When the data is not linear, linear regression works poorly because it presupposes a linear connection between the dependent and independent variables. This leads to biassed and erroneous forecasts. The model’s residuals (errors) are likely to be big and not uniformly distributed, suggesting a poor fit. Furthermore, the R-squared value, which represents the proportion of variation explained by the model, will be low. Alternative approaches for dealing with nonlinear relationships include polynomial regression, decision trees, and more complex techniques like as neural networks and support vector machines. These approaches are more successful in capturing and modelling underlying patterns. join demo class contact Durga Online Trainer
