JADS Premaster - Introduction to Machine Learning Summary

1.‌‌Introduction‌ ‌
Machine‌‌learning‌ ‌
Giving‌‌computers‌‌the‌‌ability‌‌to‌‌learn‌‌without‌‌explicitly‌‌programming‌‌them.‌ ‌
‌




‌
‌
Supervised‌‌learning‌ ‌
When‌‌a‌‌computer‌‌learns‌‌using‌l‌abeled‌‌‌data.‌‌The‌‌input‌‌is‌‌features‌‌and‌‌the‌‌output‌‌is‌‌the‌‌
predicted‌‌label.‌ ‌
● Classification‌ ‌
● Regression‌
‌
Unsupervised‌‌learning‌ ‌
When‌‌a‌‌computer‌‌actually‌‌learns‌‌using‌u‌ nlabeled‌‌‌data.‌‌The‌‌input‌‌is‌‌classified‌‌into‌‌one‌‌or‌‌more‌‌
categories.‌ ‌
● Clustering:‌‌input‌‌divided‌‌into‌‌groups‌ ‌
● Dimensionality‌‌reduction:‌‌visualize‌‌data‌‌with‌‌too‌‌many‌ ‌
● Anomaly‌‌&‌‌novelty‌‌detection‌ ‌
● Association‌‌rule‌‌learning‌ ‌
‌
CRISP-DM‌ ‌




‌ ‌
‌




1‌ ‌

, Data‌‌mining‌ ‌
1. Define‌‌Goals:‌‌business‌‌and‌‌data‌‌mining‌‌experts‌‌have‌‌to‌‌define‌‌goals‌‌and‌‌measures‌‌of‌‌
success‌ ‌
2. Obtain‌‌Model:‌‌pre-process‌‌the‌‌data‌‌and‌‌apply‌‌data‌‌mining‌‌algorithms‌ ‌
3. Evaluate‌‌Results:‌‌use‌‌pre-specified‌‌measures‌‌to‌‌evaluate‌‌the‌‌models‌ ‌
4. Deploy:‌‌deploy‌‌the‌‌model‌‌if‌‌the‌‌evaluation‌‌is‌‌successful‌ ‌
‌




‌
‌
Dependent‌‌variable‌ ‌
A‌‌variable‌‌that‌d
‌ epends‌‌‌on‌‌other‌‌variables.‌ ‌
‌
Independent‌‌variable‌ ‌
A‌‌variable‌‌that‌d
‌ oes‌‌not‌‌depend‌‌‌on‌‌other‌‌variables.‌ ‌
‌
Confusion‌‌matrix‌ ‌
‌ PREDICTION‌ ‌

A‌ ‌ True‌ ‌ False‌ ‌
C‌
T‌
True‌ ‌ True‌‌Positive‌ ‌ False‌‌Negative‌ ‌
U‌
A‌
L‌ ‌ False‌ ‌ False‌‌Positive‌ ‌ True‌‌Negative‌ ‌
‌
Precision‌ ‌
The‌‌fraction‌‌of‌‌relevant‌‌instances‌‌among‌‌the‌‌retrieved‌‌instances.‌ ‌
‌
true positives
precision = true positives + f alse positives ‌
‌
Recall‌ ‌
The‌‌ability‌‌of‌‌a‌‌classifier‌‌to‌‌find‌‌all‌‌positive‌‌instances.‌ ‌
‌
true positives
recall = true positives + f alse negatives ‌‌


2‌ ‌