I have implemented some clustering algorithms. As the data is high dimensional, I am not able to plot it. And thus, I am not really aware of the validation techniques for clustering algorithm.
My goal is to know whether or not the algorithm gives correct output
Hi Madhav,
Although evaluating the cluster algorithm is a critical task (it is unsupervised learning), you can still do such evaluation by either converting the cluster problem into a classification problem and use the associated evaluation metrics or use a clusterer-based density approach, where the log-likelihood can be utilized for the sake of evaluation.
HTH.
Samer Sarsam, PhD.
Hai Madhav,
It's a nice question. You can use PCA to reduce the dimension. Of course, you will lose some information of your original data.
I'm wondering that it should be easy task. But, the validation output of clustering problem is not good enough in result. You can test it by using real classification data and run the clustering algorithm. This methods is usually implemented in unlabel data so that it's very difficult to know the corect output. If you wanna know the accuracy of your clustering algorithm, make sure that you have the real label of your data. Note that every case may be a unique to be solved.
Best Regards,
- Badges
- Science topic
- Similar topics
- Correction