DescriptionAbundant and real-time data from manufacturing and healthcare industries opens up opportunities previously unimaginable to capture fast changing customer bases, identify new sources of profit, and devise personalized medicine. This thesis consists of three essays: the first essay develops a machine learning algorithm based on a novel pharmacokinetic pharmacodynamic model for inpatients under warfarin therapy to make personalized predictions on blood clotting times on a rolling horizon. The second essay develops a set of methodologies to detect patterns in the online sales data of a children-shoes manufacturer in China to predict price elasticity. The third essay solves a challenging production scheduling problem in a three-stage supply chain. Specifically, the first essay designs a pharmacokinetic pharmacodynamic model to make individualized and adaptive international normalized ratio (INR) predictions for warfarin inpatients in changing clinical status. We tested a new model on 60 inpatients at Columbia Medical Center. The model personalizes four submodels and minimizes the number of parameters to be estimated. Prediction accuracy was assessed by prediction error, absolute prediction error and percentage absolute prediction error. The INRs (International Normalized Ratio, to measure blood clotting times) were accurately predicted 5 days into the future. Median prediction error: 0.01–0.12; median absolute prediction error: 0.17–0.5 and median percentage absolute prediction error: 9.85–26.06%. Patients exhibit interindividual and intertemporal variability. The model captures the variability and provides accurate and personalized INR predictions. In the second essay, we study the problem of pricing and promotion for an online seller of children shoes and try to estimate the price elasticity for different products at different seasons. By testing the various models to capture the seasonality, price elasticity and sales inertia, we may accurately forecast the sales lift for various price discounts, so as to provide a foundation for data-driven price / promotion optimization. In the third essay, we study a three-stage supply chain, where the second stage refers to the operations of external contracted manufacturers. The external manufacturers each has multiple time windows in which the resources needed for the outsourced operation are available. Both internal and external operations are non-instantaneous, and the required processing times are subject to various resource constraints. The problem is to assign and sequence a given set of customer orders to both the internal and the external processes so the total tardiness in the order fulfillment is minimized. We present mathematical models that define different variations of this problem, analyze the special cases that can be solved in polynomial times, and then develop heuristic algorithms that can be applied to quickly solve the problems with a reasonable quality. In particular, we show that a heuristic solution based on the 3D linear assignment algorithm has a potential to be an effective approach for this type of problems.