Projects in Data Science and Statistics

Sentiment Analysis of NYC’s Mayor Eric Adams’ Speeches

Problem-Solving in Action

Web Scraping

Text Analysis

Animation

This project came from a question that I had about how the sentiment in NYC Mayor Eric Adams’ speeches might have changed over the last few years given some of the different…

Forecasting with {modeltime} - Part V

Selecting the best model for each city

Time Series

Forecasting

Machine Learning

This is the final (planned) post for this series on the modeltime (Dancho 2023) package I will identify the best model to forecast the Case-Shiller Home Price Index using an…

Forecasting with {modeltime} - Part IV

Tuning many global models

Time Series

Forecasting

Machine Learning

Up to this point in the project I acquired data in Forecasting with {modeltime} - Part I, performed some standard time series analysis in Forecasting with {modeltime} - Part…

Forecasting with {modeltime} - Part III

Global vs. iterative modeling

Time Series

Forecasting

Machine Learning

In my previous post on this topic – Forecasting with {modeltime} - Part II – I performed a basic analysis of the Case-Shiller HPI time series for four U.S. cities (Dallas…

Forecasting with {modeltime} - Part II

Time series analysis

Time Series

Inferential Statistics

Last time on Forecasting with {modeltime} - Part I I wrote about gathering the data for this project from various sources. Here I’ll go into the first part of the analysis…

Forecasting with {modeltime} - Part I

Time Series

Data Acquisition

Data Wrangling

I learned about the modeltime (Dancho 2023) package – developed by the folks at Business Science – in 2021 and hadn’t had much chance to use it except for a few small coding…

Topic Modeling with R

Text Analysis

Topic Modeling

For this project I’ll be following an example of performing topic modeling with R using the Tidytext format introduced by Julia Silge and David Robinson in their “Text…

Introduction to cepumd

Package Development

Data Wrangling

The purpose of cepumd is to make working with Consumer Expenditure Surveys (CE) Public-Use Microdata (PUMD) easier toward calculating mean, weighted, annual expenditures…

Optimal Threshold vs. Upsampling

Dealing with Unbalanced Data

Machine Learning

Classification

In any binary classification analysis it is common to have to deal with heavily unbalanced data, i.e., the frequency of the majority class in the dependent variable is…

Opening the ML ‘Black Box’ with Shapley Values

A Tidy Approach

Machine Learning

Prediction

Recently I worked on a project which required explanation of ML algorithms more precisely than with precision (I’ve heard you should always open with a bad joke). Being…