Exam AWS Certified Machine Learning - Specialty topic 1 question 92 discussion

D should be the more comprehensive answer. If it's not correlated, you can't make use of it in a linear regression A lot of others say B, but low variance can also be due to the nature/typical magnitudes of the variable itself

Answer B. Is not the best solucion prior can use other analysis. https://community.dataquest.io/t/feature-selection-features-with-low-variance/2418 If the variance is low or close to zero, then a feature is approximately constant and will not improve the performance of the model. In that case, it should be removed. Or if only a handful of observations differ from a constant value, the variance will also be very low.

D is the best answer as it is mentioned multivariable linear regression applied where correlation is strong between dependent and independent variables.

D: We should remove features that are strongly correlated with each other and weakly correlated with the target: https://androidkt.com/find-correlation-between-features-and-target-using-the-correlation-matrix/ You can evaluate the relationship between each feature and target using a correlation and selecting those features that have the strongest relationship with the target variable.

I think D is the correct answer. If I remember correctly, Benjamini-Hochberg Method is essentially answer D if you consider the Hypothesis to be: the feature is powerfully influential to the target. My problem with B is that the variance can be easily affected by the scale. In the question, the number of bedroom's variance is very low, while the sqrt of the house has a high variance, both of these could be very useful. Furthermore, zip codes are included, and it is safe to assume the variance of zip codes can be high, but the information is very limited, especially if you use them as numerical instead of categorical features.

B is correct but the answer in D is better.

D is preferred over C because the goal is to predict the sale price of houses, which is the target variable. By checking the correlation of each feature against the target variable, the machine learning specialist can identify which features are most relevant to the prediction of the sale price and which are less relevant. Removing features with low correlation to the target variable helps reduce the complexity of the model and potentially improve its accuracy. On the other hand, a heatmap showing the correlation of the dataset against itself (C) doesn't directly address the relevance of the features to the target variable, and so it's not as effective in reducing the complexity of the model.

Answer should be D, THIS is feature elimination /selection during feature Enginering. Choice c is so close just to confuse test takers to pick the wrong choice! See below C and D answers -- C should have been correct if the question asked about how to visualize correlation among independent variables! PROVIDED second sentence in C needs to be removed or to say which feature you will eliminate in such case then the one with low correlation against target out of those two. C. Build a heatmap showing the correlation of the dataset against itself. Remove features with low mutual correlation scores. D. Run a correlation check of all features against the target variable. Remove features with low target variable correlation scores.

The multiple regression model is based on the following assumptions: There is a linear relationship between the dependent variables and the independent variables The independent variables are not too highly correlated with each other yi observations are selected independently and randomly from the population Residuals should be normally distributed with a mean of 0 and variance σ

I think the answer is D. If the model is a decision tree or something like that, I don't think it is possible to make a decision based only on the direct correlation with the target variable. But in multiple linear regression, the only thing that matters is the relationship between the target variable and the feature variable. B, if the standard deviation is small but not zero, then we have information.

B is correct.

To eliminate extraneous information. So, the answer is D.

Correct answer is D. The reason B is wrong because it is difficult to reason out why would you plot a histogram? Absolutely unnecessary step and distraction choice.

D is not the proper answer. Here is why: It says that it is comparing with the target variable (dependent variable), which implies it is comparing the correlation between the dependent and independent variables. This type of comparison is usually done after a model is constructed in order to prevent assessing the predictive strength of the model. To compare the target label, the label you wish to predict, with the other variables before - is premature and will likely result in weakening your model. Variables with low variance has very less information and the inclusion of which will likely weaken the model performance. Hence, B.

Answer is D. https://deep-r.medium.com/difference-between-variance-co-variance-and-correlation-ea0b7ddbaa1

Answer C. Heatmaps is used to visualize for correlation matrix https://towardsdatascience.com/better-heatmaps-and-correlation-matrix-plots-in-python-41445d0f2bec

The problem with correlation tasks is it capture linear relations only. So, I would go with B