BCOR 2205 Exam Questions and Answers 2023

Which of the following is/are part of the success criteria of our model. A) Who uses the model B) How much the model can drive C) All the above - Answer- C The who and what we are studying - Answer- Unit of analysis We don't need to consider ethical reasons when developing our machine learning model. TRUE FALSE - Answer- False In the following example, which is the entity: account, ..... - Answer- Account Data used by your organization that is not collected by org is called - Answer- External Data Which of these is not a data source business use A) SME B) ER Database C) CSV D) None of these are use above - Answer- A Select all that apply (ER databases) A) Customer number is primary key for customer table B) Order number is primary key for orders table C) Customer number is the foreign key for order table D) All of the above - Answer- D Acquire explore and manipulate data is first step in ML life cycle True False - Answer- False The foreign key must be unique and non null True False - Answer- False The three types of dirty data mentioned in class are: inaccurate data, missing data, and duplicates. True False - Answer- True The primary key in one table becomes the foreign key in another associated table True False - Answer- True Examples of target leakage A) Trying to see if someone will open an account by using account number as feature B) When predicting if someone applying for an auto loan will default and credit score at time of application is feature C) When predicting customer call churn and length of call to call center when we know churn calls last longer - Answer- A If the following accuracy scores were found. What would the cross validation score be if using 5-fold-cross-validation (20/20, 10/20, 5/20, 15/20, 20/20) A) 1 B) 70 (maybe) C) .70 D) Not enough info - Answer- C The goal of machine learning is to build models with.. A)Higher prediction B) High Generalization capability C) Both - Answer- C Lets assume we have 100 rows of data, what will the size be of the holdout (20%)? What about each fold? - Answer- 20% holdout, 16 rows per fold If the original validation proportion is not representative of the overall population, then the resulting model may appear to have high ______ when in reality it just happens to fit the unusual validation set well. A) Accuracy B) Precision C) Correlation D) Values E) None of the above - Answer- A The large gap between data collected and info decision makers require is? - Answer- Analysis gap would be an example of A) Data B) Information - Answer- A Which of the following is / are characteristics of Auto ML A) The user does not need to define features B) Models will be ranked C) Models can be run simultaneously