<p>This study modeled aquatic insect abundance and biodiversity indices in the southeastern coastal regions of Bangladesh using statistical and machine learning algorithms. Insect abundance and water quality parameters were analyzed across multiple sites in 2024. A statistical model, i.e., multiple linear regression (MLR) and three machine learning models, namely, multilayer perceptron (MLP) random forest regression, and support vector regression, were used to predict insect abundance and Shannon–Wiener index. The results showed that MLP outperformed, achieving R² improvements of 19.49% and 30.73% for insect abundance and Shannon–Wiener index, respectively, compared to MLR. Moreover, analysis of the optimal input combinations showed that the MLR and MLP models explained over 70% and 80% of the variance, respectively, when using BOD, DO, WH, WS, pH, and WT, contributing to 90.57% and 85.82% of the total prediction accuracy of these models. Similarly, for the Shannon–Wiener index, the MLR and MLP models explained over 65% and 80% of the variance, respectively, when using NH₃, CO₂, WT, NO₂, pH, and TDS, resulting in 94.93% and 90.02% of the overall prediction accuracy. This study emphasized the role of water quality in modeling aquatic insects and demonstrated the advantages of machine learning-based approaches in ecological modeling.</p>