A Systematic Study of Hyperparameter Tuning for Environmental Text Classification: Implications for Environmental Management

Environmental management increasingly relies on rapid and precise information analysis to resolve critical environmental problems. This study evaluated the effectiveness of hyperparameter tuning and its impact on automatic environmental text classification performance using different Machine Learning (ML) classifiers and term-weighting schemes. Our results indicated that hyperparameter tuning generally enhanced classification performance, with the eXtreme Gradient Boosting (XGBoost) classifier showing the highest performance. The study also highlighted the trade-off between performance improvement and computational cost, i.e., enhanced classification accuracy at the expense of increased execution time. Notably, hyperparameter sensitivity varied among ML classifiers. For example, the Multinomial Naive Bayes classifier was less sensitive to hyperparameter tuning under certain term-weighting schemes. These findings provide new insights into the relationships between hyperparameter optimization, classification performance, and computational efficiency in environmental text classification. They offer valuable guidance for selecting and tuning classifiers to support better-informed decisions in environmental management.