Power of Transforming Data Before Feeding Into Linear Regression(in this case SLR)!
Without beating around the bush, my agenda is to make you realize how important it is to transform the data before building a linear regression model.
Consider this data and its corresponding transformations:
| feature | squared | log base 10 | squared | sqrt | |
|---|---|---|---|---|---|
| 1 | -10 | 100 | 2 | 1.414213562 | 1.189207115 |
| 2 | -20 | 400 | 2.602059991 | 1.613090199 | 1.270074879 |
| 3 | 1000 | 1000000 | 6 | 2.449489743 | 1.56508458 |
| 4 | -10000 | 100000000 | 8 | 2.828427125 | 1.681792831 |
| 5 | 10000 | 100000000 | 8 | 2.828427125 | 1.681792831 |
| 6 | 100000 | 10000000000 | 10 | 3.16227766 | 1.77827941 |
No see the difference in the transformed data from the eyes of a SLR:
