Page 86 - Data Science Algorithms in a Week
P. 86
Decision Trees
4. Here the decision tree algorithm may not perform that well without any
processing of the data. If we considered every class of a temperature, then 25
degrees Celsius would still not occur in the decision tree as it is not in the input
data, so we would not be able to classify how Mary would feel at 16 degrees
Celsius and at 3km/h windspeed.
We could alternatively divide the temperature and wind speed into the
intervals in order to reduce the classes, so that the resulting decision tree
could classify the input instance. But it is this division, the classification of in
what intervals 25 degrees Celsius and 3km/h should be, that is the
fundamental part of the analysis procedure for this type of problem. Thus
decision trees without any serious modification could not analyze the
problem well.
[ 74 ]
