Real-time supervised learning with Weka (on Android phone)

Question

I want to use accelerometer data to (try) make predictions what activities the user is doing (simple activities). Let's say I have a bunch of training instances, where is a single training instance and xn is the class label. After training, I want to take in the data, transform it, and then output the classification of the activity in real-time (or close to it).

First, any suggestions? Second, I will have class labels for the training set but not the test set. How should I calculate accuracy. I won't just be able to look at the label because the test set doesn't have labels. Lastly, I just want to make sure that Weka won't complain if the test set doesn't have any class labels.

I was leaning to using supervised learning, but I could be argued out of it.

Answer 1

It seems like a supervised ML problem. And if you want to use Weka GUI, yous have to label your test data. And then pass them into weka and see what is the classification result from Weka. The label you preset to your test data will not influence your result.

If you do not know what are the test data should be, that will be a unsupervised ML problem. Since supervised problem, for training model and evaluate it(N-fold crossvalidation), you have to know the ground truth.