Benjamin Soltoff

Benjamin Soltoff is an assistant senior instructional professor in computational social science at the University of Chicago. He’s the associate director of the Masters in Computational Social Science program and teaches courses in research design, programming in R, data visualization, and machine learning. He holds a Ph.D. in political science from Pennsylvania State University. He develops training workshops for learners in academia and industry on data science techniques using R with an emphasis on reproducible workflows, and he’s an RStudio-certified trainer. For more information, you can view his personal site.

projects by Benjamin Soltoff

Machine Learning with R for Text Data

4 weeks · 4-6 hours per week average · INTERMEDIATE

Picture this: You’re an academic researcher tasked with helping social scientists determine the U.S. government’s responsiveness to public demands. A clear expression of this responsiveness is examining the types of policies legislators seek to advance. In this series of liveProjects, you’ll apply machine learning to generate predictions of the policy focus of each congressional bill in a legislation dataset. Leveraging tools widely used by data scientists and academic researchers—including R, the tidymodels framework, feature engineering techniques, and ML algorithms—you’ll perform exploratory data analysis (EDA) to prepare for predictive modeling, preprocess the text data, develop core ML models, and train DL models.

DL for Text Classification

1 week · 4-6 hours per week · INTERMEDIATE

Predict the future! You’re an academic researcher working on a project that predicts what policy areas the U.S. government will prioritize. To achieve your goal, you’ll train three kinds of deep learning neural networks on a legislation dataset (a CSV file containing one row for every bill introduced in the U.S. Congress). With the resulting text classifications, you’ll predict the area of focus for future policy bills.

Extending ML for Text Classification

1 week · 4-6 hours per week · INTERMEDIATE

Imagine you’re an academic researcher working on a project for predicting trends in the U.S. government’s policy-making priorities. Using modern techniques for text data feature engineering, you’ll fit a set of models, subsample the training data to minimize bias, evaluate the models’ performance using a test-set of observations, and leverage a tidy workflow to explain how a model generates specific predictions.

ML for Text Classification

1 week · 4-6 hours per week · INTERMEDIATE

Play the role of an academic researcher preparing a machine learning model to predict the U.S. government’s focus for new policy legislation. You’ll process the legislation dataset with resampling and feature engineering techniques, employ a range of algorithms, including penalized regression and XGBoost, to fit a series of ML models, evaluate the effectiveness of the models, and tune them accordingly.

EDA for Text

1 week · 4-6 hours per week · INTERMEDIATE

Step into the shoes of an academic researcher tasked with predicting which areas will be the focus of the U.S. government’s policy-related efforts. In this liveProject, you’ll prepare for predictive modeling by exploring the policy areas and text descriptions in legislation data, using statistical visualizations and ggplot2, and identifying notable trends and outliers.