Overview
Current Completed
Computer Science
iesepulveda.com
This is the website you're currently on! All the HTML/CSS was either coded by myself, or modified from an open-source project credited in this site's github repository.I am currently using Github Pages to host the static front end. After purchasing my domain from Google Domains, I configured ivan-sepulveda.github.io to forward to iesepulveda.com while iesepulveda.com pulls the HTML & CSS from the ivan-sepulveda.github.io repository linked below. This summer, I aim to use either Digital Ocean or Nearly Free Speech to host a back-end that can run my Python, Java, and R scripts.
Economics, Business, & Finance
Determining Feature Importance in Securities
Spring 2019
Using various classification methods, I sought to determine the most relevant variables in a stock’s performance; securities were evaluated in a binary manner to reflect if they had surpassed the Standard and Poor 500’s performance for the equivalent time period.
Retail-Fashion Stock Correlation
Fall 2018
Social media is an increasingly significant aspect of our daily lives; millennials and generation-X especially. As these generations becoming an increasing economic force, it begs the question, does social media play a role in our economy? I seek to find a correlation between the social media presence of clothing retailers and their stock index.
German Credit
Spring 2019
Credit Risk Model: Predictive Models have been used or are being used across industries and functional areas. Credit Risk Scorecard is one of the most successful and probably one of the early adopters of predictive modeling in a corporate set up. A scorecard is an objective, repeatable and measurable way to assess the risk of credit (personal loan, mortgage or credit card) applicants. In this scenario, customers' demographic and behavioral data is available and the aim is to predict the probability of default. The target variable is "class" which takes value as "Good" and "Bad". The aim is to build a credit risk scorecard (just a set of probability values) based on the sample data provided.
.jpg)
Online Advertisements
Spring 2019
I performed classification using Logistic Regression and KNN. The goal of the project was to Predict who is likely going to click on the Ad on a website based on the features of a user. Some features included Daily Time Spent on a Site, Age, Area Income and Country. My logistic regression based model was able to achieve 97.1% accuracy.
Data Visualization
Population Displacement due to Typhoon Yolanda
Spring 2019
Jaime Caballero and I wrote a script to plot the magnitude of families displaced by Typhoon Yolanda based on province. The code is written in R and uses ggplot along with other APIs.
Natural Language Processing
Analysis of Restaurant Reviews
Spring 2019
Straightforward R-script to stem reviews and find which word stems correlated with whether or not a restaurant received a like.