Abstract: In order to obtain the optimal prediction model for the shear strength of an unsaturated clay, the long and short-term memory neural network model (LSTM) is used as the basis, and four biological algorithms, namely Harris Hawk algorithm (HHO), pigeon swarm algorithm (PIO), Sparrow algorithm (SSA) and Capuchin algorithm (CAP) are also used in this paper. The clay dry density, water content, pore ratio and temperature are used as the input data. Four types of optimization models are constructed in this paper. The results show that: the HHO-LSTM model has the best fitting effect. The slope of the fitting equation of the model was 0.991, the root-mean-square error, the average absolute error, the determination coefficient and the model efficiency coefficient are 1.328 kPa, 1.125kPa, 0.991 and 0.991, respectively. The GPI is 1.875. The accuracy indexes of simulated internal friction angle are 0.693, 0.584, 0.969 and 0.954 degrees, respectively with a related GPI of 1.805. The HHO-LSTM model can be used as a recommended model to predict the shear strength of the unsaturated clay.

Key words: Unsaturated Clay, Shear Strength, Long and ShortTerm Memory Neural Network Model, Biological Algorithms, Harris Hawk Algorithm