Overview

1. Importance of Preprocessing

2. Model Candidates

3. Cross-Validation

4. Using Random Forests to Answer our Research Question

5. Conclusion

Importance of Preprocessing

• data preprocessing converts raw data and signals into data representation suitable for application through a sequence of operations

• objectives: include size reduction of the input space, smoother relationships, data normalization, noise reduction, and feature extraction

• even though the process is very time consuming, it has huge impact on data modelling. The quality of the preprocessing may vary from case to case.

Size Reduction of the Input Space:

objective: get a reasonable generalization with a lower dimensionality of the data set without losing the most significant relationship of the data.

Smoother Relationships:

• original problem transformed into a simpler problem. It means that the associated mappings become smoother. The transformations can be obtained from intuition about the problem.

Normalization (Scaling):

• Unnecessarily complex relationships can be avoided by normalizing (or scaling) each of the input variables so that the variance of each variable is equal.

Noise Reduction:

• A sequence of data may involve useful data, noisy data, and inconsistent data. Preprocessing may reduce the noisy and inconsistent data.

Feature Extraction:

-If key attributes or features characterizing the data can be extracted, problem encountered can be easily solved.

Model Candidates

Research Question:

• explore 3 model candidates in our attempt to answer the research question:

1. classification and regression trees (CART)

2. bagged trees

3. random forests

Non-Ensemble Techniques: Motivation for Ensembles

Classification and Regression Tree (CART)

• term Classification And Regression Tree (CART) used as a generic term to describe the class of models that use trees to either predict the class to which data belongs (classification) or predict a real number (regression tree)

• Decision trees can be thought of as machine-generated business rules