Ensemble Technique on Predictive Analysis and Fraud Orders Detection using Supervised Machine Learning Algorithms in Supply Chain Management

Abstract

Background: In this research article, the researcher developed a predictive model on fraud orders detection using ensemble approach of supervised machine learning algorithms in supply chain management. Fraud orders are the significant research issues in business industries with respect to supply chain management and logistics management activities it creates a misleading statistic and disrupting the entire business process. The researcher pointed out some of the significant research issues on fraud orders detection in supply chain management.

Method: The researcher used the ensemble techniques on predictive model which are based on different supervised machine learning algorithms. This research article intended to the comparative research study on different supervised machine learning algorithms and its accuracy level such as Logistic Regression 0.69, Random Forest Classifier 0.89, K-Neighbours Classifier 0.74, Gaussian-NB0.67, Decision Tree Classifier 0.88. This predictive model is verified at 89% accuracy level and can be capable to handle imbalance training datasets and predict the sales and orders are in category of fraud or not.

Results: The researcher handled the imbalance datasets with accuracy level of 89% to identify the orders are in category of fraud or not. The researcher used the sales and orders datasets from Kaggle and refined the data with data pre-process process. During the data analysis process the data are passed through the different supervised machine learning algorithms and finally the researcher found that Random Forest Classifier given the 89% accuracy level to classify those orders are in category of fraud or not. One of closer predictive model-based Decision Tree Classifier which is also given the 88% accuracy level and very close to Random Forest Classifier.

Conclusion: Finally, the researcher concluded that the ensemble approach of predictive model is based on Logistic Regression, Random Forest Classifier, K-Neighbours Classifier, Gaussian-NB, Decision Tree Classifier on Fraud Orders Detection Using Supervised Machine Learning Algorithms in Supply Chain Management. This predictive model is verifying at 89% accuracy level to classify whether the orders are in category of fraud or not. The researcher assure that the predictive model would be benefited for the industries in supply chain management and logistics management to identify the sales and orders are fraud or not and enhanced the business process and operational activities.