Risk Prediction

This paper presents an automatic prediction system using a neural network for risks of an international financial institution-funded project. In test stage, 3-fold cross validation method was applied to the Accountability Counsel and World Bank datasets to evaluate the system performances using seven indicators (Micro and Macro-Averaged Precision, Recall, F1 score, and Hamming loss). Experimental results show that the Hamming loss, micro- and macro- precision, recall, and F1 score of multilabel method were at best 0.06%, 88.10%, 72.57%, 99.96%, 99.71%, 93.45% and 58.34% respectively. This research demonstrated that this neural network can be used for reducing the dimension of feature space and that the proposed model is a promising application as an efficient automatic risk classification system for potential risks of IFI−financed projects.

supervised learning, multilabel, classification, regression, neural network, automatic detection, public sector investment, international finance

Feel free to read the report, download the code, or test the model.