The rise in cyanobacterial harmful algal blooms (CHABs), driven by eutrophication and climate change, necessitates understanding cyanotoxin conditions to mitigate risks. However, limited studies have explored the influencing factors of cyanobacterial community and cyanotoxins in freshwaters using large datasets and data-driven models (DDMs). This study examines cyanobacterial composition across regions and relates it to microcystin-LR, -RR, -LA, -YR, -LF, and -LY as well as to the taste and odor compounds geosmin and 2-methylisoborneol. It also assesses the impact of environmental factors using interpretable machine learning. In the Nakdong River (NR) and Daecheong Lake (DL), differences in cyanobacterial community emerged regardless of nitrogen to phosphorus (N/P) levels. Warmer NR areas exhibited prevalent Microcystis during CHAB events. 1300-1700 Microcystis cells/mL produced 0.03 μg microcystin-LR/L (model accuracy 95 %) and DDMs predicted geosmin and 2-methylisoborneol less accurately. Random forest models using environmental factors predicted Microcystis dominance with 52 % accuracy at NR and DL, highlighting water temperature (positively) and nutrient levels (total N/P or NH₃-N/PO₄-P, negatively) as the primary factors influencing Microcystis dominance. Higher N/P contributed to higher Microcystis dominance only at DL when the water temperature was high. Phosphorus input was significantly affected by rainfall in NR but not DL. This study demonstrates toxicity predictions using environmental variables through DDMs and underscores the need to manage nutrient pollution sources to prevent Microcystis proliferation and MC-LR in surface waters.