There's no science without jargon, is there?
predictors: what we use to predict what we want or are interested in
synonyms: input, independent variables, features, explanatory variables, variables, attributes
response: what we want to predict or are interested in
synonyms: output, dependent variable, target
fit a model: create a model from observations
synonyms: train
training data: the data we use to create the model
test data: data we reserve to assess the quality of the model we created with the training data
synonyms: holdout data
signal variables: predictors that are actually related to the response
noise variables: predictors that are unrelated to the response
prediction: we want to estimate the response accurately
inference: we want to understand how predictors influence the response
supervised learning: working with a response variable - prediction or inference
unsupervised learning: no response variable - clustering
regression problems: dealing with a quantitative (numerical) response
classification problems: dealing with a qualitative (categorical) response
parametric models: assume a functional form for the model and estimate its parameters
non-parametric models: make no assumptions about the model's form and estimate it directly
No comments:
Post a Comment