🐚 Abalone Age Classification Using Neural Networks

This project focuses on building a neural network capable of classifying the age group of abalones into three categories: Young, MiddleAge, and Old.

📊 Problem

The original dataset is highly imbalanced:

• ~2,000 samples each for Young and MiddleAge
• Only ~36 samples for Old

This imbalance posed significant challenges in achieving fair and accurate classification across all classes.

---

🧠 Initial Model (Imbalanced Dataset)

Architecture:

• Input layer
• Hidden layer 1: 16 neurons, ReLU, Dropout 0.1
• Hidden layer 2: 32 neurons, ReLU, Dropout 0.1
• Output layer: Softmax
• Batch Normalization after each layer

Configuration:

• Optimizer: Adam
• Loss: Categorical Cross-Entropy
• Learning Rate: Decaying schedule
• Epochs: 450 (Early stopping at 271)
• Batch size: 32

Despite fine-tuning, this setup struggled due to the class imbalance.

---

🔁 Balanced Dataset: ShuffledDataset.csv

To address the imbalance, I created a modified and shuffled version of the dataset.

Updated Architecture:

• Input layer
• Hidden layer 1: 64 neurons, ReLU, Dropout 0.3
• Hidden layer 2: 32 neurons, ReLU, Dropout 0.2
• Output layer: Softmax
• Batch Normalization included

Configuration:

• Optimizer: Adam
• Loss: Categorical Cross-Entropy
• Learning Rate: Decaying schedule
• Epochs: 100 (with Early Stopping)
• Batch size: 32

---

✅ Final Results

With the balanced dataset and improved architecture, the model achieved:

• Precision: 0.86
• Recall: 0.80
• F1-Score: 0.82
• Loss: 0.04

---

📁 Files

• imbalanced_model.ipynb: First model on original dataset
• balanced_model.ipynb: Improved model using ShuffledDataset.csv
• ShuffledDataset.csv: Modified dataset with better class balance

---