| Summary: | 碩士 === 實踐大學 === 資訊科技與管理學系碩士班 === 107 === The corporate financial distress prediction model has always been an important research subject for both academic and industry community. The early warning of corporate financial distress is an important issue for investors and creditors. In addition to the bad debts brought to the bank, the company’s non-warning default will bring huge losses to investors or creditors. Therefore, how to establish a reliable corporate financial distress prediction model to assist investors and creditors making decisions becomes an extremely important issue. There are many related works of literature on financial distress prediction models. Most of the research is based on the financial ratio. In the past, scholars have proposed many financial distress prediction models. The purpose of this study is to use a neural network to construct a corporate financial distress prediction model based on the company's financial ratio so that investors can refer to the short-term or long-term financial crisis forecast.
This study used “Back Propagation Neural Network” to construct corporate financial distress prediction model. The sample of this study was analyzed by Taiwan listed company as an empirical sample from 2000 to 2015. The simulation data was selected from 152 financial crisis companies and 385 financially sound companies. Based on the past literature, 12 financial ratios were used in this study as input variables. During computer simulations, two-thirds of company data was used as training data, and the remaining one-third was used as test data. To evaluate the performance of the prediction model, four performance indicators which were accuracy, precision, recall, and F1-score were used. Computer simulations showed that the model proposed in this study had the good predictive ability in the forecast of one year, the first two years, and the first three years before the financial crisis, and can be used as a reference for assessing whether the company's finances are sound.
|
|---|
