Application Of Improved LSTM Excess Reserves In The Bank Business

With the rapid development of commercial economy,the commercial bank’s branch network to undertake a large number of cash transactions.Because of the cash business uncertainty,commercial bank branches have been unable to accurately estimate the daily need to use cash reserves that cover.Aiming at the problem,this paper put forward based on improved grey Wolf LSTM neural network combination model of the algorithm and its application in bank forecasts cover.First of all,introduced the model improved the Wolf algorithm,the original gray Wolf algorithm based on impact factor is proposed on the basis of nonlinear strategy and Newton iteration to enhance improvement strategies,to make up for a widespread species algorithm easy to fall into local traps and the optimization result is not stable,the precision is not high question;Second,portfolio model using gray Wolf algorithm optimization ability quickly,the LSTM neural network’s parameters optimization search,solve the limitation of its super LSTM neural network parameter problem;Finally,the application of the combined neural network model of building in terms of commercial bank branch cash cover prediction,solve the branches of commercial Banks can predict cash usage problem.The experimental stage,random fuxin region of a bank branch network operating cash usage history data,using different models to forecast the data.Results show that compared with the traditional prediction method ARIAM algorithm and simple LSTM neural network error between the predicted values and the real value of the LSTM neural network combination model based on improved grey Wolf algorithm error reduced by 72% and 16%,respectively,with the capacity for data accurate forecasts.In the experiments on real production data set show that the model prediction data have the characteristics of high precision,high stability.