Solving Non-Linear Problem in Linear Way - A Data Science Perspective |
Author(s): |
| Ankur Tutlani , Microsoft India (R&D) Pvt. Ltd. |
Keywords: |
| Data Mining, Machine Learning, Predictive Analytics, Linear Regression, Feature Engineering |
Abstract |
|
In this paper, we discuss about the tradeoff between using non-linear regression algorithms and linear regression algorithm for non-linear problems. This tradeoff arises because non-linear regression algorithms are difficult to explain and implement in the production environment but provide better accuracy. On the other hand, linear regression algorithm is easy to explain to the business and implement in the production environment but has relatively less accuracy in case of non-linear regression problems. The paper starts with defining non-linear regression problem and proceeds with explaining how to solve non-linear regression problem using linear regression by transforming the feature space. The paper concludes with illustrating some of the methods which can help achieve better accuracy from applying linear regression algorithm to the non-linear regression problem. |
Other Details |
|
Paper ID: IJSRDV5I31172 Published in: Volume : 5, Issue : 3 Publication Date: 01/06/2017 Page(s): 1342-1344 |
Article Preview |
|
|
|
|
