https://github.com/sanskaryo/cosmic-classifier-iitr

https://github.com/sanskaryo/cosmic-classifier-iitr

Last synced: 2 months ago
JSON representation

• Host: GitHub
• URL: https://github.com/sanskaryo/cosmic-classifier-iitr
• Owner: sanskaryo
• Created: 2025-03-23T03:22:10.000Z (2 months ago)
• Default Branch: main
• Last Pushed: 2025-03-23T03:22:32.000Z (2 months ago)
• Last Synced: 2025-03-23T04:23:48.397Z (2 months ago)
• Size: 2.93 KB
• Stars: 0
• Watchers: 1
• Forks: 0
• Open Issues: 0
• Metadata Files:
  • Readme: README.md

Awesome Lists containing this project

README

        
# Cosmic-classifier-IiTR

🌌 Cosmic Classifier: Planet Classification Model

🚀 Project Overview

A machine learning model to classify planets into 10 different categories based on their physical and atmospheric characteristics. This project was developed for the IIT Roorkee Techfest ML Model Event.

Author: Sanskar Khandelwal

Cogni ID: cogni2048376

📊 Dataset Overview

Input Features (10 attributes)

Atmospheric Density (kg/m³)

Surface Temperature (Kelvin)

Gravity (m/s²)

Water Content (0-100%)

Mineral Abundance (0-1)

Orbital Period (Earth days)

Proximity to Star (AU)

Magnetic Field Strength (Tesla)

Radiation Levels (Sieverts/year)

Atmospheric Composition Index (0-1)

Output Classes (10 German Categories)

Bewohnbar

Terraformierbar

Rohstoffreich

Wissenschaftlich

Gasriese

Wüstenplanet

Eiswelt

Toxischetmosäre

Hohestrahlung

Toterahswelt

🛠️ Technical Approach

1. Data Preprocessing

Handled missing values using KNN imputation

Converted categorical variables (Magnetic Field Strength, Radiation Levels) to numeric values

Normalized features using StandardScaler

Removed rows with missing target values

2. Exploratory Data Analysis (EDA)

Analyzed feature distributions

Created correlation heatmaps

Visualized class imbalance

Studied feature relationships

3. Model Development

Implemented and compared multiple models:

Support Vector Machine (SVM)

K-Nearest Neighbors (KNN)

XGBoost

Random Forest

Logistic Regression

4. Model Optimization

Used GridSearchCV for hyperparameter tuning

Implemented SMOTE for handling class imbalance

Performed cross-validation

Optimized for accuracy and F1-score

5. Final Model Selection

Selected SVM as the best performing model with:

Best parameters: {'C': 10, 'gamma': 'scale', 'kernel': 'rbf'}

Validation Accuracy: 0.9042

F1 Score: 0.9043

Precision: 0.9046

Recall: 0.9042

🎯 Key Achievements

Achieved 90.42% accuracy on validation set

Successfully handled class imbalance

Robust handling of missing values

Efficient feature engineering

🧪 Challenges Faced

Data Quality Issues

Missing values in multiple features

Class imbalance in target variable

Noise in the dataset

Feature Engineering

Converting categorical variables to meaningful numeric values

Handling different scales of features

Model Selection

Choosing between multiple algorithms

Optimizing hyperparameters

Balancing accuracy and computational efficiency

💡 Why This Approach?

Data Preprocessing

KNN imputation preserves relationships between features

StandardScaler ensures fair comparison of features

Model Selection

SVM chosen for its ability to handle non-linear relationships

RBF kernel captures complex patterns in the data

GridSearchCV ensures optimal parameter selection

Class Imbalance Handling

SMOTE creates synthetic samples for minority classes

Helps prevent model bias towards majority classes

📈 Results

The final model achieved impressive results across all metrics:

High accuracy across all classes

Balanced precision and recall

Consistent performance on validation set

🔍 Future Improvements

Feature engineering based on domain knowledge

Ensemble methods combining multiple models

Advanced techniques for handling outliers

Further optimization of hyperparameters

🛠️ Technologies Used

Python 3.x

scikit-learn

pandas

numpy

matplotlib

seaborn

XGBoost

imbalanced-learn

📝 Code Structure

The implementation is organized in the following steps:

Data loading and initial exploration

Preprocessing and feature engineering

Model development and comparison

Hyperparameter tuning

Final model selection and evaluation

Prediction generation

🎉 Conclusion

The developed model successfully classifies planets into their respective categories with high accuracy, demonstrating the effectiveness of the chosen approach in handling complex astronomical data with various challenges.

This project was developed as part of the IIT Roorkee Techfest ML Model Event.