This thesis investigates the practical applications of operations research and management science in three distinct fields: non-emergency health services, traditional manufacturing, and online retailing. Despite the prevalent utilization of management science, these sectors continue to face ongoing challenges such as inefficient processes, resource wastage, and subpar customer satisfaction. Our objective is to examine these issues comprehensively and propose potential solutions by developing practically applicable algorithms.

The first chapter centers on non-emergency transportation, drawing inspiration from real-world challenges faced by a non-emergency medical transport service provider in Hong Kong. We investigate a selective dial-a-ride problem with uncertain travel times, striving to optimize service coverage while adhering to specific service quality requirements. A mean-variance risk index is introduced to regulate the violation of service requirements, demonstrating favorable properties in the context of routing optimization. By incorporating selective dial-a-ride optimization with mean-variance risk control, we formulate both individual and collective risk models. To address the resulting nonlinear integer programming formulation, we introduce an efficient branch-and-price-and-cut algorithm. The effectiveness of our proposed models and algorithm is validated through extensive experimentation using a case study. Our risk models offer practitioners flexibility and reliability by including a risk parameter that balances service rate and service quality. Depending on their preferences, practitioners may choose between the individual or collective risk model.

The second chapter presents a novel Joint Order Acceptance and Resource Constraint Scheduling Problem (JOARCSP) model, addressing the complexities of Order Acceptance Scheduling (OAS) and Resource-Constrained Job Scheduling (RCJS) in the context of Industry 4.0. We collaborated with a Chinese manufacturing company, MUF, to develop an OAS system based on their actual production processes. The proposed model employs a Lagrangian relaxation-based optimization algorithm and local search heuristic to demonstrate significant improvements in target value and reduced solution time compared to MUF's original split-rolling model algorithm. Our study highlights the importance of simultaneously considering order selection, resource allocation, and production scheduling, and showcases the effectiveness of heuristic algorithms in solving complex optimization problems. The proposed approach offers valuable insights for enterprises aiming to optimize manufacturing processes within the framework of Industry 4.0, ultimately enhancing production efficiency, customer satisfaction, and profitability. Our contributions include the proposal of a joint OAS and RCJS problem and the development of an efficient and effective algorithm for JOARCSP, providing a reference for similar enterprises facing real-world production challenges.

The third chapter explores the realm of online retailing, a field witnessing increased consumer preference for online shopping. To augment user experience and customer retention, numerous e-commerce firms heavily invest in expediting delivery services. Establishing frontend warehouses in urban areas can often facilitate one-day or half-day delivery services, albeit with limited capacities due to space constraints. This paper introduces a series of stochastic models designed to optimize assortment and inventory decisions while boosting order fulfillment rates for frontend warehouses. To address our models, we propose solutions based on sample average approximation. We evaluate our models using real-world data from Tmall supermarket, and the experimental results unequivocally demonstrate the effectiveness of our methods.