Analytics1 A quick 10-question basic quiz 1.The purpose of a holdout sample is to a. Fit a model more tightly to the data b. Control overfitting c. Allow secondary modeling of residuals d. Reduce dimensionality 2.A key step in K-nearest neighbors is a. Calculation of conditional probabilities b. Use of multi-category predictors c. Rule generation d. Calculation of inter-record distance 3.Generation of rules in Association Rules relies on a. Data on individual transactions b. Data on user attributes c. Predictions made in a supervised learning routine d. Processing of unstructured text data 4.A Naive Bayes classifier is especially useful for a. Data with multi-category predictors b. Data with mixed numerical and categorical predictors c. Data with no holdout set d. Data with hierarchical clusters 5.A 2x2 confusion matrix is used for a. Fitting a logistic regression b. Assessing the accuracy of a classifier c. Exploratory data analysis d. Interpreting clusters 6.Principal components analysis is useful a. For reducing dimensionality b. For transforming multi-category data to binary (0/1) data c. As an optional intermediate component in a neural network d. In determining the optimum size of the holdout set. 7.One shortcoming of neural nets is a. Their inability to fit complex data patterns b. They rely on too many assumptions c. The models they produce are not easily interpretable d. They can only handle numerical predictor variables 8.An important advantage of classification and regression trees (CART) is a. They produce easily-interpretable decision rules b. The resulting model is relatively stable, if data change c. They are not prone to overfitting d. They rely on normality assumptions that provide robust results 9.Which of the following is NOT true of hierarchical clustering? a. Customer segmentation is a popular application of clustering b. The dendrogram can help decide the best number of clusters c. Normalization (standardization) is generally required d. The method is robust to changes in the distance metric. 10.In time series forecasting, autocorrelation a. Can help us make predictions b. Indicates lack of seasonality c. Suggests that a polynomial regression is in order d. Is not usually detectable First Name Last Name Email(Required)
