Linear Regressions

Source: bigdata-madesimple

Scikit-learn is a Python module for machine learning. It can be used for regression, classification, clustering, model selection, and dimensionality reduction.

In linear regressions, Y represents the depending variable and X represents the independent variable(s).

from sklearn.linear_model import LinearRegression is required to use the scikit-learn linear regression functions.

Important functions:

• lm.fit() - fits a linear model
• lm.predict() - Predict Y using the linear model with estimated coefficients.
• lm.score() - Returns the coefficient of determination (R^2)
• lm.coef_&lm.intercept_ - give the coefficients and estimated intercepts of a model.

Normally a regression won’t be performed on an entire data set. Data should be split into training and test sets. This allows us to find a model based on fit to training data, then test it against real data to see how well it predicts outcomes.

Division of data for training and tests must be done randomly to prevent bias in the model.

Residual plots can be used to visualize errors in data. In a residual plot, data should be scattered around line zero without structure. If the data appears to show structure, the model missed something and should be revised.