HELLO, I'M
Christina Joslin.
Data Science & Applied Statistics Student
Christina Joslin
About
MY BACKGROUND
With a passion for data science, machine learning, and statistical methods, I have developed a robust academic foundation and practical experience to tackle complex challenges across various industries. As an undergraduate student and Stamps Scholar at Purdue University, I am currently a research assistant in the Sustainable Transportation Systems Research Group. I have also held teaching assistant positions in Purdue's Department of Computer Science
and The Data Mine Corporate Partners program. ​
My research interests lie in applying advanced deep learning techniques, such as convolutional neural networks (CNNs) for computer vision, and recurrent neural networks (RNNs), including Long Short-Term Memory (LSTMs), Gated Recurrent Units (GRUs), and Transformer Networks for time series analysis and natural language processing, utilizing both R and Python. Additionally, I am interested in unsupervised learning, particularly in employing descriptive analysis techniques such as clustering methods.
Research Publications
WHAT I'VE WRITTEN
Used or New? Investigating Consumer Preferences in the Electric Vehicle Market
Bruno Cesar Krause Moras, Christina Joslin, Konstantina Gkritza, Ph.D
This paper was accepted for presentation at the Transportation Research Board (TRB) Annual Meeting 2025
(TRBAM-25-02448), January 5–9, 2025.
​
​​​​
​
Presentations & Awards
WHERE I'VE PRESENTED
Fall 2024 Undergraduate Research Expo
(Poster Presentation)
Purdue University, November 18, 2024
How Does Experience Affect Perceptions of Electric Vehicle Characteristics? A Users vs. Non-Users Comparison
Christina Joslin, Bruno Cesar Krause Moras, Konstantina Gkritza, Ph.D
​
NEXT-GEN Transport Systems Conference
(Poster Presentation)
Purdue University, September 21-22, 2024
Best Poster Presentation Award
Does Experience Shape Perceptions? A Comparison of Users' and Non-Users' Views on Electric Vehicle Characteristics
Christina Joslin, Bruno Cesar Krause Moras, Konstantina Gkritza, Ph.D
The Data Mine Corporate Partners Symposium
(Poster Presentation)
Purdue University, April 24, 2024
Caterpillar Inc. Electric Vehicle Charging Operations
Christina Joslin - Team Leader/Teaching Assistant for the CAT EV Charging Operations Team consisting of 12 graduate/undergraduate students.
​Provisional Patent Submitted April, 2024
Methods And Systems For Charging Electric Machines with On-Site Mobile Charging Stations
Spring 2024 Undergraduate Research Symposium
(Poster Presentation)
Purdue University, April 9, 2024
Public Adoption of New versus Pre-owned Electric Vehicles
Christina Joslin, Bruno Cesar Krause Moras, Konstantina Gkritza, Ph.D
The Data Mine Corporate Partners Symposium
(Poster Presentation)
Purdue University, April 24, 2023
BASF Forecasting Analogue Years for Corn & Soybeans
Christina Joslin - Student member of BASF Corn & Soybean Team. ​
Portfolio
WHAT I’VE CREATED
Dog and Cat Breed Classifier with MobileNetV2 Transfer Learning
​​​Description: This project implements a Convolutional Neural Network (CNN) in Python to classify images from 104 dog and cat breeds, utilizing a dataset of 29,000 images. The MobileNetV2 architecture, known for its efficiency and scalability, was used with transfer learning to achieve an F1 score of 84% on the test set. Key components include data preprocessing, model training with early stopping, and comprehensive evaluation metrics, making the model suitable for large-scale image classification tasks.​​​​​​​​
Disaster Tweet Classifier: Fine-Tuning DistilBERT for NLP
​​​
Description: This project builds and fine-tunes a Natural Language Processing (NLP) model in Python to classify disaster-related tweets using Hugging Face's Transformers library. It covers text tokenization, model fine-tuning with a pre-trained DistilBERT, performance evaluation, and achieving an F1 score of 83% on the test set.​​​​​​​​
Amazon Review Classifier: Fine-Tuning Multilingual BERT
Description: This project develops an NLP pipeline for multilingual sentiment analysis using Python and the pre-trained BERT model nlptown/bert-base-multilingual-uncased-sentiment. It includes data preprocessing, fine-tuning with Hugging Face's Trainer, and evaluation with metrics such as accuracy and F1-score. The fine-tuned model classifies text sentiment (negative, neutral, positive), achieving 77% accuracy and 76% F1 on the test set using a stratified subset of 84,000 examples.
CNN Digit Recognizer
​Description: This project implements a Convolutional Neural Network (CNN) in Python to recognize handwritten digits from the MNIST dataset. It includes data preprocessing, model training, evaluation, and saving the trained model. The GitHub version achieves a test accuracy of 99.0%, while the Kaggle version attains approximately 98.8% accuracy.
Note. In the GitHub repository, the entire MNIST dataset is loaded directly from TensorFlow, resulting in differences in data preprocessing between the Kaggle Notebook and the GitHub repository.
​​​​​​​​​
Titanic Survival Prediction
with XGBoost Classifier
​Description: This project develops a comprehensive machine learning pipeline in Python to predict Titanic passenger survival using the XGBoost classifier. The process encompasses data preprocessing—employing standard scaling, one-hot encoding, and K-Nearest Neighbors (KNN) imputation—followed by hyperparameter optimization via RandomizedSearchCV. The model's performance is thoroughly evaluated and achieved a test accuracy of 78% with final predictions exported to a CSV file.​​​