Olivia Jones

Detail-oriented with strong skills in R, Python, and Machine Learning. Experienced in ensuring data integrity and enhancing research capabilities through effective collaboration and communication.

• Data Visualization
• Machine Learning
• Data Cleaning & Wrangling
• Statistical Analysis
• Mathematical modeling and Optimization
• Dimensionality reduction
• Geometric Modeling
• High-Dimensional Data Analysis
• Interdisciplinary Applications of Mathematics

Erowid Psychedelic Reports Analysis - Python, NLP (Hugging Face Transformers), Selenium, BeautifulSoup, scikit-learn, NLTK, PCA, LSA, Clustering

• Collected and processed 40,000 user-submitted trip reports from Erowid.org using web scraping (Selenium, BeautifulSoup) and stored them in a structured dataset for analysis.
• Conducted sentiment analysis using a Longformer Transformer model to classify experiences into five categories: very negative, negative, neutral, positive, very positive.
• Applied Latent Semantic Analysis (LSA) and K-means clustering to identify semantic themes across 48 psychedelic substances, uncovering key experience patterns such as mystical, auditory, dissociative, and empathy-related effects.
• Performed data cleaning, lemmatization, TF-IDF transformation, PCA, and clustering to extract meaningful insights while removing confounding factors like substance names and administration methods.
• Created semantic similarity matrices and cluster maps to visualize substance relationships

Drum Audio Classifier - Python, scikit-learn, SVM, Decision Trees, Ridge/Lasso Regression, CREPE

• Developed a machine learning pipeline to classify drum audio samples (e.g., kicks, snares, hats, rides) using features like core frequency, amplitude, and average activation extracted from WAV files with the CREPE library
• Engineered a balanced dataset by cleaning raw audio data, applying filtering, and handling class imbalances through under sampling techniques.
• Implemented and compared multiple machine learning models: Decision Tree Classifier (optimized via cross-validation, achieving), Ridge and Lasso Logistic Regression (analyzed feature importance), Support Vector Machines (SVM) with various kernels (linear, RBF, poly, sigmoid), achieving up to 97% accuracy in binary classification tasks.

Machine Learning Model: Titanic – Python, R, scikit-learn, nnet, Random Forest, SVM, Logistic Regression

• Developed and evaluated four machine learning models (Logistic Regression, SVM, Random Forest, Neural Network) to predict Titanic passenger survival, achieving accuracies up to 85%
• Engineered features like FarePerClass and LogFare, applied one-hot encoding, and visualized relationships via heatmaps to enhance model performance and interpretability.
• Addressed class imbalance challenges, optimized hyperparameters (e.g., tree count in Random Forest, hidden layers in Neural Network), and analyzed model performance using confusion matrices, precision-recall tradeoffs, and F1-scores