A Modified Wavelet Neutral Network Model for Measuring Goodwill

Abstract

Recently years have witnessed the development of cultural enterprises. As one of the most valuable intangible assets, goodwill valuation for cultural enterprises has gained a lot of attention. An efficient method is to use Wavelet Neural Network (WNN) model for learning the predicted value of goodwill given a set of indicators. However, there are some issues of the basic WNN model. On one hand, the randomly determination of the initial state of the neural network leads to the possibility of converging to a local optimal point. To solve this problem, we propose to employ Genetic Algorithm (GA) for optimizing the initial parameters before model training. On the other hand, the training cost of basic WNN is typically big and its convergence speed is relatively slow. To accelerate the convergence speed, we introduce Levenberg-Marquardt (LM) algorithm for training. Besides, we conduct experiment to evaluate the performance of our proposed GA-LM-WNN algorithm.