Research on urban ecological environment vulnerability prediction method based on FII-LSTM

Constructing scientific and effective prediction methods for the vulnerability of urban ecological environment is of great significance for effectively managing and controlling the risks of urban ecological environment vulnerability. Existing prediction methods often overlook the self-similar characteristics of the evolution trends in complex nonlinear systems, resulting in low accuracy of the prediction results. A prediction method for urban ecological environment vulnerability based on the Fractal Interpolation (FII) - Long Short-Term Memory (LSTM) model was proposed. Using the FII model, constructed an affine transformation matrix as the Iterated Function System (IFS). By estimating the scaling factor of the FII model and using equidistant sampling measures, interpolation points were generated within the interpolation interval according to certain rules. These generated interpolation points were incorporated into the FII model to calculate the corresponding interpolation results, describe the fractal interpolation curve, and reveal the distribution state of the data. Constructed the LSTM neural network structure, selecting MSE and Adam as the loss function and optimizer, respectively. Used the holdout method to train the LSTM model, establishing a method for predicting urban ecological environment vulnerability. Based on multidimensional data from 35 cities in China, the model's prediction accuracy and stability were validated. The comparison results indicated that the constructed FII-LSTM model had high prediction accuracy and stability, making it suitable for decision-making management in urban ecological environment vulnerability prediction. The FII-LSTM model was used to predict the ecological vulnerability of the sample cities in 2021-2025 and to make policy recommendations.