#What it does
- Fits a linear relationship between input Features and the target variable.
- Linear relationship occurs when the change in the target variable (dependent variable) is pr oportional to the change in an input feature (independent variable).
#How it Works
It minimizes the sum of squared residuals ([[3. Fitting the Model - Least Squares Approach|least squares]])
$$ SSR = \sum_{i=1}^{n} (y_i - (\beta_0 + \beta_1 x_i))^2 $$#Preconditions
- Assumes linearity and homoscedasticity
- No multicollinearity
#Evaluation
- $R^2$ [[Y2Q2/Machine Learning Foundations/R-squared|R-squared]]
- Mean Squared Error (MSE)
#Advantages
- Simple
- Interpretable
#Limitations
- Limited to linear relationships
- Sensitive to outliers
Latent Semantic Analysis (LSA)