This video is the second part of the lecture on Linear Regression. We begin by reviewing what the Stochastic and Batch Gradient appraoches look like in the vecctor matrix notation that we have used in the previous video. We then look at the metrics used to measure the efficiency of a Linear Regression model such as the R-square score.
Next we move on to look at the "Data Assumptions" - what kind of data does Linear Regression give good results on. In particular we review ideas such as Linearity, Multicollinearity, Residue Normality, Homoscedasticity - what they mean, their impact and how to detect or measure them. We talk about Pearson Coefficient and about VIF (Variance Inflation Factors).
Next we illustrate a real-life problem, and an example of the approach and steps you would take to solve it using Linear Regression. Finally we list a few questions / exercises that would provide further insights to the thoughtful viewer.
Contents:
6:40 - Stochastic and Batch Gradient
20:53 - R-square score
38:38 - Linearity
43:10 - Pearson Coefficient
52:09 - Multicollinearity
59:49 - VIF
1:17:25 - Normality
1:22:37 - Homoscedasticity
1:25:43 - Approach to a real-life problem
1:44:50 - Questions / Exercises