Hyperparameter Tuning and Its Application in Sentiment Analysis with Convolutional Neural Network

In this paper, we explore hyperparameter tuning methods for a convolutional neural network (CNN) applied to sentiment analysis on the IMDB movie reviews dataset. Four popular hyperparameter optimization methods including random search, hyperband search, Bayesian optimization, and genetic algorithm are employed to search for the optimal hyperparameters. The best parameters from each tuning strategy are then used to train a CNN model, and their classification performance is evaluated and compared based on confusion matrices and accuracy metrics. Experimental results demonstrate that although all methods yield competitive performance with accuracies ranging between 87 % and 88 %, the model tuned by the genetic algorithm obtains the highest validation accuracy. This work highlights the effectiveness of automated hyperparameter tuning in enhancing CNN-based sentiment analysis.