70-773 무료 덤프문제 온라인 액세스

You have cloud and on-premises resources that include Microsoft SQL Server and a big data environment in Apache Hadoop.
You have 50 billion fact records.
You need to build time series models to execute forecasting reports on the fact records.
What should you use?

• A.RxLocalParallel on the SQL Server database
• B.RxHadoopMR on the Hadoop cluster
• C.RxLocalseq on the SQL Server database
• D.RxSpark on the Hadoop cluster

You plan to analyze data on a local computer. To improve performance, you plan to alternate the operation between a Microsoft SQL Server and the local computer.
You need to run complex code on the SQL Server, and then revert to the local compute context.
Which R code segment should you use?

• A.sqlCompute <- RxInSqlServer(connectionstring = sqlConnString, shareDir = sqlShareDir,wait = sqlWait, consoleOutput = sqlConsoleOutput)rxSetComputeContext("local")x <- 1:10rxExec(print, x, elemType =
  "cores", timesToRun = 10)rxSetComputeContext("sqlCompute")
• B.sqlCompute <- RxInSqlServer(connectionString = "Driver=SQL Server;Server = myServer; Database = TestDB; Uid = myID; Pwd = myPwd;")sqlPackagePaths <- RxFindPackage(package = "RevoScaleR", computeContext = sqlServerCompute)
• C.sqlCompute <- RxInSqlServer(connectionstring = sqlConnString, shareDir = sqlShareDir,wait = sqlWait, consoleOutput = sqlConsoleOutput)rxSetComputeContext("local")x <- 1:10rxExec(print, x, elemType =
  "cores", timesToRun = 10)rxSetComputeContext("RxLocalParallel")
• D.sqlCompute <- RxInSqlServer(connectionstring = sqlConnString, shareDir = sqlShareDir,wait = sqlWait, consoleOutput = sqlConsoleOutput)rxSetComputeContext("sqlCompute")x <- 1:10rxExec(print, x, elemType
  = "cores", timesToRun = 10)rxSetComputeContext("local")

You are running a large logistic regression for 1,000 feature variables by using the LoisticRegression() function in the MicrosoftML package. All of the predictor variables are numeric.
Currently, you specify the input variables separately by using the following formula.
Outcome ~ Feature000 + Feature001 + Feature002 + ... + Feature999
You discover that it takes 20 minutes to estimate each model.
You need to reduce the amount of time required to estimate each model without losing any information in the predictors.
What should you do?

• A.Use concat() to create a single array variable named Features, and then specify a new formula named Outcome ~ Features.
• B.Use selectFeatures() to select the features that provide the most information about the outcome variable.
• C.Use princomp() on the correlation matrix of Features, and then use only the first 100 principle components to reduce the number of input variables.
• D.Use stepControl() to perform stepwise regression to limit the number of variables that contribute to the model.

You have one-class support vector machines (SVMs).
You have a large dataset, but you do not have enough training time to fully test the model.
What is an alternative method to validate the model?

• A.Use Principal Components Analysis (PCA)-Based Anomaly Detection.
• B.Replace the SVMs with two-class SVMs.
• C.Perform feature selection.
• D.Use outlier detection.

You have a slow Map Reduce job.
You need to optimize the job to control the number of mapper and runner tasks.
Which function should you use?

• A.rxExec
• B.RxLocalParallel
• C.RxComputeContext
• D.RxHadoopMR