Analysis of English Writing Text Features Based on Random Forest and Logistic Regression Classification Algorithm

Abstract

The characteristics of English writing text in natural scenes are characterized by low character detection rate, difficulty in small character detection, and various character detection categories. In order to improve the classification effect of English-written texts, solve the problem of feature loss and precision reduction of the prediction model based on dimensionality reduction neural network classifier in the analysis process. An improved stacking model combining random forest and logistic regression is proposed to analyze the characteristics of written English texts. The model uses multiple undersampling, trains multiple random forests as primary classifier, and uses logistic regression as secondary classifier. Experimental results show that this model can effectively improve the classification efficiency of unbalanced text classification. While ensuring the main features, the accuracy of prediction is substantially improved. It is proved that the model has high practicability in analyzing the features of English writing texts.