Please enable JavaScript.
Coggle requires JavaScript to display documents.
Chapter 7: Decision Analytic Thinking I (The Confusion Matrix (False…
Chapter 7: Decision Analytic Thinking I
Evaluating Classifiers
classification model takes an instance for which we do not know the class and predicts its class
Binary
often called positive/negative
bad outcome is a "positive" example
normal or good outcome is a negative example
positive examples are worthy of attention or
alarm
negative example = uninteresting or benign
Plain accuracy and it problems
classification accuracy is popular cause its easy to measure
accuracy is a common evaluation metric - because it reduces classifier performance to a single number
The Confusion Matrix
one sort of contingency table
confusion matrix separates decisions made by classifier, making explicit how one class is being confused for another
False Positives
Negatives classified as positive
False Negatives
Positives classified as negative
Problems with Unbalanced Classes
skewed/unbalanced because such a small number of unusual ones in a large population
Problems with Unequal Costs and Benefits
makes no distinction between false positive and false negative errors
Generalizing Beyond Classification
when we are applying data science to actual application it is vital to ask what is important in the application? what is the goal? are we assessing the results of data mining appropriately?
Expected Value
decomposes data-anlaytic thinking into
structure of the problem
elements of the analysis that can be extracted from the data
element of the analysis that need to be acquired from other sources
The weighted average of values of the different possible outcomes, where the weight given to each value is its probability of occurence