Dimensionality Reduction Algorithm With Ordered Pair Of Normalized Real Numbers

Before classifying high-dimensional data, dimension-ality reduction algorithms are often used to process the data to achieve the purpose of reducing computational complexity and interference of redundant information, and improving the correct rate of classification. However, various dimensionality reduction algorithms have their advantages and disadvantages, and their performance varies greatly in different datasets, and it is difficult to choose the optimal dimensionality reduction algorithm. To address this problem, this paper introduces a new mathematical concept of ordered pair of normalized real numbers into the field of dimensionality reduction algorithm for the first time, and puts forward the new method of combining different dimensionality reduction algorithms through ordered pair of normalized real numbers, to realize the complementarity of different dimensionality reduction algorithms, improve the accuracy of subsequent classification. In addition, this paper also proposes the nearest neighbor classification algorithm of ordered pair of normalized real numbers, which realizes the classification of ordered normalized real number pairs. Through experiments on multiple real data sets and comparing a variety of improved k-nearest neighbor algorithms, it is verified that the proposed method has better performance.