1.Which Machine Learning algorithm uses the ExpectationMaximization (EM) algorithm? List the high-level steps for this machine-learning algorithm. Just copy the algorithm from the PDSH book PDF file.

ANS:

2.One way to reduce the number of features in a dataset for which you are performing linear regression is to use backward stepwise regression. You start out with a full model and systematically drop features one at a time from the model using a logical procedure. We used this feature reduction method along with statmodels OLS regression in Lab 6. Which of the statistics returned from OLS were used to decide which feature to drop?

Drop the feature with the highest R-squared value.

Drop the feature with the lowest MSE value

Drop the feature with the highest coefficient p-value.

ANS:

3. A statistical model is developed by training the machine learning algorithm using training data. In most cases, this is just a subset of all the possible data for the problem for which the model is being developed. We want to develop a model that also works well with unseen data, called test data. The models that we build can overfit or underfit the data. With this in mind, which of the following statements is false:

Using Ridge regression versus linear regression on a dataset can improve predictability by reducing overfitting.

A model that underfits the data, does not predict well on the test data, and does not predict well on the training data because there are too many features in the model.

Overfitting occurs when the model fits too well on the training data, and the model does not fit as well on the test data.

Eliminating some of a models less significant features can help reduce overfitting.