Page 38 - FINAL CFA II SLIDES JUNE 2019 DAY 3

READING 8: MULTIPLE REGRESSION AND ISSUES IN REGRESSION ANALYSIS
Classification and Regression Trees (CART)
MODULE 8.10: SUPERVISED AND UNSUPERVISED MACHINE LEARNING

Classification trees are appropriate when the target variable is categorical, used when the target is binary (e.g., IPO will be
successful vs. not successful).
Logit and probit models, are used when the target is binary but are ill-suited when there are significant nonlinear relationships
among variables. In such cases, classification trees may be a viable alternative.

Regression trees are appropriate when the target is continuous.

Neural Networks (Artificial Neural Networks)
Constructed with nodes connected by links. The input
layer is the nodes with values for the features
(independent variables). Each node uses an activation
function, typically a nonlinear function, to generate a
value from the weighted average of the input values from
those nodes linked as inputs to each hidden node. There
may be multiple hidden layers with linked nodes. Again,
each node in a hidden layer will use an activation function
to the values from other nodes. There is typically a single
node in the output layer that is the prediction of the model.
Linking the information in the input layer to multiple nodes
in the hidden layers, each with its own activation function,
allows the neural network to model complex nonlinear
functions to use the information in the input variables well.

The researcher must determine the structure of the network called the hyperparameters. Additional layers can improve the
predictive accuracy of neural networks. Deep learning nets (DLNs) are neural networks with many hidden layers (often more
than 20).