Sentiment Analysis Of New Energy Vehicle Consumer Comments Based On Text Mining

Due to the improvement in national living standards,cars have transitioned from being luxury items to becoming necessities.Alongside traditional fuel-based cars,there is a significant emphasis on promoting the development of new energy vehicles by the government.In the era of the Internet of Things(Io T),it has become the norm for users to share their experiences with goods online.These user reviews contain emotional tendencies and offer substantial potential for analysis.However,with the rapid growth of online text data and the unique characteristics of irregularity and uncertainty exhibited by textual content,the conventional single-model approach is no longer sufficient to meet the demands of text mining.As a result,this paper proposes an optimization of sentiment classification models in the fields of machine learning and deep learning,respectively.It introduces a multi-level model that incorporates several single models to provide consumers and merchants with a reliable means of assessing sentiment tendencies.Furthermore,the paper utilizes the LDA topic analysis model to offer merchants valuable insights for improvement.In this paper,we selected user reviews of new energy vehicles from Auto Zone and Pacific Car.com as the analysis text.Using Python software,a collection of user reviews for 20 new energy vehicles spanning various price levels was obtained based on vehicle sales rankings.After careful data screening and pre-processing,a total of 47,300 reviews were gathered for analysis.This paper concentrates on the following research using machine learning,deep learning sentiment classification base models,and topic models:(1)The paper focuses on optimizing sentiment classification models in machine learning.By conducting reference comparison tests,the ada-boost model based on SAMME.R and the SVM model based on the Gaussian kernel function are established.Both models exhibit improvements in accuracy,precision,recall,and F1 value to a certain extent.Specifically,the ada-boost model achieves a recall of 82.34%,while the SVM model achieves an accuracy of84.98%.Leveraging the advantages of machine learning,such as low hardware requirements,it provides a suitable model for the initial screening of unlabeled text by new energy vehicle consumers for sentiment tendency.(2)The paper conducts a multi-layer fusion of sentiment classification models in the field of deep learning.To address the limitations of the BiGRU model in handling long sequences and the Text CNN model’s inability to accurately extract data location features,the multi-fusion models BiGRU-ACNN and ERNIE-BiGRU-CNN are introduced.The BiGRU-ACNN model incorporates an attention mechanism to obtain feature representations with contextual information,while the ERNIE-BiGRU-CNN model enhances the model’s ability to understand the original text by incorporating the ERNIE preprocessing model.Comparative experiments with single models demonstrate that both established models effectively improve the judgment of text sentiment tendencies.Among them,the ERNIE-BiGRU-CNN model performs best,achieving accuracy,precision,recall,and F1 values exceeding 89%,with precision improved to 92.29%.The optimized structure and algorithm of this model enable it to meet the requirements of merchants for sentiment analysis of large-scale data.(3)To provide more targeted suggestions,the paper employs Python software to create word cloud maps of theme features for each sample dataset.Additionally,the LDA theme model is employed to analyze the distribution of consumer preference themes across different price points of new energy vehicles and for different new energy vehicles within the same price point.These analyses provide valuable reference opinions for businessmen to improve their offerings,ensuring they better meet consumer needs and enhance the market competitiveness of their products.