Saumya Parikh

Graduate Student at Arizona State University

QuizIt: An Adaptive Quizzing Platform

I designed an online quizzing application that provides questions to the users based on their preference and proficiency. The application measures and keeps track of the user performance, based on a set of parameters such as the accuracy of their answers, the time taken, the average performance for each question, etc. The application was developed in a team of 4 using MongoDB, React, Express and Node JS. My focus was on designing the algorithm and the flow of the application, and later helping out in improving user interface and building visualizations.

Visualizing Climate Change

As part of a course on Data Visualizations (CSE 578) at ASU, I worked in a team of 4 students to develop detailed and comprehensive visualizations on climate change. We used data from the World Bank open data catalog, National Snow and Ice Data Center and Berkeley Earth. I worked on the two visualizations for Ice Extent, and the animated world map shown on the Combined page, and helped my teammates with the Parallel Coordinated Plot, again on the Combined page. All the visualizations were created on D3.js.

Analyzing Advertisements Dataset

Created a logistic regression model to predict whether a user will click on an ad or not. For this, I used a dummy dataset consisting of various features, such as user's age, daily time spent on site, city, gender, etc. I divided the entire dataset into training and testing sets by 70:30 ratio. The model could classify the test data with an average of 92% precision and 92% recall.

Working on Iris Dataset with SVM and GridSearch

We were introduced to the popular multivariate dataset during the Data Mining class at ASU. Since it was readily available in Seaborn, I tried analyzing and creating an SVM model for classifying the data. While the default parameters (kernel coefficient and penalty parameter) were good enough to classify the data, I used the GridSearch technique anyway to learn its basic syntax in Scikit-learn. The test dataset was classified with an average of 96% precision and 96% recall.

Learning TensorFlow and Deep Learning with MNIST Dataset

Got my feet wet in the field of Deep Learning by trying out the MNIST Digit Recognition problem, aka "the Hello World problem for Deep Learning". I followed the tutorial provided by Google step-by-step to understand the finesse and power of TensorFlow and get a hang of the basic process of pattern recognition with deep learning. Now that I have understood the overall application of neural networks and the deep learning method, I'll try to get a clear understanding of the theory behind it and then work with more problems.

Online Drum Machine

I am trying to create an online drum machine that would allow users to play a beat pattern, enable them to select samples, adjust the tempo, pitch, and visualize the sound. It's a work in progress. I am learning the various functionalities that the Web Audio APIs and Nexus UI provide and trying to add them in my project. Music is my greatest passion, so working on this has been really fun and stress-busting.