[Objective] The changing patterns of plant diversity in the Daqing Mountain Nature Reserve in Inner Mongolia under different environmental factors were analyzed. And the relationship between plant diversity in mountainous areas and the environment was clarified， in order to provide a scientific basis for the research， evaluation， protection and comprehensive management of the diversity in mountain ecosystems. [Methods] The deep learning algorithm were employed to construct a plant diversity index model， and the model accuracy was validated. Subsequently， the spatial distribution of plant diversity in Daqing Mountain was predicted， and the various patterns of plant diversity under different environmental factors were analyzed. [Results] ① There were a total of 108 plant species in the study area， belonging to 77 genera and 31 families. The plant diversity on the shady slopes was greater than that on the sunny slopes. ② The slope had the greatest relative contribution （42%） to the Shannon-Wiener index （H’）， Simpson dominance index （D） and Pielou evenness index （J）， followed by temperature vegetation dryness index （TVDI， 25%）， temperature （17%）， NDVI （8%） and solar radiation （8%）. Temperature and solar radiation had the largest relative contribution to the Margalef richness index （R） （38%）， slope （9%）， aspect （8%）， and NDVI （7%）. ③ The predicted results of H’， D， J， and R all showed strong agreement with measured values， with mean absoute error （MAE） values of 0.08， 0.03， 0.03， and 0.05， and mean square error （MSE） values of 0.020， 0.003， 0.002， and 0.004， respectively. Further linear regression analysis between simulated and observed values in the training set revealed that the R2 values for each diversity index reached 0.86， 0.93， 0.92， and 0.99， respectively. ④ The value range of H’， D， J and R in the Daqing Mountain were 3.87， 0.83， 0.95 and 4.12， respectwely. ⑤ H’， D， and J were linearly negatively correlated with slope， TVDI， land surface temperature （LST） and solar radiation， and linearly positively correlated with NDVI. R was linearly negatively correlated with LST， solar radiation and slope， and linearly positively correlated with aspect and NDVI. Overall， plant diversity was linearly negatively correlated with LST， solar radiation and slope， and linearly positively correlated with NDVI. [Conclusion] Deep learning methods can be feasibly used to predict the spatial distribution of plant diversity in mountainous landforms. This method can deepen our understanding of the complex relationship between plant diversity and environment.