Data Science and Machine Learning Essentials - Edx

QUESTION 1

(1 point possible)

You have trained and tested a regression model, and plotted the predicted labels against the actual label values in the training and test data. The resulting chart shows strong correlation between the predicted labels and the training data, but the actual label values from the test data vary significantly from the predicted values.

What can you conclude about your model?

 The model will generalize well. The model is under-fitted. The model is under-fitted. - incorrect The model is over-fitted. The model will work better as an unsupervised learning model.

QUESTION 2

(1 point possible)

You have amassed a large volume of customer data, and want to determine if it is possible to identify distinct categories of customer based on similar characteristics.

What kind of predictive model should you create?

 A recommender system. A clustering model. A regression model. A classification model.

 

QUESTION 3

(1 point possible)

You have created a two-class classification model, and used it to generate a ROC curve chart from a dataset of test data. The area under the curve is completely below the straight diagonal line through the X,Y axis.

What can you conclude about your classification model?

 The model predicts labels that are less accurate than could be achieved by guessing. The model predicts labels that are more accurate than could be achieved by guessing. The model will generalize well when applied to unknown label values. The model exhibits the highest possible level of accuracy.

<a href="http://click.linksynergy.com/fs-bin/click?id=HJJ4jiwN8VY&offerid=311675.10000122&subid=0&type=4&LSNSUBSITE=LSNSUBSITE"><IMG border="0"   alt="Rakuten Affiliate Network Welcome Program" src="http://ad.linksynergy.com/fs-bin/show?id=HJJ4jiwN8VY&bids=311675.10000122&subid=0&type=4&gridnum=1"></a>

QUESTION 4

(1 point possible)

You are creating an experiment in Azure ML Studio, and you want to remove two columns from a dataset before preparing the data for modeling.

Which two of the following modules can you use to remove the unwanted columns?

 Project Columns Split Join Apply SQL Transformation Clean Missing Data

- unanswered

Note: Make sure you select all of the correct options—there may be more than one!

QUESTION 5

(1 point possible)

You are creating an experiment in Azure ML Studio based on a dataset containing weather data. You want to add a column to the dataset. The new column should contain the existing Temperature value squared.

Which three of the following modules could you use to create the calculated column?

 Project Columns Execute R Script Metadata Editor Apply SQL Transformation Execute Python Script

- unanswered

Note: Make sure you select all of the correct options—there may be more than one!

QUESTION 6

(1 point possible)

You create an Azure ML experiment and add a dataset. You visualize the data in an attempt to identify which columns contain outliers.

What indication of outliers should you look for?

 Values on the extreme right hand side of a histogram. Extreme values unrelated to the values of other variables in multiple chart views of the data relationships. Extreme values in any single chart view of the data relationships. Extreme values in a Q-Q Normal plot of the variables.

QUESTION 7

(1 point possible)

You discover that there are missing values for an unordered numeric column in your data.

Which three approaches can you consider using to treat the missing values?

 Substitute the text “None”. Forward fill or back fill the value. Remove rows in which the value is missing. Interpolate a value to replace the missing value. Substitute the numeral 0

- unanswered

Note: Make sure you select all of the correct options—there may be more than one!

QUESTION 8

(1 point possible)

You are preparing data to train a machine learning model. You want to use all applicable data to train the model, but you must ensure that variables with a wide range of numeric values do not dominate the training of the model.

What should you do?

 Remove numeric columns that contain a wide range of values. Scale numeric columns that contain a wide range of values. Convert numeric columns with a wide range of values to string features. Clip the values of the numeric columns to replace the top and bottom 20% with the mean value.

QUESTION 9

(1 point possible)

You are creating an experiment based on real estate transactions data in which each sale of a property is indicated by a unique transaction ID field. The data is entered into the source database manually, and some records may have been entered more than once. You plan to use the data to train a regression model that predicts property prices.

What should you do to reduce bias in your data that might affect the predictive model?

 Use only the first 50% of the rows to train, score, and evaluate the model. Exclude the transaction ID column from the training set for the model. Remove rows containing duplicate transaction IDs, retaining only the first instance. Scale the transaction ID column to normalize the data.

QUESTION 10

(1 point possible)

You are preparing data for a predictive model. After several iterative experiments to remove detrimental features and evaluate the model, you have reduced your dataset to five features, including one feature the presence of which neither improves or worsens the performance of the model.

What should you do?

 Use all five remaining features to train the model. Remove the feature that has no effect on model performance. Remove all features other than the one that doesn’t affect model performance. Keep iteratively removing features until only the most impactful one remains.

QUESTION 11

(1 point possible)

You create an experiment to train, score, and evaluate a regression model. You must ensure that the training, testing, and evaluation of the mode is unbiased, with no inter-dependency on the data.

