Domain Specific Predictive Analytics: A Case Study With R

As part of our research work Predictive analytics, we are interested to perform experiments on the areas, Supply Chain Risk Management, Credit Scoring and Bankruptcy Prediction. When comparing to previous studies on this topic, our research is novel in the following areas. All the experiments carried out in this paper have used three different application specific data repositories that are described in detail in Design and implementation section. Focused on making use of traditional predictive techniques Linear Discriminant Analysis (LDA), Quadratic Discriminant Analysis (QDA) and compared their performance with respect to Accuracy, Misclassification, Precision, Recall, prevalence and F-Score. we compared the performance of classification algorithms like: Naive Bayes, K Nearest Neighbor with respect to Error in Classification. Analyzed the performance of Model Averaging generation algorithm with respect to average of Markov Blanket size, Neighbourhood size and Branching factor. The main finding from our research is LDA is very good choice when modeling Supply Chain Risk Management, Credit Scoring and Bankruptcy Prediction.