Scholarlink

<p>Due to the rise of mobile phones that are often the only cameras on hand, people share images on social media or while giving feedback on a product, service, or venue. Therefore, analyzing people’s feelings through facial expressions has become crucial, making sentiment analysis (SA) an essential tool for understanding user opinions and emotions in various fields, including marketing and social media analysis. However, traditional approaches to SA mostly focused on textual data, neglecting the rich information that images can provide, especially in conveying sentiments through facial expressions. Additionally, the application of feature selection techniques that can optimize classification accuracy remains unexplored. In order to make use of images for SA, an integrated approach for evaluating sentiment from images is proposed that includes feature extraction using a novel multi-transform fusion method, optimal feature selection using dual moth flame optimization, influenced by the moth flame optimization algorithm, and an ensemble classification approach to predict the overall sentiment as positive, negative, or neutral. The system’s ability to analyze image sentiment on datasets such as FER2013 and JAFFE is demonstrated through experiments. The proposed model, when compared to the current systems, shows improved accuracy by an average of approximately 3.42% and 7.76% on the two datasets, respectively.</p>

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

Optimization for Image Sentiment Analysis Using Novel Dual Moth Flame Algorithm

  • 期刊论文

  • 发表日期：

    2025年2月3日

查看全文
  • Siddhi Kadu,
  • Bharti Joshi,
  • Pratik Agrawal

摘要

Due to the rise of mobile phones that are often the only cameras on hand, people share images on social media or while giving feedback on a product, service, or venue. Therefore, analyzing people’s feelings through facial expressions has become crucial, making sentiment analysis (SA) an essential tool for understanding user opinions and emotions in various fields, including marketing and social media analysis. However, traditional approaches to SA mostly focused on textual data, neglecting the rich information that images can provide, especially in conveying sentiments through facial expressions. Additionally, the application of feature selection techniques that can optimize classification accuracy remains unexplored. In order to make use of images for SA, an integrated approach for evaluating sentiment from images is proposed that includes feature extraction using a novel multi-transform fusion method, optimal feature selection using dual moth flame optimization, influenced by the moth flame optimization algorithm, and an ensemble classification approach to predict the overall sentiment as positive, negative, or neutral. The system’s ability to analyze image sentiment on datasets such as FER2013 and JAFFE is demonstrated through experiments. The proposed model, when compared to the current systems, shows improved accuracy by an average of approximately 3.42% and 7.76% on the two datasets, respectively.