Data Scientist @Uber, MSDS @USF, IIT Bombay Alumnus, www.linkedin.com/in/alvira-swalin

Editor of USF-Data Science

May 15, 2018 · 1 min

Thanks for the blogs, it’s really useful.

1

Hi Etienne,

Thank you for reading. Here y, is the variable for which you are trying to predict (label) and x, z are the predictors. If you don’t have any label, and you can see that feature having missing values is dependent on other features from correlation matrix, then I would advise you to try predicting the missing…

2

Apr 21, 2018 · 1 min

Hey!
I read both the articles and found them quite interesting.

1

Hi Pushkar,

I am glad you read it. I agree some of the features you suggested will definitely improve the predictability, the only problem is that I don’t have that data. Whatever analysis I have done currently is based on the scraped data of HTML. If only I could get more information, I will try to incorporate those features. Anyways, thanks for suggestions.

Apr 14, 2018 · 1 min

the numerator in residuals is not variance of residuals.

1

Hi Sai! The reason why I wrote it as variance of residuals is because I was assuming the mean of residuals to be zero. Therefore to write variance of residuals, I can use the below formula which shows that sum of squares of residuals can be written as variance of residuals.

4