The Financial Statements Of Listed Companies In Fraud Identification

While financial reporting fraud in china capital market has become more prevalent and costly in recent years, fraud detection has been badly lagging. The paper first selects the red flags of fraudulent financial statements, and then, develops the prediction mode using the data in shanghai stock market. The effect of the prediction mode is evaluated at last. In the case of prior probability of fraud financial reporting in population is too small, we will not get the enough samples to estimate the predict mode. If we adapt unequal probability sample design, the parameter estimation methods need to be modified. The paper present the methods to fit logistic model and neural network model under unequal probability sample design. In addition, because the neural network output can only be seen as the distance to classification surface which is different from what we want, the probability of fraud financial reporting, the paper present the condition of training the neural network to output the posterior probability.