Classification_with_decision_tree

Who to use ?

• to compile

gcc *.c -Wall -O3 -fopenmp -lm

• to run

./a.out

Features

Decision tree for classification

• geni and entropy as split criteria

• Dataset_bagging and feature_bagging

• load a train and test tree from a file

Random forest with :

• geni and entropy as split criteria

• Dataset_bagging and feature_bagging

Results

MNIST

Dataset : 60000 instances, 784 features, 10 classes
Dataset : 10000 instances, 784 features, 10 classes
Subproblem : 60000 instances, 784 features, 10 classes
RandomForest : 20 trees, 10 classes
treeCount = 20, nodeCount = 114392
train = 0.994, test = 0.943
Elapsed time: 62.921 seconds

Pendigits

Dataset : 7494 instances, 16 features, 10 classes
Dataset : 3498 instances, 16 features, 10 classes
Subproblem : 7494 instances, 16 features, 10 classes
RandomForest : 20 trees, 10 classes
treeCount = 20, nodeCount = 14786
train = 0.996, test = 0.953
Elapsed time: 0.385 seconds

Fashion

Dataset : 60000 instances, 784 features, 10 classes
Dataset : 10000 instances, 784 features, 10 classes
Subproblem : 60000 instances, 784 features, 10 classes
RandomForest : 20 trees, 10 classes
treeCount = 20, nodeCount = 187722
train = 0.984, test = 0.850
Elapsed time: 242.810 seconds

Letter

WARNING

PTHREADS make some dataleaks... 288 bytes in 1 blocks are possibly lost

Authors

@fareanor3 @Solp1