The model usage time of k-nearest neighbor (kNN) method is sensitive to the dataset size (i.e., amount of training/stored data). When the dataset is large, the model usage time could be prohibitively long. A machine learning scientist suggests that the kNN's model usage time can be reduced by approximately 50% if you can separate the original data into two partitions. Such strategy can be recursively applied so that the time can be further reduced by approximately a half again. Comment on the effectiveness (classification accuracy) and efficiency (classification time including model construction time and model usage time) of this strategy. Note that this is an open question and make your own assumption