Leveraged Nearest Neighbors algorithm for efficient TSP computation.

Integrated Nominatim API for automated city coordinate retrieval.

Employed GIF-based animated tour generation for interactive route visualization.

Implemented SARSA algorithm for training an agent in a gaming environment.

Conducted experiments with various hyperparameters to analyze their impact on agent performance.

Evaluated the effectiveness of different hyperparameter configurations to optimize agent learning in the gaming environment

Developed a deep learning model using PyTorch for object localization, enabling precise identification and localization of objects within images.

Implemented advanced techniques such as data augmentation, transfer learning with pre-trained models, and custom dataset creation to enhance model performance and accuracy.

Successfully trained and evaluated the model, achieving significant improvements in object localization accuracy compared to traditional computer vision techniques.

Developed a deep learning-based road segmentation model using UNet architecture and EfficientNet-B0 encoder.

Applied extensive data augmentation techniques to improve model performance and generalization.

Achieved high accuracy in road segmentation on a large-scale dataset, demonstrating the potential for real-world road detection applications.

Built a cutting-edge web application on NYC Collision Analysis using Python, Streamlit, PyDeck, GCP, BigQuery, SQL, and Looker Studio.

Extracted and processed extensive collision data from NYC Open Data, storing it in GCP BigQuery for advanced analysis.

Created interactive visualizations and a user-friendly interface to explore collision data, with filtering and mapping capabilities.<

Trained Support Vector Machines (SVMs) with linear, radial, and polynomial kernels on the OJ dataset.

Evaluated training and test error rates for each SVM model and identified optimal cost values.

Utilized machine learning algorithms, including Neural Network, CART, and Random Forest, to analyze the Cleveland heart disease dataset.

Compared the performance of the models based on metrics such as mean squared error, accuracy, and error rate.

Found that Random Forest achieved the highest accuracy (83.14607%), followed by the Neural Network (77.72006%), and CART model.

Created a binary variable, mpg01, based on car mileage to facilitate prediction using the Auto dataset.

Explored the association between mpg01 and other features through scatterplots, boxplots, and correlation analysis.

Applied different classification models such as LDA, QDA, Logistic Regression, and KNN to predict mpg01 and calculated their respective test error rates.

Loaded the required libraries for the assignment, including ISLR2, ISLR, bootstrap, boot, leaps, klaR, class, GGally, corrplot, caret, and rpart.

Performed best subset regression analysis on the Boston data from the ISLR2 package, using AIC, BIC, five-fold cross-validation, and ten-fold cross-validation for model selection.

Observed different results for the selected models using AIC, BIC, five-fold cross-validation, and ten-fold cross-validation.

Utilized Google Cloud Platform's BigQuery to analyze Chicago crime data.

Explored crime patterns and trends in Chicago using various queries and data analysis techniques.

Gained insights into the types of crimes, their frequency, geographical distribution, and temporal variations to inform decision-making and resource allocation for crime prevention efforts.

Performed preprocessing and feature engineering on genome data by removing irrelevant features, normalizing the data, and selecting relevant genetic markers.

Trained a random forest model to detect autism, leveraging its ability to handle complex datasets like genome data.

Explored and compared the performance of boosting algorithms (GBM, Adaboost, XGBoost) to improve the random forest model's baseline performance.