My goal is to get a smartphone names from Twitter. So this is what I followed:

1- I extracted 100K tweets using the keyword “smartphone”.

2- I Applied LDA after applying ngram tokenization and cleaning.So, I got noisy results such as:(giveway, international, apple, iphone_6, samsung_s5,news…)

3- I filtered the results using a smartphone list (iphone_6, samsung_s5,iphone_4s,…) extracted from DBpedia in order to remove the noise.

Is what I did supervised or unsupervised Machine learning?

  • $\begingroup$ Yes I know that LDA is unsupervised. However I wanna know if my add-on cleaning filter makes my methodology as supervised? If no, I wanna know the supervised solutions in my case. Thanks. $\endgroup$ Jul 15 '16 at 10:35
  • $\begingroup$ Thanks for clarifying. I've added a few edits to help direct @Hima Varsha original answer to your remaining question. Thanks! $\endgroup$
    – AN6U5
    Jul 15 '16 at 16:41

The basic idea behind supervised and unsupervised learning is: In supervised learning, one possesses a training data set that includes the desired target output feature. Examples of popular supervised learning algorithms are linear regression, and support vector classification etc. In unsupervised learning, we do not possess a training data set with the target output feature i.e. we do not know what our output or results look like. An example is clustering. There are really good explanations out there about this.

LDA is an unsupervised learning algorithm and the process you described can be classified as unsupervised learning. The filtering step that you describe does not make the algorithm supervised because the target smartphones have not been directly correlated to the training data and therefore is only serving as a guide to restrict the cardinality of the target variable.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.