Parker Nisbet's repositories
xbox-game-pass
Xbox Game Pass subscription value quantification and visualization.
retail-rfm-analysis
Recency, Frequency, and Monetary Value (RFM) analysis and customer segmentation for transactional data. Unsupervised cluster learning to delineate customer types.
animal-attribute-clustering
A quick comparison of two (K-Means and hierarchical/agglomerative) clustering methods for text-based animal species classification.
billboard-artist-frequency
Frequency analysis of historical Billboard Hot 100 lists based on an artist's supporting vs main artist numbers.
dask-emnist-classification
Dask-parallelized project, contrasting GaussianNB and LightGBM models for EMNIST handwritten character classification.
drug-use-and-health
A brief look at drug use trends in the United States, as they relate to: earliest age of consumption, and impact on mental health.
game-ab-testing
A/B testing impact of progression system changes on player retention / interaction. Non-parametric hypothesis testing and power transformations for non-normally distributed data.
mnist-svm-tuning
Optimizing LinearSVC models trained on the MNIST Handwritten Digits dataset, includes ensemble methods and bayesian optimization.
multiclass-perceptron-boundaries
Quick visualization of linear decision boundaries for a scratch-implemented perceptron classifier. Model evaluates loss function with each weight / bias update and will store away best performing parameters for later use.
online-purchase-propensity
In progress. Modeling shopper purchase propensity for a ficticious online retail website.
heart-descent-methods
Partial scratch-implementation of coordinate descent, using a specified feature selector method to choose and later update a coordinate's corresponding weight. This behavior mimics the "fit" function of most descent-compatible machine learning algorithms.
mnist-digit-classifier
Gaussian naive Bayes classifier for digits in the MNIST dataset. Similar in nature to my other repo ("newsgroup-naive-bayes"), albeit instead of multinomial document classification, this repo explores gaussian image classification. Covariance smoothing utilized to minimize error rates to the ~4% realm.
newsgroups-naive-bayes
Multinomial naive Bayes newsgroup document classification without relying on pre-built sklearn modules. Smoothing and inverse document frequencies utilized to improve model accuracy.
nintendo-game-ratings
EDA and association rule mining for a Metacritic-sourced dataset of Nintendo games, with comprehensive game-specific data like review scores, release date, developers, genres, and ESRB rating.
parkernisbet
Config files for my GitHub profile.
tweet-sentiment-nlp
A notebook covering tweet natural language processing, from data munging to predictive model creation and evaluation. Sentiment analysis will be used to classify tweet body text as positive or negative.
vertebral-neighbors-classifier
Comparison of scratch-implemented and sklearn nearest neighbors methods for classification of vertebral patient data. Using multiple distance metrics, models were k-value optimized to reduce error rates.