Published: March 2019
The 5 questions machine learning can answer
Data science predicts answers to questions using algorithms.
It's helpful to think about an algorithm as a recipe and your data as the ingredients. An algorithm tells how to combine and mix the data in order to get an answer. Computers are like a blender. They do most of the hard work of the algorithm for you and they do it pretty fast.
It might surprise you, but there are only five questions that data science answers:
Question: Is this A or B?
Algorithm family: Classification
Example: Will this copier fail in the next two months? Yes or no?
Question: Is this weird?
Algorithm family: Anomaly detection
Example: Is this credit card charge normal?
Question: How much - or - How many?
Algorithm family: Regression
Example: What will the Q1 expenses be?
Question: How is this organized?
Algorithm family: Clustering
Example: Which car models have the most brake problems?
Question: What should I do next?
Algorithm family: Reinforcement learning
Example: For a humidity control system: adjust humidity or leave as is?