What should you do?

 Train, score, and evaluate the model with the same data. Split the data into two non-overlapping subsets. Then train the model with the first subset, and score and evaluate the model with the second subset. Split the data into two non-overlapping subsets, and train the model with the first subset. Then join the first and second subsets and use the joined dataset to score and evaluate the model. Split the data into three non-overlapping subsets. Then train the model with the first subset, score the model with the second subset, and evaluate the model with the third subset.

QUESTION 12

(1 point possible)

When assessing the residuals of a regression model you observe the following:

• Residuals exhibit a persistent structure and are not randomly distributed with respect to values of the label or the features.
• The Q-Q normal plots of the residuals show significant curvature and the presence of outliers.

Given these results, which two of the following things should you try to improve the model?

 Cross validate the model to ensure that it will generalize properly. Try a different class of regression model that might better fit the problem should be tried. Create some engineered features with behaviors more closely tracking the values of the label. Add a Sweep Parameters module with the Metric for measuring performance for classificationproperty set to Accuracy.

QUESTION 13

(1 point possible)

When exploring the k-means clustering of a dataset you continue to increase the number of clusters one-by-one until you observe that the projection of the first two principle components show:

• The major and minor axes of each of the ellipses are of similar lengths.
• The directions of major axes of the ellipses are relatively aligned.

Based on these observations, what should you do?

 Publish the experiment as it is; the number of clusters chosen fits this dataset well. Try an alternative modeling approach to clustering. Reduce the number of clusters chosen. Increase the number of the clusters chosen.

QUESTION 14

(1 point possible)

You create an Azure ML experiment based on a dataset that includes a numeric label. You plan to create a binary (two-class) classification model to predict this label.

What should you do to prepare the data for the classification model?

 Apply a logarithmic transformation to the values of the label and features using the Normalize Data module. Z-Score normalize the values of the label and features with the Normalize Data module. Quantize the label so that it has two categorical levels or values with the Quantize Data module. Quantize the features so they all have two categorical variables values with the Quantize Data module.

QUESTION 15

(1 point possible)

You create a binary (two-class) classification machine learning model. When evaluating the model, you observe the following metrics:

• Accuracy: 0.9
• Area under the curve (AUC): 0.8
• Recall: 0.3

What can you conclude about the performance of your model?

 Proportionally, false positives predominate the errors. Proportionately, false negatives predominate the errors. There are equal numbers of true positives and true negatives. There are equal numbers of false positives and false negatives.

QUESTION 16

(1 point possible)

You create a regression model, and then cross-validate it.

You observe the following results:

• The values of the performance metrics change significantly from fold to fold.
• The standard deviation of the performance metrics is close to the value of the metrics.

What can you conclude about your mode?

 The model is a good fit to the data. The model will likely generalize well. The model will likely not generalize well. The model is insensitive to the test data in the cross validation folds.

QUESTION 17

(1 point possible)

MYou create an Azure ML experiment. You must create a Decision Forest Regression model, but you aren’t sure of the optimal parameter values to use for your data and prediction objectives.

What should you do to ensure optimal performance of the model with the least effort?

 Use a Bayesian Linear regression model, which has fewer parameters. Use the Sweep Parameters module to determine the optimal parameters. Leave the default values of the Decision Forest regression model unchanged. Create multiple experiments with every possible combination of the parameter values.

QUESTION 18

(1 point possible)

You have created custom function in R or Python that returns the square root of the values that are passed to it. You want to use the custom function in multiple Azure ML experiments.

What should you do?

 Add the code file to a ZIP archive and upload it to Azure ML as a dataset. Copy and paste the code into an Execute R Script or Execute Python Script module in every experiment where you plan to use it. Export the output of every experiment to a CSV file, and then run the custom function against the exported data in a local development tool. Convert the code to the equivalent SQL, and use the Apply SQL Transformation model in each experiment where you want to use the function.

QUESTION 19

(1 point possible)

You create an experiment that uses a Train Matchbox Recommender module to train a recommendation model, and add a Score Matchbox Recommender module to generate a prediction. You want to use the model in a music streaming service to recommend songs for the currently logged in user.

Which recommender prediction kind should you configure the Score Matchbox Recommender module to use?

 Item Recommendation Related Items Rating Prediction Related Users

QUESTION 20

(1 point possible)

You publish an Azure ML experiment as a web service for use by a client application that must generate predicted sales values for high volumes of product data asynchronously.

Which endpoint should the application use to consume the web service?

 The URL of the experiment in Azure ML Studio. The Request-Response Service URL for the web service. The Batch Execution Service URL for the web service. The URL of an Excel workbook in a OneDrive folder.