TDSM 7.23
From The Data Science Design Manual Wikia
Revision as of 20:11, 11 December 2017 by Xiaoxi2002cwz (talk | contribs)
The correlation coeff. will show how each metric is related to the final results. We can use feature selection to see how effective a feature is in predicting the results. Some feature selection algorithms are Linear Regression, Random-Forest and Randomised Lasso.
Also, it is important to combine the predictive power with certain problem. For example, when we predict earthquake, we expect to enhance the recall of the model. However, when we predict terrorist, precision plays a more important role.