Predicting the Growth Value of Technology Enterprises with an Optimized Back-propagation Neural Network

Abstract

Accurate assessment of the development value of technology-based small and medium-sized enterprises (SMEs) is beneficial to the effective support of these enterprises. This paper briefly introduced technology-based SMEs and factors affecting their development values. Then, the development value of enterprises was evaluated using a back-propagation neural network (BPNN) algorithm. The BPNN algorithm was improved by particle swarm optimization (PSO). The specific improvement method was using particles in the PSO to represent adjustable parameters in the BPNN algorithm. Each particle represented a parameter scheme, and during the training of the improved BPNN, the iteration of the particle swarm in the PSO was used to replace reverse parameter adjustment based on training error. Finally, the PSO-BPNN algorithm was simulated and compared with the extreme learning machine (ELM) and traditional BPNN algorithms. The results showed that the ELM quickly obtained the weight parameters by generalized inverse matrix. The PSO-BPNN algorithm had faster convergence than the traditional BPNN algorithm in training. The former converged to stability after about 600 iterations, while the latter converged to stability after about 800 iterations. Moreover, the improved BPNN algorithm had a smaller mean square error after stabilization. The PSO-BPNN algorithm had the smallest calculation error for the development value assessment number (1.00%-1.25%) and correlated significantly with the market value growth rate (P = 0.002). The PSO-BPNN algorithm can effectively evaluate the development value of technology-based SMEs.