Oil Spill Classification Project🛢️🌊

Introduction🎯:

This machine learning project focuses on predicting the occurrence of oil spills based on various features provided in the dataset. The dataset consists of a set of masked parameters which influence the likelihood of an oil spill. The goal of this project is to build a machine learning model capable of accurately classifying the presence or absence of oil spills (target feature) using several classification algorithms. The challenge is to handle class imbalance, high-dimensional data, and ensure accurate predictions for both the majority and minority classes.

Problem Statement🚨:

• Objective: Predict the occurrence of oil spills based on masked feature.
• Target Variable: target (indicating the occurrence of an oil spill).
• Features: A combination of numeric and categorical features representing environmental and operational factors.
• Challenge: The dataset suffers from class imbalance, with the majority of the data belonging to the Category 0 (no spill).

Tools & Technologies Used:🛠️

• Programming Language: Python🐍
• Libraries & Frameworks:
  • Pandas: For data manipulation and analysis.
  • NumPy: For numerical computing and array manipulation.
  • Matplotlib & Seaborn: For data visualization (graphs, plots, and statistical analysis).
  • Scikit-learn: For machine learning models, preprocessing (scaling, PCA), model evaluation, and hyperparameter tuning.
  • XGBoost: For gradient boosting and classification.
  • Imbalanced-learn: For handling imbalanced datasets using SMOTETomek.
  • Statsmodels: For calculating the Variance Inflation Factor (VIF) to check for multicollinearity.
  • Pickle: For model serialization and saving/loading models.

ML Model Used:🤖

Various classification models were implemented to predict the target variable:

• Logistic Regression
• Ridge (L2) Logistic Regression
• Lasso (L1) Logistic Regression
• K-Nearest Neighbors (KNN)
• Decision Tree
• Random Forest
• XGBoost

Model Performance 🚀

The following models were tested, and their accuracy scores on the test data are as follows:

Model	Accuracy
Logistic Regression	0.985
Ridge (L2) Logistic Regression	0.987
Lasso (L1) Logistic Regression	0.984
K-Nearest Neighbors (KNN)	0.992
Decision Tree	0.956
Random Forest	0.995
XGBoost	0.995

Conclusion📈

• Best Performing Model: XGBoost achieved the highest accuracy of 0.995, making it the preferred model for this task.
• Model Deployment: The final model was saved using pickle for deployment and future use in making predictions on new data.

How to Use 🛠️

1. Clone the repository.
2. Install the necessary dependencies listed in requirements.txt.
3. Load the model using:

   import pickle
   model = pickle.load(open('Oil_spill_app.pkl', 'rb'))

4. Preprocess your input data (scale it and apply PCA).
5. Predict oil spill occurrences:

   predictions = model.predict(your_input_data)