Data Dimensionality Reduction Based On Evolutionary Algorithm

The advent of the information age has added a lot of excitement to human life,but with it comes the occurrence of data disasters.The most effective way to solve data disasters is data dimensionality reduction,which is divided into feature extraction and feature selection.Feature extraction is a mapping between the original data.To extract relatively important data features,feature selection is to select important feature data from the original data.In this paper,feature extraction is used as a means of data dimension reduction.Some of the most popular feature extraction algorithms such as PCA(principal feature extraction),LE(Laplacian feature mapping),LLE(local feature mapping)are effective feature extraction methods.However,there are many influencing factors in popular learning algorithms,such as the number of K-nearest neighbors.Less,the selection of kernel function and dimensionality reduction have always been the main factors that perplex the effect of popular learning.Additional data noise has a lot of influence on the extracted data features,so there will be more problems to be solved in popular learning.1.Aiming at the problem of K-nearest neighbor data selection and dimensionality reduction in ISOMAP,immune cloning algorithm is applied to deal with K-nearest neighbor and dimensionality reduction in ISOMAP.Evolutionary combination makes K-nearest neighbor and dimensionality reduction optimal in different datasets.It can only extract the features of a single data set to achieve the specific analysis of a specific data set and formally unifies the problem of K-nearest neighbor and dimension reduction combination optimization,which makes the important features extracted and the relationship between the data clearer.2.2.In LLE(Local Feature Mapping),it is difficult to select K-Nearest Neighbor data and dimension reduction.Immune cloning algorithm is applied to deal with K-Nearest Neighbor and dimension reduction in LLE,so that both K-Nearest Neighbor and dimension reduction can achieve the optimal combination,which can provide better processing conditions for LLE.Among them,immune cloning algorithm is used to deal with K-Nearest Neighbor and dimension reduction in LLE.Because LLE is a local operation,the computational complexity is not large,so the noise removal process is given after choosing a reasonable dimension(the previous LLE chooses feature vectors continuously according to the priority of the eigenvalues,and this time in the case of limited dimension,random combination selection).The experimental results show that the effect is better than before.Much better.Considering the iterative computational complexity of evolutionary algorithms,we should also use a larger mutation probability to effectively improve the global search ability.