• A Broad overview of the Model Building Process
  • Train-test split
  • Preprocessing
• Explore Model Validations with Cross-Validation Methods
  • Hyper parameter tuning
  • Alliviate Over-fitting

• Creates a unified interface for modeling and prediction
  • A total of over 200 Models and Algorithms
• Pipeline ready modeling
• Added Parallel computing efficiency

• Gain domain knowledge (if possible)
  • Gain intuition of the problem or data
• Estimate model parameters - learned parameters
  • In linear regression, intercept and slope(s)
• Determine tuning parameters, hyperparameters, structural parameters
  • In KNN, estimating the K is not data dependent
• Calculate the performance of the model, and how it will generalize to new data
• Repeat

Caret implements a unified interface (syntax) for all of R's most popular models. The basic syntax is as follows:

model <- train(formula = y ~ x, 
              data = dataset, method = 'lm'
              trControl = 'additional options')

Add smoother to scatterplots