Flight Anomaly Detection using GAN Models

• Developed and applied an ensemble of GAN (Generative Adversarial Network) models MADGAN, FenceGAN, and FGAN attaining industry-leading AUC scores (>70) for anomaly detection in flight data. • Achieved exceptional AUC scores of 97 on benchmark datasets using MADGAN, FenceGAN and FGAN, demonstrating

Facial Emotion Recognition

• Utilized LeNet CNN architecture to create a robust emotion recognition model and achieved a notable test accuracy of 56% on an extended dataset of 35,887 images. • Performed data preprocessing techniques to enhance model performance, including image augmentation and normalization, investigated mod

Colorectal Cancer Classification with CNNs

• Implemented and compared various CNN architectures (LeNet, AlexNet, modified GoogLeNet) for classifying cancer types from histology images. • Achieved 83% accuracy with both EnhancedClassifier and GoogLeNet, significantly surpassing LeNet's 12% accuracy.

Most Streamed Spotify Songs 2024

The objective was to gain insights into music streaming trends, popular artists, and the performance of tracks over the year.Utilized Python libraries including Pandas for data wrangling, Matplotlib and Seaborn for static visualizations, and Plotly for interactive charts.

Credit Card Fraud Detection

Developed and evaluated machine learning models like Logistic Regression, Random Forest, Decision Tree, Gradient Boosting, XGBoost, LightGBM for detecting fraudulent credit card transactions.

German Credit Numeric Project

The goal of this project was to develop and evaluate classification models for predicting creditworthiness based on the German Credit dataset. The focus was on comparing traditional binary classifiers with advanced deep learning approaches to determine the most effective model for this task.

Superstore Analytics Dashboard

The Dashboard is designed to provide an insightful and comprehensive analysis of sales data across various dimensions, including city, region, customer segment, product performance, and time trends.