Overview
Alexander Brown / (119 Words, 1 Minutes)
Overview
This post is a landing page to showcase some of the more traditional (non-spatial) machine learning projects I worked on as part of my M.S. in Spatial Data Science.
To preserve academic integrity, I won’t show any of the code or models without a specific request from someone unaffiliated with USC.
If you are interested in any of my work. Please email me with a specific project you’d like to see/discuss and the reason you’d like to see it. Generally, I won’t share the repository unless we’ve had a prior discussion.
Projects
Non-parametric Classification for Biomedical Anomaly Detection
Using parametric kNN classifiers on biomedical data. Exploring effects of different Ks, polling methods, and distance metrics.
Factory Energy Output Prediction from Sensor Data
Predicting factory energy output from sensor data. EDA, univariate, multivariate, and Knn regression, L1/L2 regularization, non-linear predictors, hypothesis testing.
Time Series Classification of Human Movement Patterns
Classification of human movement patterns from wearable biometric sensors. Time domain feature extraction, binary/multi-class logistic regression models, gaussian/multinomial naive bayes, model analysis (roc/auc, confusion matrix, visual separation), case controlled sampling.
Interpretable Classification & Prediction Models for Sensitive Data
Exploring various classification/prediction methods where interpretability is major concern (i.e. crime prediction, medical diagnoses). Tree based methods (hierarchical decision trees, random forests, L1 penalized gradient boosting tree, pruning), PCR regression, Lasso regression.
Predicting & Classifying Air Truck Brake Failures
Exploring various methods for treating class imbalance on truck air brake sensor data. Over/under sampling, iterative imputation for mising data, balanced random forest classification, XGBoost Classification.
Multi-class/Multi-label Classification of Frog Species from Audio Signals
Exploring multiclass/multilabel classification of frog species based on audio signals. Feature extraction, hamming/exact match loss, gaussian SVM kernels, clustering methods.
Deep Learning Text Sentiment Analysis
Text sentiment analysis (classification) using deep learning (MLPs, Convolutional Networks, LSTMs).