The Optimization Research Of Double Transformation Algorithm In Multi-dimensional Sequence Data Analysis

With the widespread use of monitoring equipment,a large amount of multi-dimensional time series data can be collected in the fields of transportation,meteorology,and finance.The time series has the characteristics of randomness,continuity and periodicity.The time series is predicted and analyzed,and the predicted results can effectively guide production and life.The least squares support vector machine(LSSVM)model is a commonly used prediction model.The value of the penalty factor and radial basis width directly determines the prediction accuracy of the model.When predicting multi-dimensional time series,the high computational cost makes the LSSVM model unable to quickly determine the optimal parameter value.How to optimize the parameter selection of the LSSVM model and improve the prediction accuracy of the model is a research direction with practical application value.This thesis proposes a double transform algorithm(DTA),which uses the unitary transform and the hyperbolic rotation transform to solve the inverse of the Hessian matrix and updates the weight of the cost function.The Newton-Raphson algorithm(NRM)is then used to iteratively update the predicted value.When the predicted error is less than a set threshold,the predicted value is output.The DTA algorithm does not make any statistical assumptions on multi-dimensional time series data,so it retains the convergence and stability of the NRM algorithm and reduces the time complexity from o(7)M ~2(8)to o(7)M(8).Apply the DTA algorithm to the LSSVM model,first initialize the penalty factor and radial basis width of the LSSVM model,then iteratively update the parameters using unitary transformation and hyperbolic rotation transformation and find the optimal parameter values,and bring the parameter values into the LSSVM expression and output the prediction model.In order to verify the effectiveness of the algorithm,UCI KDD standard data set and Traffic Flow Data standard data set were selected in the Python language environment for simulation experiments.First of all,the bilateral algorithm(DTA)mentioned in this paper is compared with batch gradient descent method(BGD),Newton-Raphson method(NRM)and particle swarm optimization(PSO),and the stability and calculation time of different optimization algorithms are compared.Then use the above four optimization algorithms to select the penalty factor value and radial basis width value of the LSSVM model,and evaluate the prediction performance of the optimized LSSVM model.The simulation results show that the DTA algorithm can effectively reduce the calculation time and optimize the parameters under the premise of ensuring the stability of the algorithm.The LSSVM model optimized by the DTA algorithm has smaller errors,higher prediction accuracy,and better prediction results.Finally,the DTA-LSSVM algorithm is applied to the actual traffic business,the algorithm can effectively predict the road traffic in the next7 days.