In this paper, both traditional and new-born machine learning based techniques are reviewed, and some cases from previous papers are analyzed. Through this paper, it is found that machine learning based methodology, compared to the conventional ones, indeed provides better performance in increasing the accuracy of demand forecasting and reducing the cost of daily optimization and operation cost. This research is aimed to give a general review of techniques, and find out whether it is possible for machine learning based techniques can actually help companies to manage the supply chain with higher efficiency.