Define the Problem

First define the number of output variables (univariate, bivariate, multivariate), the machine learning paradigm of the problem, then the type of problem, then the algorithm(s) to be used:

Supervised

• Regression - Predicting a value
• Classification - Predict a class

Unsupervised

• Clustering - Group ungrouped data
• Association Rule Mining - Find associations between two or more classes

Reinforcement

Data Exploration

Numerical Exploration

Take measures of central tendency

Check for missing values

Data Visualization

Histograms

Scatter Matrices

• These plot every numerical variable against every other numerical variable to help you visually seek out correlations

Data Preprocessing