Home / Summaries / Class notes - Intelligent Data-analyse / learning-values-output

Intro Data Mining, Decision trees

Q: What are three types of learning?

1. Supervised learning ( for prediction, classification): next to the input values, the output values (classes) are also given Example for weather dataset: •Input values: outlook, temperature, humidity, windy •Output values: play? yes / no 2.Unsupervised learning ( for clustering): no output values are available Example for weather dataset: no classes à remove the “play” column •Need to “guess” possible classes, i.e., clusters of data points 3.Reinforcement learning ( for learning a strategy): only an output (e.g., a game output) is generated at the end of game

4 important questions on Intro Data Mining, Decision trees

What is the difference between classification and association?

·Classification rule: Predicts value of a given attribute (the classification of an example)
·Association rule: Predicts value of arbitary attribute (or combination)

What are the 3 ways for biasing the search in Machine Learning?

•language bias, due to limitations in the description language (e.g., format of rule set description, structure of the function)
•search bias, due to the particular search strategy used (no exhaustive search, e.g., heuristic search from general-to-specific till a certain level of fit has been obtained)
•overfitting-avoidance bias, favoring simple models

What are three types of learning?

1. Supervised learning(for prediction, classification):
next to the input values, the output values (classes) are also given
Example for weather dataset:
•Input values: outlook, temperature, humidity, windy
•Output values: play? yes / no
2.Unsupervised learning(for clustering): no output values are available
Example for weather dataset: no classes à remove the “play” column
•Need to “guess” possible classes, i.e., clusters of data points
3.Reinforcement learning(for learning a strategy):
only an output (e.g., a game output) is generated at the end of game