Question

(a) Which points would you consider outliers by visual inspection? Mark these points on a printout of the figure. (b) Using the following table, state which points would be considered outliers using the distance to kth-nearest neighbor algorithm (assume k-10). Use a threshold of 0.5 on the distance to the 10th nearest neighbor to determine outliers. xy Distance to 10th nearest neighbor 1.02 5.04 4.00 2.50 4.02 5.39 3.31 3.73 1.11 5.09 0.94 4.48 4.07 5.38 4.17 5.85 4.63 3.92 0.15 2.01 0.42 1.13 0.56 0.38 0.72 0.90 (c) Using the following table, state which points would be considered outliers using the LOF algorithm. Use a threshold of 2.5 for the LOF score to determine outliers. x | У | Distance to 10th nearest neighbor 0.83 4.88 1.02 5.04 4.00 2.50 4.02 5.39 3.31 3.73 1.11 5.09 0.94 4.48 4.07 5.38 4.17 5.85 4.63 3.92 2.69 1.02 3.88 0.95 1.57 2.35 (d) Based on your answers to parts (a)-(c), what can you say about the relative performance of the distance to kth-nearest neighbor and LOF algorithms?

Answer 1

. dnetioral dolasel shoon menhonol の 1f) axe belou eloto These ace刊 clefined Clustess 回. The points sich ase masted th th Squate as the outhes they lie ve's CaLise ch ase outie ase t lence, above

b Given neighho 四,15 s).01 0.42 1-13 16 0.S6 1.08 3.04 4 03 S 3 3.31 3-73 1-111, S09 07 S 38 4.17 S.8S 63 3.42 0.38 0-72 0.90 wing neverl a bove table. Points to the P toe .See the chet ance ,he distance 9.ol & [./.3 있 maximum past on toth neaset nei