analytics final

prediction error

prediction - error

regression trees

decision trees (numeric outcome)

classification trees

decision trees (class outcome)

most impure Gini Index

0.5

most impure Entrophy index

1

classification tree example

majority vote prediction

regression tree example

prediction average

KNN needs to be

normalized/standardized

why is K picked in KNN

to minimize the root mean squared error

timely predicitons

KNN is not good for

recall + formula

tp/obs default (up and down division)

precision + formula

tp/total obs predicted (left to right)

TP & FP increase & FN & TN decrease

when reducing a threshold

accuracy rate formula

tp + tn / total number of observations

recall interpretation

70% of the observed fraud transactions were correctly classified by the model

precision interpretation

given that the model predicted fraud there is a % probability that the transaction is fraud

pure

entrophy or gini index value of 0 indicates that all observations are pure

unsupervised

KNN learning method classification

unparametric modeling method

KNN regression modeling method

training & test data set

what does an analytical file need to be split into when building a KNN model

classification tree

once a binary feature is selected it cannot be selected again