm is demonstrated by experiments.Finally,the algorithm research in this paper is summarized and prospected.