VANSH AGRAWAL

Performed feature engineering with One Hot Encoding on a 500k+ transactions dataset, enhancing model recall and interpretability

Applied SMOTE sampling and Robust Scaler for normalisation, used z-test for feature selection, ensuring reliable model inputs

Devised a model with XGBoost & ANN, achieving an F1 Score of 0.86, an AUC of 0.92 & improving fraudulent identification by 25%

Achieved an accuracy of 98% at a rate of 40 fps by leveraging AprilTag in conjunction with Deeptag library for a resilient car detector

Applied OpenCV techniques, including colour thresholding, Gaussian blur and convolutions, to extract lane points on road images

Attained 76% accuracy on lane detection by implementing the RANSAC algorithm to fit two lanes among the extracted lane points

Built an ML model leveraging Random Forest and XGBoost algorithms on a large-scale dataset of 120,000 records and 11000+ features

Enhanced dataset quality and robustness by using Standard Scaler for normalisation and KNN Imputer for addressing missing values

Achieved RMSE score of 11.39 by using correlation matrix, PCA, and Exploratory Data Analysis to analyse data points systematically