You have labelled data and are trying to predict a label based on known features. Historical data predicts likely future events. Applies already learned concepts from a training dataset to a test dataset.
Supervised learning consists of 2 parts
- Regression
- Classification
Learners
- Weak : Model which explains the variance more or less accurately
- Strong : Model which explains the variance accurately
Bias – Variance Tradeoff
Bias : how much on average is my predicted values different from actual values. High bias is underfitting.
Variance : how different will predictions be, at the same point, if different samples are taken from the same population. High variance causes overfitting.
high bias = high (pred – act) = high error = underfit model
high variance = high sensitivity in changes in data = overfit model
The Data Science Pitstop 