🔧 Optimization Algorithms

This part contains implementations of optimization algorithms applied to classic computational problems.
The projects demonstrate skills in heuristic search, evolutionary algorithms, and combinatorial optimization using Genetic Algorithms, Nearest Neighbor, and Local Search techniques.

---

📂 Problems Solved

1. Traveling Salesman Problem (TSP) - Berlin52 Dataset

• Problem Type: Combinatorial Optimization
• Dataset: Berlin52 TSP instance (52 cities in Berlin)
• Algorithms Implemented:
  • Nearest Neighbor Heuristic
  • Genetic Algorithm with 2-opt Local Search
• Optimization Focus: Minimize total tour distance for city visitation route.

---

2. Iris Dataset Clustering - Medoid Selection

• Problem Type: Clustering Optimization
• Dataset: Iris flower dataset (150 samples, 4 features, 3 classes)
• Algorithm Implemented: Genetic Algorithm for Medoid Selection
• Optimization Focus: Find optimal medoids to minimize intra-cluster distances.

---

🛠️ Technical Projects & Skills Demonstrated

🔹 Heuristic Search Methods

• Nearest Neighbor: Greedy constructive heuristic for TSP initialization.
• Path Optimization: Sequential city selection based on proximity.

🔹 Evolutionary Algorithms

• Genetic Algorithm Framework: Population initialization, selection, crossover, mutation.
• Fitness Evaluation: Distance-based fitness functions for both TSP and clustering.
• Adaptive Parameters: Dynamic mutation rates based on generation progress.

🔹 Local Search Integration

• 2-opt Optimization: Edge exchange for local improvement in TSP.
• Hybrid Approach: Combine global search (GA) with local refinement.

🔹 Solution Representation & Encoding

• Permutation Encoding: Path representation for TSP solutions.
• Binary Encoding: Medoid selection for clustering.
• Constraint Handling: Ensures valid tours and cluster assignments.

🔹 Performance Optimization

• Distance Matrix Precomputation: Efficient Euclidean distance calculations.
• Elitism Strategy: Preserving best solutions across generations.
• Termination Conditions: Target distance thresholds or generation limits.

🔹 Problem-Specific Adaptations

• TSP Operators: Order crossover, path mutation.
• Clustering Metrics: Intra-cluster distance minimization.
• Medoid Assignment: Optimal cluster center selection.

---

🛠️ Technical Environment

• Languages: Python
• Libraries: NumPy, SciPy, scikit-learn
• Algorithms: Genetic Algorithms, Nearest Neighbor, 2-opt Local Search
• Domains: Combinatorial Optimization, Clustering, Route Planning
• Skills: Algorithm Design, Heuristic Methods, Evolutionary Computation, Performance Tuning

---

📊 Key Achievements

• TSP Optimization: Reduced tour distance from 11,397 (NN) to 7,598 (GA + 2-opt) → 33% improvement.
• Clustering Solution: Automated medoid selection for Iris dataset partitioning.
• Algorithm Hybridization: Demonstrated effectiveness of combining global (GA) and local (2-opt) search methods.

---