在近幾年來，隨著電腦及硬體設備的提升和相關理論基礎的研究累積，人工智慧已經在各領域當中發光發熱，並且得到許多實務上的應用，而在近幾年金融科技發展浪潮之下，人工智慧在財務領域上的應用卻不及在其他領域上的發展和應用。在本文中提出利用人工智慧當中著名的深度學習模型 - 類神經網路模型來探討公司違約風險的預測，透過利用增加隱藏層、以及提出與一般類神經網路模型較為特殊模型架構和自定義的損失函數設定，來克服不平衡資料所帶來的問題，並且作為本文中所主要使用的類神經網路模型。

　　在本研究中所使用的資料是來自於2012年1月至2016年12月的台灣公司資料，而資料中變數可分為以下五個種類，以（1）總體經濟因素（2）負責人基本資料（3）企業特性（4）財務比率以及（5）企業授信資料，在實驗預測結果顯示，無論是在整體預測力和違約金額方面，相較於其他經典模型如羅吉斯模型、隨機森林模型，在本篇論文所提出的類神經網路模型預測力表現最佳，並且在違約金額是最小的，為違約預測之最好的模型。

In recent years, with the increase of computer and hardware equipment and the accumulation of relevant theoretical foundations, artificial intelligence has been glowing in various fields and has been applied in many practical applications, and with the wave of FinTech development, the application of artificial intelligence in the financial field is not as good as the development and application in other fields. In this paper, we use the deep learning model of artificial intelligence model - neural network model to forecast of company credit default risk, through the use of increased hidden layers, and the proposed special model architecture and dynamic loss function setting to overcome the problems caused by imbalance data.

The data used in this thesis are the company data from January 2012 to December 2016, and the variables of the data can be divided into the following five categories : (1) the macroeconomic variables (2) the basic information variables in company’s charger (3) firm’s characteristics variables (4) Financial ratio and (5) firm’s credit information variables. In the final prediction results, the overall predictive power and default amount, compared with other classical models such as the Logistic Regression model, Random Forest model, the neural network model proposed in this paper is the best predictive power and the default amount is the smallest, which is the best model for default prediction.

中文摘要 I
Abstract II
誌謝辭 III
CONTENTS IV
1. Introduction 1
2. Methodology 5
2-1 Architecture of Neural Network Model 5
2-1-1 Basic Architecture of Neural Network Model 5
2-1-2 Architecture of Neuron 5
2-1-3 Special Architectures of Neural Network Model 6
2-1-4 Embedding 6
2-2 Forward Propagation 7
2-2-1 Linear Combination and Non-Linear Transformation 7
2-2-2 Activation Functions 8
2-2-3 Loss Functions 10
2-2-4 Special Loss Function 12
2-3 Backward Propagation 13
2-3-1 Gradient Descent Algorithms 13
2-4 Confusion Matrix and Benchmarks 15
2-4-1 Confusion Matrix 15
2-4-2 Benchmarks 16
3. Data 17
3-1 The Company Data 17
3-2 The Economics Data 17
4. Empirical Result 23
5. Conclusions 25
Reference 26

Altman, E. I. (1968). Financial ratios, discriminant analysis and the prediction of corporate bankruptcy. The journal of finance, 23(4), 589-609.
Bahrammirzaee, A. (2010). A comparative survey of artificial intelligence applications in finance: artificial neural networks, expert system and hybrid intelligent systems. Neural Computing and Applications, 19(8), 1165-1195.
Beaver, W. H. (1966). Financial ratios as predictors of failure. Journal of accounting research, 71-111.
Berkson, J. (1944). Application of the logistic function to bio-assay. Journal of the American Statistical Association, 39(227), 357-365.
Byanjankar, A., Heikkilä, M., & Mezei, J. (2015, December). Predicting credit risk in peer-to-peer lending: A neural network approach. In 2015 IEEE Symposium Series on Computational Intelligence (pp. 719-725). IEEE.
Chiang. (2018). Deep Learning Applied to Default Risk Prediction –Evidence from Lending Club.
Iturriaga, F. J. L., & Sanz, I. P. (2015). Bankruptcy visualization and prediction using neural networks: A study of US commercial banks. Expert Systems with applications, 42(6), 2857-2869.
Jin, Y., & Zhu, Y. (2015, April). A data-driven approach to predict default risk of loan for online Peer-to-Peer (P2P) lending. In 2015 Fifth International Conference on Communication Systems and Network Technologies (pp. 609-613). IEEE.
Odom, M. D., & Sharda, R. (1990, June). A neural network model for bankruptcy prediction. In 1990 IJCNN International Joint Conference on neural networks (pp. 163-168). IEEE.
Ohlson, J. A. (1980). Financial ratios and the probabilistic prediction of bankruptcy. Journal of accounting research, 109 131.
Po-Cheng Huang. (2009). Artificial Neural Network and Discriminate Analysis on the Forecast Model of Corporate Loan Default.
Rafiei, F. M., Manzari, S. M., & Bostanian, S. (2011). Financial health prediction models using artificial neural networks, genetic algorithm and multivariate discriminant analysis: Iranian evidence. Expert Systems with Applications, 38(8), 10210-10217.
Rosenblatt, F. (1958). The perceptron: a probabilistic model for information storage and organization in the brain. Psychological review, 65(6), 386.
Rumelhart, D. E., Hinton, G. E., & Williams, R. J. (1988). Learning representations by back-propagating errors. Cognitive modeling, 5(3), 1.
Tien-Yu Chen. (2009). A Study of the Prediction Models for Overdue Cash Card Loans Using Logistic Regression, Neural Network and Two-Stage Integrated Approach.