Becky Lytle

Location & Weather Data Analysis

Summer 2017

This past summer, I interned at Bluecore!

I explored customer location data (from discrete ‘events’ on ecommerce sites) using Google BigQuery + visualized data using D3.js (see a cool D3 map I made here!)

Additionally, I developed a heuristic to take customer location data and estimate each customer’s primary location, including developing an approach to leverage k-means clustering to remove outlier cities.

Afterward I wrote production-ready code to process tens of terabytes worth of initial customer event data, keeping track of billions of customers’ locations daily. This at-scale implementation required the use of BigQuery User-Defined Functions written in JavaScript.

To tie all of this data exploration back to weather, I implemented code to create / update weather forecast tables in BigQuery (using a weather API) for 22,000+ unique locations seen in customer events.

Beyond Penn's Treaty

May 2016 - May 2017

Beyond Penn's Treaty is a quaker transcription project housed in a Django site; it aims to make information from various 18th and 19th century quaker journals accessible to the public.

I transformed a prototype of the project into its final product. I also designed a dynamic document viewer using JavaScript and implemented a Django-Haystack based search engine to sift through various manuscripts and thousands of profiles of people, groups, and places.

Algorithm Visualizations

Fall 2016

In my free time, I decided to create a lil side project using my D3.js knowledge paired with my knowledge of different algorithms. So far, I've created 2 "games" that reflect the respective problems that Dijkstra's algorithm and Prim's algorithm solve.

Data Analysis of Lancaster Ave Dataset

Spring 2016

I worked with a dataset called the Lancaster Avenue Project in my Data Science and Visualization class. Throughout the semester, I implemented different data analysis technniques (PageRank and network analysis, k-means clustering, linear regression) using Python in order to draw conclusions about the data.

Here, you can see my visualization and conclusions drawn from applying network analysis to my data. Click here to download my final paper from this project!

Registrar's Project

Fall 2016

In my Analysis of Algorithms class we were asked to design and implement an approximation algorithm for an NP-Complete problem called 'The Registrar's Problem.' Given an input of classes, students (and their preferences for classes), rooms/sizes, professors, and time slots, we created a valid schedule that attempted to maximize total student preferences honored while still maintaining a reasonable time complexity.

Afterward, we came up with ways to alter the problem (and the algorithm) in order to make recommendations to the registrar for future class scheduling.