•We build a stochastic programming model for supply chain planning under uncertainty.•The model handles multi-product, multi-echelon supply chains with backorders.•Our supply uncertainty combines ...supplier random yield and stochastic lead times.•We apply our model to a large real-world wind turbines supply chain.•We show theoretical and numerical results for the impact of supplier uncertainty.
With today's rapidly changing global market place, it is essential to explicitly include uncertainty in supply chain planning models. In this paper, we develop a two-stage stochastic programming model for the comprehensive tactical planning of supply chains under demand and supply uncertainty. The model handles multi-period, multi-product, and multi-echelon supply chains. It incorporates backorder penalties with general cost structures. The considered supply uncertainty combines supplier random yield and stochastic lead times, and is thus the most general form of such uncertainty to date. We illustrate how our model directly applies to the planning of the supply chain of one of the world's biggest manufacturers of wind turbines. We illustrate theoretical and numerical results that show the impact of supplier uncertainty/unreliability on optimal procurement decisions. We also quantify the value of modeling uncertainty versus deterministic planning.
•We model a new quantity discounts supply chain planning problem.•The model is very hard to solve using leading commercial solvers.•We develop MIP-based local search algorithms for our problem.•We ...apply our model to a realistic food supply chain.•We show the efficiency of our algorithms in getting high quality solutions quickly.
Supplier selection with quantity discounts has been an active research problem in the literature. In this paper, we focus on a new real-world quantity discounts scheme, where suppliers are selected in the beginning of a strategic planning period (e.g., 5 years). Monthly orders are placed from the selected suppliers, but the quantity discounts are based on the aggregated annual order quantities. We incorporate this type of cost structure in a multi-period, multi-product, multi-echelon supply chain planning problem, and develop a mixed integer linear programming (MIP) model for it. Our model is highly intractable; leading commercial solvers cannot construct high quality feasible solutions for realistic instances even after multiple hours of solution time. We develop an algorithm that constructs an initial feasible solution and a large neighborhood search method that combines two customized iterative algorithms based on MIP-based local search and improves such solution. We report numerical results for a food supply chain application and show the efficiency of using our methodology in getting very high quality primal solutions quickly.
Machine learning (ML) models have been quite successful in predicting outcomes in many applications. However, in some cases, domain experts might have a judgment about the expected outcome that might ...conflict with the prediction of ML models. One main reason for this is that the training data might not be totally representative of the population. In this paper, we present a novel framework that aims at leveraging experts’ judgment to mitigate the conflict. The underlying idea behind our framework is that we first determine, using a generative adversarial network, the degree of representation of an unlabeled data point in the training data. Then, based on such degree, we correct the machine learning model’s prediction by incorporating the experts’ judgment into it, where the higher that aforementioned degree of representation, the less the weight we put on the expert intuition that we add to our corrected output, and vice-versa. We perform multiple numerical experiments on synthetic data as well as two real-world case studies (one from the IT services industry and the other from the financial industry). All results show the effectiveness of our framework; it yields much higher closeness to the experts’ judgment with minimal sacrifice in the prediction accuracy, when compared to multiple baseline methods. We also develop a new evaluation metric that combines prediction accuracy with the closeness to experts’ judgment. Our framework yields statistically significant results when evaluated on that metric.
•The outputs of machine learning models often conflict with experts’ judgment.•We resolve the conflict by incorporating the judgment into the existing model.•A new metric is proposed to assess the quality of prediction models.•We run experiments on synthetic data and real-world case studies on IT and Finance.•Our framework produces better results than the existing ML baselines.
•We consider a revenue change prediction problem.•We propose a framework that casts the problem as a classification one.•Our method maximizes prediction precision while minimizing sacrifice in ...accuracy.•Our method treats class imbalance that is typical in datasets of this problem.•We validate our method on real-world datasets and compare it with prior art methods.
In business environments where an organization offers contract-based periodic services to its clients, one crucial task is to predict changes in revenues generated through different clients or specific service offerings from one time epoch to another. This is commonly known as the revenue change prediction problem. In practical real-world environments, the importance of having adequate revenue change prediction capability primarily stems from scarcity of resources (in particular, sales team personnel or technical consultants) that are needed to respond to different revenue change scenarios including predicted revenue growth or shrinkage. It becomes important to make actionable decisions; that is, decisions related to prioritizing clients or service offerings to which these scarce resources are to be allocated. The contribution of the current work is twofold. First, we propose a framework for conducting revenue change prediction through casting it as a classification problem. Second, since datasets associated with revenue change prediction are typically imbalanced, we develop a new methodology for solving the classification problem such that we achieve maximum prediction precision while minimizing sacrifice in prediction accuracy. We validate our proposed framework through real-world datasets acquired from a major global provider of cloud computing services, and benchmark its performance against standard classifiers from previous works in the literature.
