Research on Label Distribution Dimension Reduction Algorithm Based on MLDA

Data dimension reduction can effectively keep the important features in the data, improve the classification performance of the data. Based on this, the model of Multi-label Linear Discriminant Analysis (MLDA) is introduced to reduce the dimension of data, and the Label Distribution Dimension Reduction Algorithm based on MLDA (MLDA-LDL) is proposed. Firstly, the original data is learned to complete feature dimensionality reduction, and the sample data after feature dimensionality reduction is obtained. Then, the training model of feature and label is bulit by the classical algorithm of label distribution paradigm. Finally, the prediction is completed through the training model. Experimental results on several public data sets show clear advantages of the MLDA-LDL algorithm.