Page 12 - Data Science Algorithms in a Week
P. 12
Preface ix
Optimization for a Manufacturing Supply Chain.” The methodology uses particle swarm
optimization (PSO) in order to find stability in the supply chain using a system dynamics
model of an actual situation. This is a classical problem where asymptotic stability has
been listed as one of the problems to solve. The authors show there are many factors that
affect supply chain dynamics including: shorter product life cycles, timing of inventory
decisions, and environmental regulations. Supply chains evolve with these changing
dynamics which causes the systems to behave non-linearly. The impacts of these
irregular behaviors can be minimized when the methodology solves an optimization
problem to find a stabilizing policy using PSO (that outperformed GAs in the same task).
To obtain a convergence, a hybrid algorithm must be used. By incorporating a theorem
that allows finding ideal equilibrium levels, enables a broader search to find stabilizing
policies.
• Cutting Forces: Accurate prediction of cutting forces has a significant impact on
quality of product in manufacturing. The chapter “Estimation of Cutting Forces in turning
of Inconel 718 Assisted with High Pressure Coolant using Bio-Inspired Artificial Neural
Networks” aims at utilizing neural networks to predict cutting forces in turning of a
nickel-based alloy Inconel 718 assisted with high pressure coolant. Djordje Cica and
Davorin Kramar discuss a study that employs two bio-inspired algorithms; namely GAs
and PSO, as training methods of neural networks. Further, they compare the results
obtained from the GA-based and PSO-based neural network models with the most
commonly used back propagation-based neural networks for performance.
• Predictive Analytics using Genetic Programming: The chapter “Predictive
Analytics using Genetic Programming” by Luis Rabelo, Edgar Gutierrez, Sayli Bhide,
and Mario Marin focus on predictive analytics using genetic programming (GP). The
authors describe with detail the methodology of GP and demonstrate its advantages. It is
important to highlight the use of the decile table to classify better predictors and guide the
evolutionary process. An actual application to the Reinforced Carbon-Carbon structures
of the NASA Space Shuttle is used. This example demonstrates how GP has the potential
to be a better option than regression/classification trees due to the fact that GP has more
operators which include the ones from regression/classification trees. In addition, GP can
help create synthetic variables to be used as input to other AI paradigms.
• Managing Overcrowding in Healthcare using Fuzzy Logic: The chapter
“Managing Overcrowding in Healthcare using Fuzzy Logic” focuses on the
overcrowding problem frequently observed in the emergency departments (EDs) of
healthcare systems. The hierarchical fuzzy logic approach is utilized by Abdulrahman
Albar, Ahmad Elshennawy, Mohammed Basingab, and Haitham Bahaitham to develop a
framework for quantifying overcrowding. The purpose of this research was to develop a
quantitative measurement tool for evaluating ED crowding which captures healthcare
experts’ opinions and other ED stakeholder’s perspectives. This framework has the