🌊 Sonar Object Classification (Mine vs Rock)

An intelligent Machine Learning project that classifies sonar signals as Mine or Rock using Random Forest — with a clean and interactive Streamlit web app for real-time predictions.

🧠 Overview

Underwater sonar signals are often used to detect and differentiate between rocks and metallic objects such as naval mines.
This project analyzes sonar returns from 60 frequency bands and predicts whether the object detected is a Mine (M) or a Rock (R) using supervised learning.

The goal is to create an accurate, interpretable, and deployable model for Sonar Object Classification.

📂 Dataset

Source: UCI Machine Learning Repository

Instances: 208

Features: 60 continuous numeric values (energy values at different frequencies)

Target Labels:

M → Mine

R → Rock

⚙️ Tech Stack
Category	Tools & Libraries
Language	Python 🐍
Data Processing	Pandas, NumPy
Modeling	Scikit-learn (RandomForestClassifier)
Visualization	Matplotlib, Seaborn
Deployment	Streamlit
Model Storage	Joblib
🚀 Project Workflow
1️⃣ Data Preprocessing

Loaded sonar dataset and explored data structure.

Encoded target labels (M and R) to numeric values.

Standardized the 60 features for uniform scaling.

Split dataset into 80% training and 20% testing.

2️⃣ Model Training

Trained multiple models and compared performance.
The Random Forest Classifier performed best with:

Excellent generalization

Low variance

Robust to noise and overfitting

from sklearn.ensemble import RandomForestClassifier

model = RandomForestClassifier(n_estimators=100, random_state=42)
model.fit(X_train, y_train)

3️⃣ Model Evaluation

Evaluated accuracy, precision, recall, and confusion matrix.

Model	Accuracy
Logistic Regression	83.2%
K-Nearest Neighbors	86.4%
SVM	88.6%
Random Forest	91.3% ✅
💾 Model Saving

Saved the trained Random Forest model using Joblib:

import joblib
joblib.dump(model, 'sonar_model.pkl')


Load later for prediction:

model = joblib.load('sonar_model.pkl')

🖥️ Streamlit Web App
▶ Run the app locally
streamlit run streamlit_app.py

📋 App Features

✅ User-friendly input form (manual or file upload)
✅ Predicts instantly: Mine or Rock
✅ Displays confidence probability
✅ Visual feedback with color-coded results
✅ Model accuracy displayed on sidebar

🧩 Sample Prediction Code
import numpy as np
import joblib

model = joblib.load('sonar_model.pkl')

# Example input (60 feature values)
sample_input = np.array([[0.02, 0.03, 0.05, ... , 0.09]])
prediction = model.predict(sample_input)

if prediction[0] == 1:
    print("Mine Detected 💣")
else:
    print("Rock Detected 🪨")


📚 References

UCI Sonar Dataset

Scikit-learn Documentation

Streamlit Docs

✨ Future Enhancements

Integrate Deep Learning (ANN/CNN) models.

Add hyperparameter tuning with GridSearchCV.

Include feature importance visualization.

Deploy app publicly via Streamlit Cloud / Render / AWS.

👨‍💻 Author

Lidiya
💼 https://www.linkedin.com/in/d-lidiya-68388a331/
 | 🧠 https://github.com/alwaysalearner1234
 | ✉️ [email protected]