README - Metaheuristic Algorithms for Scheduling Problems

Overview This project addresses the Flexible Job Shop Scheduling Problem (FJSP) using various metaheuristic algorithms. It includes implementations for three major algorithms:

• 2SGA (Two-Stage Genetic Algorithm)
• JCGA (Job-Constrained Genetic Algorithm with 2D Encoding)
• TAMA (Three-stage Adaptive Memetic Algorithm)
• Plus a Statistics module for comparing the performance of the algorithms.

Files and Their Purpose:

1. 2SGA.py: Purpose: Implements the Two-Stage Genetic Algorithm (2SGA) to solve FJSP. Key Features:

• Stage 1: Optimizes operation sequence.
• Stage 2: Assigns operations to machines.
• Includes Gantt chart visualization and multiple crossover/mutation strategies. Input: .fjs dataset file (flexible job shop format). Output: Final schedule, makespan, Gantt chart. How to Run:

1. Set the input_path variable to the full path of the .fjs file (e.g., "10a.fjs").

2. Run the file directly: python 2SGA.py

3. JCGA.py: Purpose: Implements the Job-Constrained Genetic Algorithm with 2D Two-Dimensional Encoding (TDE) and Allowed Scheduling Job Set (ASJS) decoding. Key Features:

• Uses precedence constraints.
• 2D chromosome encoding (machine genes + job operation matrix).
• POX crossover and reverse mutation.
• Includes Gantt chart visualization and schedule validation. Input: .fjs dataset file. Output: Optimal schedule with makespan and Gantt chart. How to Run:

1. Ensure you have a .fjs benchmark file (e.g., "10a.fjs").

2. Import and use the main functions: from JCGA import parse_fjs_file, configure_deap jobs, constraints, operations = parse_fjs_file("10a.fjs") toolbox = configure_deap(jobs, constraints, operations)

3. Manually evolve the population and evaluate fitness, or extend with a main loop.

4. TAMA.py: Purpose: Implements the Three-Stage Adaptive Memetic Algorithm using Reinforcement Learning (Q-Learning) for local search control. Key Features:

• Integrates local search operators adaptively.
• Supports action logging, reward-based learning.
• Balances exploration and exploitation. Input: .fjs dataset file. Output: Final makespan, idle time, local search stats. How to Run:

1. Make sure the file path in parse_fjs() matches your .fjs input.

2. Initialize and run: from TAMA import TAMA, parse_fjs jobs, num_machines = parse_fjs("10a.fjs") algo = TAMA(jobs, num_machines) algo.run()

3. Statistics.py: Purpose: Aggregates and analyzes the results of multiple runs of all algorithms. Key Features:

• Computes min, mean, median makespan across benchmarks.
• Generates boxplots, critical difference diagrams.
• Performs Iman-Davenport and Nemenyi statistical tests.
• Exports CSV summaries. Input: JSON result files of the format:
• 2sga_all_runs_XX.json
• jcga_all_runs_XX.json
• final_results_tama_XX.json Output:
• makespan_summary.csv
• avg_ranks_*.csv
• Gantt and statistical visualizations How to Run:

1. Set the correct results_path pointing to your JSON files.
2. Run the script: python Statistics.py

Requirements: Install the following Python packages: pip install deap numpy pandas matplotlib seaborn scikit-posthocs statsmodels networkx

Recommended Run Order:

1. Prepare .fjs dataset files (e.g., "10a.fjs").
2. Run the algorithms individually:
   • 2SGA.py or use its functions
   • JCGA.py as a module
   • TAMA.py via the class method
3. Save each algorithm's run results as JSON (as expected by Statistics.py).
4. Run Statistics.py to evaluate and compare the performance of the algorithms.

Notes:

• Each algorithm file is designed to be standalone but may require minimal adaptation (e.g., file paths).
• All algorithms assume well-formatted FJSP .fjs benchmark files.
• The .fjs benchmark files are located in the "SetDeDate" folder
• You can customize hyperparameters such as population size, crossover/mutation probabilities, etc., within each file.