3. Consider the following table of cross validation on tree induction for a two-class classification problem,as discussed in clasS: Root node error: 1524/3100 0.49161 n- 3100 CP nsplit rel error xerror 1 0.54757282 2 0.11909385 3 0.06601942 4 0.05372168 5 0.05242718 6 0.03430421 7 0.01326861 8 0.01165049 9 0.01035599 10 0.00776699 11 0.00550162 12 0.00517799 13 0.00453074 14 0.00291262 15 0.00258900 16 0.00226537 17 0.00194175 18 0.00161812 19 0.00129450 20 0.00097087 21 0.00064725 22 0.00032362 23 0.00000000 xstd 0 1.0000000 1.025890 0.0180141 10.45242720.462136 0.0151731 2 0.3333333 0.368932 0.0139601 3 0.2673139 0.293204 0.0127297 4 0.2135922 0.238835 0.0116698 5 0.1611650 0.186408 0.0104615 6 0.1268608 0.150809 0.0095013 8 0.1003236 0.127508 0.0087912 9 0.0886731 0.119741 0.0085368 10 0.0783172 0.100324 0.0078541 12 0.0627832 0.083495 0.0071968 14 0.05177990.077023 0.0069238 15 0.04660190.073786 0.0067825 17 0.0375405 0.066019 0.0064285 19 0.0317152 0.062783 0.0062741 20 0.0291262 0.057605 0.0060178 22 0.0245955 0.055663 0.0059185 24 0.0207120 0.054369 0.0058512 26 0.0174757 0.054369 0.0058512 30 0.0122977 0.050485 0.0056440 34 0.0084142 0.045955 0.0053910 42 0.0032362 0.048544 0.0055371 52 0.0000000 0.046602 0.0054279 (a) In the "rel error" column we get a value of 0. for the 23rd row. Explain what this number means (b) In terms of error rate, how well do you think the resulting maximally-deep tree will perform on different data from the sample population class. How well will this tree classify new data from the same population your a value best? (c) Consider the tree that makes no splits i.e. the one that simply classifies according to the most likely (d) Juding from the table, what appears to be your best choice of complexity parameter a? In what sense is 3. Consider the following table of cross validation on tree induction for a two-class classification problem,as discussed in clasS: Root node error: 1524/3100 0.49161 n- 3100 CP nsplit rel error xerror 1 0.54757282 2 0.11909385 3 0.06601942 4 0.05372168 5 0.05242718 6 0.03430421 7 0.01326861 8 0.01165049 9 0.01035599 10 0.00776699 11 0.00550162 12 0.00517799 13 0.00453074 14 0.00291262 15 0.00258900 16 0.00226537 17 0.00194175 18 0.00161812 19 0.00129450 20 0.00097087 21 0.00064725 22 0.00032362 23 0.00000000 xstd 0 1.0000000 1.025890 0.0180141 10.45242720.462136 0.0151731 2 0.3333333 0.368932 0.0139601 3 0.2673139 0.293204 0.0127297 4 0.2135922 0.238835 0.0116698 5 0.1611650 0.186408 0.0104615 6 0.1268608 0.150809 0.0095013 8 0.1003236 0.127508 0.0087912 9 0.0886731 0.119741 0.0085368 10 0.0783172 0.100324 0.0078541 12 0.0627832 0.083495 0.0071968 14 0.05177990.077023 0.0069238 15 0.04660190.073786 0.0067825 17 0.0375405 0.066019 0.0064285 19 0.0317152 0.062783 0.0062741 20 0.0291262 0.057605 0.0060178 22 0.0245955 0.055663 0.0059185 24 0.0207120 0.054369 0.0058512 26 0.0174757 0.054369 0.0058512 30 0.0122977 0.050485 0.0056440 34 0.0084142 0.045955 0.0053910 42 0.0032362 0.048544 0.0055371 52 0.0000000 0.046602 0.0054279 (a) In the "rel error" column we get a value of 0. for the 23rd row. Explain what this number means (b) In terms of error rate, how well do you think the resulting maximally-deep tree will perform on different data from the sample population class. How well will this tree classify new data from the same population your a value best? (c) Consider the tree that makes no splits i.e. the one that simply classifies according to the most likely (d) Juding from the table, what appears to be your best choice of complexity parameter a? In what sense is
