Mobile Information Systems / 2016 / Article / Tab 7

Research Article

A Mobile Application for Easy Design and Testing of Algorithms to Monitor Physical Activity in the Workplace

Table 7

Deafult parameters for base classifiers: -NN, NN, and DT.

Classifier Parameters

-NN Search algorithm: linear search with Euclidean distance
Number of neighbors: 25

NN Number of hidden layers: (# attributes + # classes)/2
Learning rate: 0.3
Momentum: 0.2
Maximum number of epochs to train through: 500
Attributes normalized between −1 and 1

DT (REPTree) No depth restrictions
Minimum total weight of the instances in a leaf: 2
Minimum variance on all the data needed for splitting: 0.01
Pruning
Number of folds used for pruning: 3