Aiming at the problem of insufficient accuracy of the regional flood disaster risk quantitative assessment method, an improved support vector machine model based on the butterfly optimization algorithm was constructed and applied to the flood disaster risk assessment and spatio-temporal characteristics analysis in Heilongjiang Province in the past 15 years. The results showed that during the study period, the overall flood risk level in Heilongjiang Province fluctuated significantly in the early stage, but gradually stabilized in the later stage, and showed a spatial distribution pattern of high in the northwest and low in the southeast. Among them, the flood risk level in the Daqing area was the lowest, the risk level in the Suihua area was the highest, and the risk level in the rest of the areas had a clear downward trend with the inter-annual variation. Water production modulus, per capita GDP, monthly strongest precipitation, proportion of total output value of agriculture, forestry and fishery, natural population growth rate, number of health care beds per 10000 people, and total storage capacity of 10000 hectares of reservoirs were the key driving factors for changes in flood risk. Compared with the traditional support vector machine model and the improved support vector machine model based on the imperialist competitive algorithm, the constructed BOA-SVM model, mean absolute error was decreased by 38.15% and 9.18%, the mean square error was decreased by 58.5% and 21.56%, the mean absolute percentage error was decreased by 35.23% and 11.42%, respectively, and the model fit was excellent. The coefficent of determination was increased by 0.62% and 0.12% respectively, indicating that the BOA-SVM model had more advantages in terms of fit, adaptability, stability, reliability and evaluation accuracy. The research results can provide a model for flood disaster risk assessment, and provide reference for effective regulation and reduction of regional flood disaster risk.