Statistical Learning and Data Science (R Studio)
Budget: $30 – $250 USD
Examine a provided dataset using multiple methods (listed below) and estimate the model error involved in each using cross-validation. Implement each of the following methods:
KNN
LDA
QDA
Naïve Bayes
Use 10-fold cross-validation to estimate the error involved in each method.
Assemble all of your work in a .Rmd file and its knitted .html output. Your .Rmd file should contain all codes necessary to read in the data, perform the splitting required, implement each method, and compute the cross-validated error estimate.
Data: midterm_data.csv
Data notes: Data contains a categorical response with 3 levels (A, B, and C) and 3 numeric predictors.
Part II - Conceptual
Sketch a graphical representation of a data set with 2 numeric predictors and a categorical response with 3 or more levels (see Figure 1.4 on page 5 of the textbook for an example of what this sketch might look like), for which:
KNN will work better than LDA or QDA
LDA will work better than KNN or QDA
QDA will work better than KNN or LDA
For each sketch, give a brief explanation of why you believe that the given method will outperform the other two. (In your explanation, assume that the sketch represents the entire population of interest.)
KNN
LDA
QDA
Naïve Bayes
Use 10-fold cross-validation to estimate the error involved in each method.
Assemble all of your work in a .Rmd file and its knitted .html output. Your .Rmd file should contain all codes necessary to read in the data, perform the splitting required, implement each method, and compute the cross-validated error estimate.
Data: midterm_data.csv
Data notes: Data contains a categorical response with 3 levels (A, B, and C) and 3 numeric predictors.
Part II - Conceptual
Sketch a graphical representation of a data set with 2 numeric predictors and a categorical response with 3 or more levels (see Figure 1.4 on page 5 of the textbook for an example of what this sketch might look like), for which:
KNN will work better than LDA or QDA
LDA will work better than KNN or QDA
QDA will work better than KNN or LDA
For each sketch, give a brief explanation of why you believe that the given method will outperform the other two. (In your explanation, assume that the sketch represents the entire population of interest.)