01:49 (Q1) What are the assumptions for Linear Regression? 03:33 (Q2) What is the difference between Precision and Accuracy, can you explain in terms of Confusion Matrix and Confidence Interval? 06:36 (Q3) Normal distribution question 09:36 (Q4) What is the power of a hypothesis test? Why is it important? 11:33 (Q5) What is the difference between K nearest neighbors and K means? 13:19 (Q6) Explain Random Forest in Layman terms. 17:40 (Q7) What is K fold cross-validation? Why do we use it? 19:31 (Q8) Explain how we can handle missing values in our data? 22:21 (Q9) What is the difference between a Bar Graph and a Histogram? 24:17 (Q10) What is a Box and Whisker Plot and when should we use it?
Much appreciating your efforts. Useful video to get a fair idea on what diff type of questions one can face. I have one small observation as follow: Explanation of question 6, is more of a feature selection rather than how actually a random forest works. It should be first deciding some most important parameters (feature selection) and then asking people what is best out of 3 offers based on selected parameters. People will respond as per their experience and learning and recommend an offer to go ahead. The highest number of people recommending an offer is what I will go ahead. Now, in technical term, these number of people are my trees (forest).
16:00 Random Forest LinkedIn Example. There are 10 companies you got a job offer. You show 3 of them to your LinkedIn contacts. If change the example like this we consider both bagging and subspace sampling. (Bagging-Bootstrap Aggregation is to use some of the data, subspace sampling is to use some of the features. )
I like the way you tried it, but unfortunately you lack explanation , when you say linear relationship please be specific about linear relation between what? if i have a quadratic data, you mean i can't run linear regression?
Not sure about the explanation for question 1. Nor why linear regression requires the assumptions for best linear unbiased estimators. Can still fit a linear regression just fine without them, albeit it might invalidate confidence intervals and interpretability. Who cares what the assumptions are if MSE is low?
do you think the MSE will be low if the assumptions are not true? Low MSE in itself means that the residuals are small and the predicted linear regression line closely represents the actual pattern in data.
Aaa... You ggggggive excellent explanation.. Please use perfecttttttt mic .. Because a lotttt of unnecessary sounds coming from your mouthhhh.. Those are getting irritattttted me.
for someone who says "communication is key to data science" you seem to use the visual medium of video very poorly... Reading exactly what's on screen 10x faster than you reading it out really bugged me. I realize actual video can take a lot longer to make, or cost a lot more. But PowerPoints with text dumps haven't been 'good communication' since 2000. Sorry if this was harsh, great info. I just muted it and scrolled through each slide - which would've been a better experience on slideshare sadly.