Volcanic Eruption Prediction

This Kaggle competition was launched by Italy's Istituto Nazionale di Geofisica e Vulcanologia (INGV) which focuses on geophysics and volcanology. The institute monitors seismic and volcanic activity all over the country.

This project consists of data exploration and pre-processing to ensure that the data would be fit for themodel we would be using for training and predicting. As mentioned previously, our goal is to fit a modelthat would predict the volcano's "time to eruption" based on the sensor data logs from the different sensorson the volcano.

LoL Esports Data Visualizations

Processed and analyzed League of Legends Esports champion data using Python to show the drafting frequency of each champion over different game updates. Displayed findings using interactive data visualizations.

Web Server and Client

Multithreaded Web Server and Single-threaded Client implementation on Python.

This enables the user to run a multithreaded web server that would be able to handle multiple requests concurrently. The client initiates a connection to the server and would display then message status and page content of the requested file..

Web Crawler and Inverted Index Builder

This project crawls through the web and creates and index for different websites for use in a search engine.

SF Fire Department Calls

Exploratory data visualizations of Fire Department Calls for Services in San Francisco on December 2016

This data set includes information about 911 calls and fire department call services in San Francisco on December 2016. Each of the 911 calls is recorded with an incident number, time, location, and type. The time given includes the date and time of the day. The location includes an address, zip code, and neighborhood district. The call type includes citizen assist, traffic collision, citizen assist, medical incident,vehicle fire, structure fire, outside fire, fuel spill, gas leak, electrical hazard, elevator, alarms, and others ,which are then group labeled as alarm, non life-threatening, potentially life threatening, and fire.

Simulating Simpson's Paradox

This project focused on simulating the Simpson's paradox in a large scale simulated healthcare dataset. The dataset we worked with was created by reverse engineering admission histories and engineering analytical features from simulated data sources on an existing dataset about hospital readmission rates for diabetic patients. The simulated data sources include grocery purchases and wearable vital sign sensors and were used to engineer features such as percent calories from carbohydrates and minutes spent walking. In order to simulate Simpson's paradox, we extended this dataset by adding a genetic trait to act as a statistical confounder.

SF Tree Maintenance Service Requests

TThe datasets used in this visualization are from SF OpenData - geographical data showing the different supervisorial districts in SF and Tree Maintenance Data.

The data was subsetted to focus on service requests which are still "open". The data is visualized using a Non-Proportional Symbol Map.

Java 8 API Visualization

The hierarchical data used for the different visualizations is the Java 8 API hierarchy. The CSV file with the Java 8 hierarchy is from the Java 8 SDK Documentation. Different hierarchical visualization techniques are used to visualize the data.

The different hierarchy visualizations include Traditional Node Layout with Straight Edges, Circular Dendogram Node Layout with Curved Edges, Circle Packing and Treemap

Sentiment Analysis Web Application

Created and deployed a web application using AWS SageMaker that uses a recurrent neural network (PyTorch) to perform sentiment analysis on movie reviews.

NOTE: Web Application currently not live