Open Access

ARTICLE

AI-Driven Sentiment Analysis: Understanding Customer Feedbacks on Women’s Clothing through CNN and LSTM

Phan-Anh-Huy Nguyen^*, Luu-Luyen Than

Faculty of Economics, Ho Chi Minh City University of Technology and Education, Ho Chi Minh City, 700000, Vietnam

* Corresponding Author: Phan-Anh-Huy Nguyen. Email:

Intelligent Automation & Soft Computing 2025, 40, 221-234. https://doi.org/10.32604/iasc.2025.058976

Received 25 September 2024; Accepted 28 February 2025; Issue published 14 April 2025

Abstract

The burgeoning e-commerce industry has made online customer reviews a crucial source of feedback for businesses. Sentiment analysis, a technique used to extract subjective information from text, has become essential for understanding consumer sentiment and preferences. However, traditional sentiment analysis methods often struggle with the nuances and context of natural language. To address these issues, this study proposes a comparison of deep learning models that figure out the optimal method to accurately analyze consumer reviews on women's clothing. CNNs excel at capturing local features and semantic information, while LSTMs are adept at handling long-range dependencies and contextual understanding. By integrating these two deep learning techniques, our model aims to achieve better performance in sentiment classification. The models were trained and evaluated on a dataset of women's clothing reviews sourced from Kaggle. The dataset was pre-processed to clean and tokenize the text data, and word embeddings were used to represent words as numerical vectors. The CNN component of the model extracts local features from the text, while the LSTM component captures long-range dependencies and contextual information. The outputs of the CNN and LSTM layers are then concatenated and fed into a fully connected layer for final sentiment classification. Experimental results demonstrate that the hybrid model outperforms traditional machine learning techniques and other deep learning models in terms of accuracy, precision, recall, and F1-score. By accurately classifying sentiment, identifying key themes, and predicting future trends, our model can provide valuable insights to businesses in the apparel industry. These insights can be used to improve product design, marketing strategies, and customer service, ultimately leading to increased customer satisfaction and business success.

---

Keywords

---