Sentiment Analysis Of Micro Blog Comments Based On Integrated Learning

Weibo has the characteristics of fast user release of information and fast transmission of information,and has become the main social media for people.Netizens will express their emotional views on different events by posting microblogs and comments.Sentiment analysis of Weibo comments plays a major role in network monitoring,public sentiment guidance,and public opinion monitoring.Manually mining and analyzing the emotions contained in comments consumes a lot of manpower,and the efficiency of excavate emotions is very poorly,So more people are starting to do emotional analysis studies.This article uses different methods to analyze the sentiment of Weibo text,and the main work is as follows:Based on the shortcomings of previous word vector transformation models in determining the weight balance of text words,an optimized TF-IDF algorithm was used to improve the weight imbalance of similar corpora.The improved TF-IDF algorithm was combined with the Word2 Vec word vector transformation model,significantly increasing the weight of word vectors.We proposed the integration of emoticons with part of speech information and textual features to form a multi feature mixed vectorization matrix,which solves the problem of incomplete expression of meaning and emotion and polysemy in traditional texts,and provides support for improving the accuracy of the classifier model in the future.In the process of comparing the effectiveness of multiple classifiers,it was found that the accuracy evaluation index of traditional machine learning model algorithms is lower than that of deep learning models,but their efficiency is higher than that of deep learning models.In order to ensure the synchronous improvement of accuracy and efficiency,this paper combines the DNL classifier integrated by the decision tree,naive Bayes and Logistic regression in the traditional machine learning model with the RCCR classifier integrated by the RNN,CNN and RCNN in deep learning,that is,combining the respective advantages of the two learning models,and proposes the best1 model.The experimental results show that the Best1 ensemble model has improved in terms of accuracy,precision,recall,and F1 value compared to traditional machine learning models and deep learning models.The accuracy of the best1 classifier reaches 87.54%,which is nearly 20% higher than the DNL model with the highest accuracy in machine learning algorithms.The time used is shortened by nearly half compared to the deep learning ensemble model RCCR.From this,it can be concluded that the Best1 model has significantly improved both in terms of accuracy and efficiency.