PyData Amsterdam 2016
Decision trees are very simple methods compared to Support Vector Machines, or Deep Learning. But they have some interesting properties that make them unique. For classification, for regression, or to extract probabilities, decision trees are easy to set up, and debug. And they are excellent to get a better understanding of your data.
This talk will cover Decision Trees, from theory, to their implementation in Python.
The talk will have a very practical approach, using examples and real cases to illustrate how to use decision trees, what we can expect from using them, and what kind of problems we will need to address.
The main topics covered will include:
- What are decision trees?
- How decision trees are trained?
- Data preprocessing for decision trees
- Understanding your data better with decision tree visualization
- Debugging decision trees using common sense and prior domain knowledge
- Avoiding overfitting, without cross-validation
- Python implementation