Curious Joe

Blog of Arafath Hossain

Financial Aid App

🎫R 🎫Web App 🎫ML

One of the challenges of allocating non-federal financial aid is to make sure the aid goes to the right candidates: candidates who need the money, and who would actually choose to come if they get the extra money.

This financial aid application helps the aid administrators decide whom to award the aid and how much. It uses a Machine Learning model (Extreme Gradient Boosting) to calculate the probability of enrollment for each admitted student. It allows the users to target a specific segment of students, vary aid allocation amount, and recalculate probability to see the possible effect of aid amount on their enrollment probability.

Currently, this app helps the administrators allocate $200,000 as financial aid among approximately 3,000 students enrolled each year.

Card image cap

Retention Prediction

🎫Python 🎫Machine Learning 🎫Data Analysis

Drop out of undergraduate students after Freshman year is one of the biggest challenges for the universities, especially for the small and medium-sized ones. And limited resources make it impossible to design blanket intervention measures.

Machine Learning models are used to pre-emptively target students for intervention and make the intervention initiatives more efficient and targeted. Total five predictive models are run at different cut-off times to update the retention predictions as the students progress through the semester, and we learn more about the student.


🎫R Package GitHub Repo

fastEda is an R package to make visual exploratory data analysis easier and faster. Using this package users can quickly create multiple plots to visualize uni-variate and bi-variate relations.

Using Clustering to Understand Hollywood Movies

🎫R 🎫Web App 🎫ML 🎫Data Viz Live App GitHub Repo

A Shiny app built as an R package using golem framework. This interactive application helps users to explore the data, and use K-Nearest Neighbors (KNN) to cluster the movies for a deeper understanding of both the data and model.

Text Analysis on Youtube Videos

🎫R 🎫NLP GitHub Repo Blog Post

In this analysis different text analytics techniques such as topic modeling, sentiment analysis were used to analyze the Youtube video titles that contained ‘Bangladesh’, a south Asian country. The goal was to understand which areas were highlighted about Bangladesh, and what kind of sentiments were reflected.

How Worried are Americans?

🎫Tableau 🎫Data Viz Live Dashboard

This Tableau dashboard visualizes a Gallup survey measuring anxiety levels of Americans across all the continental states. This interactive dashboard allows users to have a quick comparative overview of different states in terms of the residents’ anxiety level.

Where are all the Low Income Students?

🎫Tableau 🎫Data Viz Live Dashboard

This Tableau dashboard visualizes data on Pell Grant recipients as shared by the U.S. Department of Education. It shows a comparison among the states in terms of the total number of Pell Grant recipients. It also allows users to see highest Pell recipient receiving institutions, and how their positions change over the years.