Research

Research Portfolio

Integrated Optimization of Predictive and Prescriptive Tasks
- Predictive Model (Regression) Integrated under Prescriptive Model (Two-Stage Shipment Linear Optimization Model)
- Data Preparation via Python{NumPy, Pandas}, and Model Solution via Python {Gurobi}.
- Karush–Kuhn–Tucker (KKT) Conditions Implementation.
- Portable Batch System (PBS) High Performance Computing Job Scheduler for Multi-Job Submission with UNIX Cluster.
- Progressive Hedging Algorithm for Decomposition Technique.
- Developed an Algorithm to Optimize Generalization Error.

Prescription-Embedded Predictive Neural Network Models
- Predictive and Prescriptive Two Neural Network Structures Integrated via Custom Loss Function.
- Data Preparation via Python{NumPy, Pandas}, and Model Solution via Python {TensorFlow}.
- SLURM High Performance Computing Job Scheduler for Multi-Job Submission.
- Integrate TensorFlow with ScipyDirect to Find Optimal Loss.

Predicting the Duration of Surgical Operations via Text Mining and Data Modelling

Data Preparation and Cleaning via MATLAB (from scratch)
Text Mining and Feature Engineering in R by Using Different Statistical Analysis.
Neural Network and Random Forest Implementations with Multiple Correspondence Analysis in R.

Other Research Portfolio

An Extensive Implementation of Recently Developed End to End Learning Models in Literature.
Evaluation and Comparison of Different Prescriptive Models via Python {Gurobi,PyTorch, TensorFlow}
Deep Learning, Convolutional and Recurrent Neural Networks, SVM Implementations in MATLAB and R with Well-Known Data Sets.
Implementation of Benders Algorithm for Two-Stage Stochastic Optimization.
Implementation of L-Shaped Method and The Multicut L-Shaped Method for Stochastic Optimization.
Designed a Genetic Algorithm for Evolutionary Data Mining (from scratch).
Designed K-Means, PCA, and AdaBoost Algorithms (from scratch).

Journal Papers

Working Papers & Under Review
- Mehmet Kolcu, Alper Murat, ”Integrated Optimization of Predictive and Prescriptive Tasks”
- Mehmet Kolcu, Alper Murat, ”Prescription-Embedded Predictive Neural Network Models”
- Mehmet Kolcu, ”Predicting the Duration of a Surgical Operation Duration via Text Mining and Data Modelling”

Work in Progress
- Mehmet Kolcu, Alper Murat, ”A survey on Integration of Data and Decision Sciences”
- Mehmet Kolcu, Alper Murat, ”Repairing Decisions through Machine Learning Models”

Conference Presentations

Joint Predictive and Prescriptive Analytics, INFORMS 2019 Annual Meeting, Seattle, WA
Controlling Generalization Error of Joint Predictive and Prescriptive Tasks, 2019 Graduate Research Symposium, Detroit, MI
Integrating Machine Learning Algorithms with Decision Problems, 2018 Graduate Research Symposium, Detroit, MI
Surgery Duration Prediction from Surgery Description, 2017 Departmental Research Presentation, Detroit, MI

Research

Research Portfolio

Integrated Optimization of Predictive and Prescriptive Tasks

Predictive Model (Regression) Integrated under Prescriptive Model (Two-Stage Shipment Linear Optimization Model)

Data Preparation via Python{NumPy, Pandas}, and Model Solution via Python {Gurobi}.

Karush–Kuhn–Tucker (KKT) Conditions Implementation.

Portable Batch System (PBS) High Performance Computing Job Scheduler for Multi-Job Submission with UNIX Cluster.

Progressive Hedging Algorithm for Decomposition Technique.

Developed an Algorithm to Optimize Generalization Error.

Prescription-Embedded Predictive Neural Network Models

Predictive and Prescriptive Two Neural Network Structures Integrated via Custom Loss Function.

Data Preparation via Python{NumPy, Pandas}, and Model Solution via Python {TensorFlow}.

SLURM High Performance Computing Job Scheduler for Multi-Job Submission.

Integrate TensorFlow with ScipyDirect to Find Optimal Loss.

Predicting the Duration of Surgical Operations via Text Mining and Data Modelling

Data Preparation and Cleaning via MATLAB (from scratch)

Text Mining and Feature Engineering in R by Using Different Statistical Analysis.

Neural Network and Random Forest Implementations with Multiple Correspondence Analysis in R.

Other Research Portfolio

An Extensive Implementation of Recently Developed End to End Learning Models in Literature.

Evaluation and Comparison of Different Prescriptive Models via Python {Gurobi,PyTorch, TensorFlow}

Deep Learning, Convolutional and Recurrent Neural Networks, SVM Implementations in MATLAB and R with Well-Known Data Sets.

Implementation of Benders Algorithm for Two-Stage Stochastic Optimization.

Implementation of L-Shaped Method and The Multicut L-Shaped Method for Stochastic Optimization.

Designed a Genetic Algorithm for Evolutionary Data Mining (from scratch).

Designed K-Means, PCA, and AdaBoost Algorithms (from scratch).

Journal Papers

Working Papers & Under Review

Mehmet Kolcu, Alper Murat, ”Integrated Optimization of Predictive and Prescriptive Tasks”

Mehmet Kolcu, Alper Murat, ”Prescription-Embedded Predictive Neural Network Models”

Mehmet Kolcu, ”Predicting the Duration of a Surgical Operation Duration via Text Mining and Data Modelling”

Work in Progress

Mehmet Kolcu, Alper Murat, ”A survey on Integration of Data and Decision Sciences”

Mehmet Kolcu, Alper Murat, ”Repairing Decisions through Machine Learning Models”

Conference Presentations

Joint Predictive and Prescriptive Analytics, INFORMS 2019 Annual Meeting, Seattle, WA

Controlling Generalization Error of Joint Predictive and Prescriptive Tasks, 2019 Graduate Research Symposium, Detroit, MI

Integrating Machine Learning Algorithms with Decision Problems, 2018 Graduate Research Symposium, Detroit, MI

Surgery Duration Prediction from Surgery Description, 2017 Departmental Research Presentation, Detroit, MI