I was assigned the task of linearizing some of my data, which exhibits a non-linear appearance. When using the distfit library, it indicated that my data's distribution is closest to a gamma function. enter image description here

The primary objective of linearization is to accurately interpret the weights, synergy, and VIF (Variance Inflation Factor) of each variable, as well as to enable the clustering of subgroups within certain features in order to reassess their weights.

A straightforward linear regression won't yield satisfactory results, so I attempted applying linearization functions (ln, log2, log10, etc.) with no effect.

My question is, how can I effectively achieve this when dealing with multiple variables, especially when visually seeing it is challenging? Additionally, some of the variables are qualitative.

Is linearization the right approach? While some models may handle non-linear relationships, they often have poor interpretability.

Example of the scatterplot of Target x feature1 (before scaling): enter image description here



Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Browse other questions tagged or ask your own question.