Principal Component Analysis Algorithm and Spatial Mapping Construction for Feature Extraction of University Language Text Data

With the arrival of the big data era, a huge amount of text data of college language is generated, and how to manage these text data efficiently and mine useful information has become the focus of many scholars.The study first preprocesses and represents the university language text data, proposes a feature screening method based on Shannon entropy and JS-scatter, and then combines the principal component analysis algorithm with the dimensionality reduction of the extracted features on this basis.Subsequently, a pre-trained high-dimensional word vector spatial mapping model is introduced to generate richer semantic representations, and a pre-trained high-dimensional word vector spatial mapping model based on the pre-trained high-dimensional word vector spatial mapping model is designed.Finally, the method proposed in this paper is tested experimentally.Under different feature dimensions, the macro-averages of this paper's method are 72%, 44.2%, 67.1%, and 3.3% higher than those of IG, PMI, ANOVA, and JS methods.At the feature dimension k=350, the macro-mean of this paper's method is 0.853, when the classification effect reaches the optimization.In the spatial mapping relationship of word vectors, the accuracy of the mapping of this paper's method also reaches 11.2% for the words with word frequency sorted from the first 5000 to the first 6000.This proves the effectiveness and feasibility of this paper's method.