MATLAB: Imbalanced data classification with boosting algorithm

classification

I am working on a binary data classification problem. The dataset is imbalanced, it consists of 92% 'false' labels and 8% 'true' labels. The number of features is 18 and I have a small number of 650 data points. I want to use boosting algorithms in matlab like 'GentleBoost' to solve this problem. I assign uniform for prior as follows:

ada = fitensemble(Xtrain,Ytrain,'GentleBoost',10,'Tree','LearnRate',0.1, 'prior', 'uniform');

but the performance is consistently poor. How should I set the parameters? Is it necessary to set a cost? How can I do this?Is there any classifier that perform better than this?

Best Answer

I think it is very difficult to assess this, because it is possible that your dataset is simply difficult to classify. (For example, maybe it is actually almost random, independent of the features.)

My suggestion would be to create an artificial dataset in which you know that the features determine the response. Test the syntax with that dataset, to ensure that you are coding it correctly.

Related Solutions

MATLAB: Can someone tell me about the algorithm used to obtain the prediction of the bagged models in Ensemble bagged tree classification in Matlab

In general ensemble classifiers combine a different variety of weak classifiers to reduce high variance. So in general the weak leraners could be any classification models, from logistic regression to nueral nets and could be using any optimization techniques to reduce loss. But in an Ensemble Bagged tree ( I think you are talking about this https://www.mathworks.com/help/stats/treebagger-class.html ) classifier the models would be weak decision trees in which a bootstrap data is drawn for each tree. Decision trees ususally use gini index or entrophy to decide their next branch. Adaboost can also be applied to any ensemble classifier. So in an Ensemble Bagged tree classification all models would be decision trees with varying data due to bagging.

MATLAB: Handling imbalanced data with patternnet

Hi,

To weight the errors during training, one can specify the "error weights" (EW) property in 'train', see here:

https://www.mathworks.com/help/deeplearning/ref/train.html#d117e35592

At prediction time, one can pass the error weights to the performance function, either by calling 'perform' with 'ew' or 'crossentropy' with 'perfWeights'.

For example (to weight the classes in the Iris dataset by 0.1, 0.2 and 0.7):

[x,t] = iris_dataset;
net = patternnet(10);
net = train(net,x,t,[],[],[0.1,0.2,0.7]');
y = net(x);
perf = perform(net,t,y,[0.1,0.2,0.7]')

Or:

perf = crossentropy(net,t,y,[0.1,0.2,0.7]')

Hope this helps!

Best Answer

Related Solutions

MATLAB: Can someone tell me about the algorithm used to obtain the prediction of the bagged models in Ensemble bagged tree classification in Matlab

MATLAB: Handling imbalanced data with patternnet

Related Question