Comparison between Nearest Neighbours and Bayesian Network for demand forecasting in supply chain management

Machine Learning has found to be playing a significant role in solving issue of demand forecasting in supply chain management, where many traditional methods result in substandard accuracies. There is a high demand of robust computational systems for predicting the trends of demands for the purpose of Inventory Management in supply chain management of an organization. Every organization has Terabytes of transactions and shipments data. These terabytes of data help in defining and implementing robust techniques that can help in identifying stochastic dependency in the historical data to determine future trends. Attributes like Consignee address, shipper, shipper address, place of delivery, weight of container and country are important for prediction supply trends. Naïve Bayes classifier is used to make decision in uncertainty and K nearest neighbor is lazy and supervised learning algorithm to determine the trends in supply chain. The purpose of this research is to bring a close comparison between Nearest Neighbor Algorithm and Bayesian Networks using confusion matrix as a performance metric and Walmart dataset has been used for simulation. The results show that Bayesian networks technique surpasses the Nearest Neighbor technique in detecting relations in dataset for prediction demand in supply chain. Bayesian networks, emerges to be robust in demand prediction instead of increasing K-neighbors in the supervised learning algorithm.