Predictive Analytics and Recommender

Predictive Analytics and Recommender


Predictive Analytics are methods to analyse big data sets of historic data in order to make predictions about the future.

Predictive Analytics are used in a high number of use cases today. Possible applications of predictive analytics are among others:

  • credit scoring (financial services)
  • predict consumer demands and plan supply chain and stock (retail)
  • predicting shopper behavior and do recommendations (retail)
  • customer segmentation, target marketing
  • fraud protection

Two different groups of techniques are commonly used to conduct predictive analytics: regression techniques and machine learning techniques. This wikipedia article gives a good first overview about the applications and techniques of predictive analytics.

Recommender Systems

Recommender systems are used to help user to make particular choices by recommending choices for him. Examples are recommendations of articles or items on a E-Commerce platform or music titles on a music streaming service.

A lot of simple recommenders are based on the approach of Collaborative Filtering (CF), This wikipedia article gives an excellent explanation about functioning and the history of CF.

A more advanced algorithm widely used in the online retail space is Alternating Least Squares Method (ALS). This article from bugro describes CF and ALS very nicely. Additionally, some Python coding on a sample movie database (MovieLens) is included in the article. All these links have been found in a Medium article on using CF and ALS in a kaggle competition on predicting user click behavior.