Scholarlink

In recent years researchers has been applying different machine learning and deep learning approaches on textual data. Preprocessing of textual data is one of the challenging tasks and it is getting enhanced day by day. Several vectorization techniques for mapping a string to a feature vector are compared and machine learning and deep learning approaches are applied for multi-class classification and binary class classification on the well-known datasets like Amazon Fine Food Review. Several vectorization techniques like Countvectorizer, TF-IDF, Hash vectorizer, Tokenizer, and embedding technique Glove are used in the comparison. The classifiers that are used in this work are Naïve Bayes, Logistic Regression, K-nearest neighbor, Multilayer perceptron (MLP), Convolution Neural Network (CNN), and Long Short Term Memory(LSTM). It has been found that all vectorization techniques are not compatible with every classifier as the accuracy depends on the creation of the feature vector. A graph has been drawn to compare the testing accuracy of different classifiers with the different tokenizers.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

A Comparative Analysis on Deep Learning Approaches for Sentiment Analysis of Food Review Data with Class Imbalance

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Pritha Banerjee,
  • Animesh Bhandari,
  • Jayita Saha,
  • Chandreyee Chowdhury

摘要

In recent years researchers has been applying different machine learning and deep learning approaches on textual data. Preprocessing of textual data is one of the challenging tasks and it is getting enhanced day by day. Several vectorization techniques for mapping a string to a feature vector are compared and machine learning and deep learning approaches are applied for multi-class classification and binary class classification on the well-known datasets like Amazon Fine Food Review. Several vectorization techniques like Countvectorizer, TF-IDF, Hash vectorizer, Tokenizer, and embedding technique Glove are used in the comparison. The classifiers that are used in this work are Naïve Bayes, Logistic Regression, K-nearest neighbor, Multilayer perceptron (MLP), Convolution Neural Network (CNN), and Long Short Term Memory(LSTM). It has been found that all vectorization techniques are not compatible with every classifier as the accuracy depends on the creation of the feature vector. A graph has been drawn to compare the testing accuracy of different classifiers with the different tokenizers.