作者: Eija Koskivaara
DOI:
关键词:
摘要: The aim of this study is to investigate the impact various pre-processing models on forecast capability artificial neural network (ANN) when auditing financial accounts. Hence, focus paper data. ANNs are selected for purposes because they capable learning complex, non-linear underlying relationships. Therefore, used model dynamics and relationships between account values in order find unexpected fluctuations. This uses a multi-layered with backpropagation algorithm. was built by using statements 31 manufacturing companies over four years. accounts were regarded as timeseries. data pre-processed different ways. Firstly, all scaled linearly. Secondly, linearly yearly basis. Thirdly, company And fourthly, best results achieved either or