Investigation of Classification Model for Text-Based Sentiment Analysis
Ian Cedric Ng Man King, Yang Li, Teoh Yan Qin, Humaira Ashraf, Uswa Ihsan, Satbir Singh Sehgal · 2024
This research is dedicated to training an effective text sentiment analysis model based on the Amazon review dataset. To evaluate model performance, Recall and F1 scores were adopted as the evaluation metric and introduced grid search and random search techniques to tune hyperparameters to ensure optimal model effectiveness. In data preprocessing, operations like data cleaning, tokenization, and feature extraction were performed. After training the sentiment analysis model with machine learning, its performance on positive and negative emotion categories was evaluated and used the confusion matrix to present the results. On the other hand, Optimization of hyperparameters through grid search and random search and imported the model using Flask framework was implemented to provide an intuitive interactive interface for users to input text and obtain sentiment analysis results, further enhancing system user experience and availability. Through comprehensive performance evaluation, the system’s robustness and accuracy was tested. In order to further enhance overall sentiment analysis performance, future work will concentrate on growing the dataset, investigating the integration of additional deep learning techniques, and experimenting with additional hyperparameter optimization strategies.