Based on the cyanobacteria comprehensive index(I_c) constructed by the satellite imageries of Lake Taihu in 2004-2018, the random forest machine learning algorithm was used to analyze the relationship between meteorological factors and I_c, and quantitatively evaluate the importance measures and contribution rate of the main meteorological features. The results show that among the main meteorological elements such as light, temperature, water and wind, temperature plays a leading role in cyanobacterial comprehensive index, followed by wind speed and precipitation, and the influence of sunshine hours may be neglected. Among them, the most important measure of importance in temperature conditions is the annual average temperature, followed by the average temperature in winter and spring. The most important in the wind factors is the average wind speed in July. The dominant factor in the water condition is the cumulative precipitation in September. The optimal random forest model simulation value is basically consistent with the actual cyanobacteria comprehensive index, and the determination coefficient is 0.91. The random forest model simulation effect is better by the 0.01 significance test. Using the random forest model simulation value to evaluate the cyanobacteria blooms in Lake Taihu, the model simulation accuracy reached 86.7%. The simulation results of the five severe grades of the year model are completely consistent. The simulation values of the six grade models of the medium grade are consistent with the five years, and the simulation accuracy of the medium and above grades is 90.9%. The model can reflect the comprehensive effects of meteorological factors on the cyanobacteria comprehensive index, and the simulation effect on medium and severe cyanobacteria blooms is better. The random forest model is helpful to understand the dominant meteorological factors affecting cyanobacterial blooms under eutrophication conditions. The predictability of meteorological factors can promote the improvement of cyanobacterial bloom prediction and early warning ability.