Upgrade to Pro — share decks privately, control downloads, hide ads and more …

Recommender Engines : A Peak into Predictive An...

Recommender Engines : A Peak into Predictive Analytics

Proposed talk on Predictive Analytics and Recommender Engines

Raghav Bali

June 12, 2016
Tweet

Other Decks in Programming

Transcript

  1. Predictive Analytics • Analysis of current and historical facts/data to

    make predictions about the future • Traditionally a field of statistics/statistical computing. • Now encompasses machine learning and data mining. Current Data Historical Data Predict Future Machine Learning / Statistics
  2. Recommender Engines • Class of Information Filtering systems • Model

    user preferences • Analyse input data to predict output similar to user preferences.
  3. Types of RE • Collaborative Filters • Content Based Filters

    • Hybrid Recommender Engines http://i.imgur.com/xlXjtOL.jpg
  4. RE: Collaborative Filters • Also termed as User Based CF

    • Users with similar behaviours and/or attributes have similar preferences
  5. RE : Content Based • Also termed as Item Based

    CD+F • Item attributes along with user personas are utilized to build preference models
  6. RE : Hybrid • Best of both worlds • Can

    be modelled using User Based CF and Item Based CF in different configurations. • Less prone to issues of sparsity and cold start.
  7. Applications • Jobs you may be interested in • Who

    to follow • Other movies you might enjoy
  8. Issues • Cold Start Problem • Sparsity Problem • Filter

    Bubble http://ebiquity.umbc.edu/blogger/2015/06/08/hot-stuff-at-coldstart/