Overview

SMF is an algorithm for learning a matrix factorization model that takes into account seasonal patterns.
SMF has the following properties:

• Online: it provides online training, and scales near-linearly with input size
• Seasonality: it models seasonal patterns, such as daily, weekly, yearly seasonality etc.
• Drift: it allows the latent factors in a matrix factorization model to drift over time.

Paper

• SMF: Drift Aware Matrix Factorization with Seasonal Patterns.
  Bryan Hooi, Kijung Shin, Shenghua Liu, and Christos Faloutsos.
  SIAM International Conference in Data Mining (SDM), 2019.
  

Code

The source codes used in the paper are available. [Download]

Datasets

The Disease dataset used in the paper is from the Tycho dataset, and can be found at https://www.tycho.pitt.edu/data/.
The Taxi dataset is originally from https://www1.nyc.gov/site/tlc/about/tlc-trip-record-data.page.
Preprocessing follows the approach used in: https://github.com/toddwschneider/nyc-taxi-data