Organic matter (OM), phosphorus (P) and potassium (K) in black soil play a crucial role during crop growth. Studying the distribution of nutrient elements in black soil and carrying out quantitative calculation of element content is of great significance to the scientific management of black soil and environmental protection. Based on 80 black soil samples collected from Nehe city, Heilongjiang province and hyperspectral measured data, the correlation between 4 forms of spectral reflectance including original, first differential, absorbance transformation, and first differential of absorbance and soil OM, P, K contents were calculated, and the correlation coefficient method was used to extract the sensitive bands. For the optimization of parameter values in the machine learning model, the bat algorithm (BA) was introduced and combined with Adaboost model. Using BA to optimize the two important parameters of Adaboost, namely, maximum iterations n and weak learner weight reduction factor v, selecting CART decision tree and determination coefficient as the weak regression learner of the model and objective function value of the parameter optimization respectively, the BA-Adaboost model was constructed for estimating soil OM, P, K contents. The results showed that the BA-Adaboost model could quickly search the optimal global parameters, and the accuracy and reliability of the Adaboost were improved significantly after BA optimization, where the prediction accuracy of OM was the highest among the three nutrient elements. And determination coefficient and root mean square error were 0.864 and 0.152 g·kg^-1 respectively, which were improved by 14.2% and 25.4% compared with before optimization. Therefore, the constructed BA-Adaboost model has potential in the hyperspectral estimation of soil element content and is an efficient estimation method.