The Role of Improved Neural Network Algorithm in Convertible Bond Market Analysis

Abstract

In order to accurately analyze the convertible bond market, reasonably determine its interest rate and volatility, reduce the risk of convertible bond buyers, and improve their purchase rate, the current convertible bond algorithm is analyzed. Combined with Black-Scholcs pricing model, Radial Basis Function (RBF) neural network algorithm is improved, and RBF neural network is used for orthogonal least squares calculation. The central position of RBF neural network is guaranteed to remain unchanged while the algorithm remains unchanged. The least square method is used to calculate the weight vector of the network again, and the width of the RBF network is set according to the obtained data, in order to find the optimal width. Relevant data samples of convertible bonds are collected by Tonghuashun Ifind Financial Data System and other systems, and the data are simulated and calculated by Matlab software, and the pricing simulation results of convertible bonds under three neural networks are compared. The results show that the modified RBF neural network algorithm can evaluate the price of convertible bonds. And its calculation error is less than the other two neural network algorithms. This indicates that the RBF neural network algorithm designed by us can accurately predict the pricing of convertible bonds in combination with Black-Scholcs pricing, which provides guidance and data support for the pricing of convertible bonds and market investment in the future.