CAR Data Science Toolkit

The CAR datas science toolkit is a collection of common data science tools and algorithms, implemented and documented as simply as possible for data journalists to learn from and understand.

Tools currently implemented include:

• Clustering algorithms: DBSCAN; k-means clustering
• Classification: Naive Bayes classifier; k-nearest neighbors
• Similarity metrics: Euclidean distance; Jaccard similarity; cosine similarity; Pearson similarity; Hamming distance
• MapReduce workflow that calculates pairwise document similarity based on TF-IDF weights.