This lecture elaborates on decision trees for classification and regression tasks. We discuss the Classification And Regression Tree, a.k.a CART, algorithm. In a classification approach, we discuss two costs that are minimized in a greedy manner using the sub-optimal CART algorithm, which are the Gini impurity and the Entropy. In a regression approach, this cost is simply replaced by the mean-squared error (MSE). Furthermore, we discuss the sub-optimality of the CART algorithm, which does not necessarily yield the optimal tree. On the other hand, obtaining the optimal tree is an NP-complete problem. The other part of the lecture is dedicated for python implementations of both Decision Tree classifiers and regressors. We also show how to plot the tree on python using graphviz. Finally, we end the lecture with a Stock Market Analysis case study with the intent of predicting buy or sell signals. We study different stocks such as google, tesla and zomedica.

⏲Outline⏲
00:00:00 Introduction
00:01:11 Decision Tree Classifiers
00:02:31 The CART Algorithm
00:05:39 Gini Impurity
00:10:43 CART Sub-optimality
00:13:31 Entropy
00:21:52 scikit-learn: Decision Tree Classifiers
00:25:38 Viewing decision trees using graphicviz
00:30:41 Plotting Decision Boundaries on Python
00:38:08 Soft Decision Tree Classifiers
00:40:45 Decision Tree Classifiers & Rotation Sensitivity
00:46:42 Decision Tree Regression
00:49:57 scikit-learn: Decision Tree Regressor
00:56:11 Stock Market Analysis: Decision Trees Predicting Buy & Sell Signals

Instructor: Dr. Ahmad Bazzi

Subscribe: https://www.youtube.com/channel/UCgC1d4JZ1Fz4t8MWLJD464w

#machine-learning #tensorflow