Applications of cloud computing are increasing as companies shift from on-premise IT environments to public, private, or hybrid clouds. Consequently, cloud providers use capacity planning to maintain ...the capacity of computing resources (instances) required to meet the dynamic nature of the demand (queries). However, there is a trade-off between deploying too many costly instances, and deploying too few instances and paying penalties for not being able to process queries on-time. An instance has multiple resource dimensions and executing a query consumes multiple dimensions of an instance's capacity. This detailed multi-dimensional management of cloud computing resource capacity is known as elasticity management and is an important issue faced by all cloud providers. Determining the optimal number of instances needed in a given planning horizon is challenging, due to the combinatorial nature of the optimization problem involved. We develop an optimization model and related algorithms to capture the trade-off between the resource cost versus the delayed execution penalty in software as a service applications from the cloud provider's perspective. We develop an exact approach to solve small to medium sized applications and heuristics to solve large applications. We then evaluate their performance via extensive computational analyses with real-world data and current cloud provider approaches. We also develop a stochastic framework and methodology to deal with demand uncertainty, and using two different randomly generated data sets (representing problem instances in practice), we demonstrate that robust solutions can be obtained.
We consider the problem of portfolio selection with risk factors, where the goal is to select the portfolio position that minimizes the value at risk (VaR) of the expected portfolio loss. The problem ...is computationally challenging due to the nested structure caused by the risk measure VaR of the conditional expectation, along with the optimization over a discrete and finite solution space. We develop a nested simulation optimization approach to solve this problem. In the outer layer, we adapt the optimal computing budget allocation (OCBA) approach to sequentially allocate the simulation budget of the outer-layer to different portfolio positions. In the inner layer, we propose a new sequential procedure to efficiently estimate the VaR of the expected loss. We present a numerical example that shows that our approach achieves a higher probability of correct selection under the same computing budget compared to three other methods.
A Cloud-Migration Feasibility Advisor Asthana, Shubhi; Megahed, Aly; Iyoob, Ilyas
2020 IEEE 40th International Conference on Distributed Computing Systems (ICDCS),
2020-Nov.
Conference Proceeding
Many organizations, driven by the need for greater productivity and lower costs, are moving their portfolio of applications to the cloud. However, there are many challenges in migrating an ...application to the cloud. That is, there are different aspects of an application that affect its cloud-migration feasibility. That is why assessing the feasibility of migrating an application to the cloud is not an easy task. Also, it is non-trivial to map which cloud solution - whether infrastructure as a service, platform as a service, or software as a service, should be used for hosting the application, if it is shown to be cloud-migration feasible. In this paper, we present a novel cloud-migration feasibility advisor that analyzes the meta data received about an application and evaluates whether it is feasible to move it to the cloud or not. In addition, the advisor uses dependency graphs and the aforementioned meta data to classify which cloud solution would suit best for the application, if it is deemed migration feasible.
The economics of the cloud model has been encouraging IT enterprises to migrate from on-premise environments to public, private, or hybrid cloud solutions. To perform such a migration, a cloud ...offering needs to be chosen and a cloud solution needs to be built. In industrial settings, cloud designers may spend days or even weeks to come up with an acceptable cloud solution at a low cost/price. Like any manual process, it is obvious that such a cloud solution design process is error prone, time consuming, and does not guarantee an optimal output, e.g. a solution with a minimum cost/price. Different from existing works that solve the problem from the user’s angle, we solve it from the cloud provider’s prospective, who aims at offering customized cloud solutions for different user requirements at low costs. Such difference requires a unique way of problem modeling. Through analyzing real business data, we abstract the problem into a general attribute–value combinations and formulate a powerful integer programming optimization model to solve it. The general form of the optimization model allows various definitions of customer requirements as well as cloud offerings. Our novel optimization approach for cloud solution design satisfies client requirements, cloud offering constraints, and produces a solution at a minimum cost in a short time, if one exists. We evaluated our solution on realistic data against two baseline approaches. The numerical results show both the effectiveness and efficiency of our approach as well as its practical potential.
•We tackle the problem of cloud solution design.•We formulate an optimization model that finds lowest cost solution designs.•We compare our approach to two baseline heuristics as well as a brute force method.•Numerical results show the efficiency and effectiveness of our approach.