Projects

Spam Email Classifier
| Machine Learning Project

Project aims to figure out the best Spam Email Classifying algorithm within the algorithms used based on misclassification rates and true negatives.
Data set collected at Hewlett-Packard Labs, classifies 4601 e-mails as spam or non-spam.
Optimal Choice : Bagging and Support Vector Machine with linear Kernel - misclassification error rate of 12% (lowest among all models and approaches) and only 2 (lowest among all models and approaches) non spam emails being classified as spam emails.

Panel-data project (PP) exploring exploring 2020 US Presidential Election results at the county level, together with county-level demographic data
Project provides analysis regarding the Presidential Election votes to GOP with explanatory variables like income, education level, etc.
Data obtained using public API from 2019 ACS 5-Year via the tidycensus package in R.
Data wrangling, tidying, joining, manipulating, and feature engineering through R
Interactive U.S Maps using Leaflet Package
Regression Analysis using SLR and MLR.

Findings :

Independently conducted a comprehensive study, including survey design and data collection, to assess economic literacy and decision-making factors among Lawrence University students.
Replicated and updated Bryan Caplan's research, focusing on the unique student population at Lawrence University and shedding light on the impact of economics coursework on students' economic beliefs.
Revealed that taking economics courses significantly impacts students' economic beliefs, making them more likely to hold free-market views and be skeptical of government intervention in the economy.
Investigated demographics, economic knowledge, and influencing factors in economic decision-making, revealing a diverse range of economic literacy levels.
Provided valuable insights into improving economic literacy, informing economic policies, and suggested avenues for future research in the field.

Findings :

The readymade garments (RMG) industry has played a crucial role in the economic development of Bangladesh, significantly impacting its GDP growth.
From 1983, the RMG sector became a major contributor to Bangladesh's exports, comprising about 84% of the total export sector.
The Multi-Fiber Arrangement (MFA) and South Asian Free Trade Area (SAFTA) had notable effects on the RMG sector's growth and economic development.
The period of RMG development (1983-2004) was characterized by steady but low growth, while the post-MFA period (2004-2019) marked a transition to high growth.
Empirical tests confirmed a strong correlation between the growth of total exports, particularly from the RMG sector, and GDP per capita in Bangladesh.
The study suggests that Bangladesh is still experiencing the transitional phase of high growth, contradicting the final stage of constant low growth predicted by the combined Solow-Romer model.
The RMG sector's growth has also positively impacted social and human development indicators such as employment rates, poverty alleviation, life expectancy, female labor force participation rate, as well as literacy rates.

Allows Lawrence University students leverage data and information on the location, past education, and current work experience of LU alumnus resulting in an estimated connection of 3,000 students to alumnus.
Extensive use of R and Power BI Query for data wrangling and feature engineering.

An R Shiny App which helps users visualize the Sales, Battery Life, and Launch Price for each type of model of iPhones since the iPhone 3 GS.
Data gathered through Web Scraping using Selector Gadget.
Project includes Data Taming, Data Tidying, Data Transformation, Data Manipulation, and Joining Data
Project includes advanced ggplot2 graphs
User gets to switch between 2 Types of graphs (Scatterplots connected by segments and a user initiated correlation)
Type 1 graph helps users visualize the Sales, Battery Life, and Launch Price for each type of model of iPhones since iPhone 3g
Type 2 graph shows the correlations between any two among the 4 variables which the user chooses through the Shiny App
Project strictly for implementation of R-Shiny, Web Scraping, Data Manupulation, and GGplot abilities. Plot conclusions may/may not be accurate due to insufficient data.