abhisaarsharma
commented
8 years ago Hi, Yeah you are right, I tried these methods before with similar results. The reasons I can think of is that they do not have any weight setting, since the rr points are less they should be given more weight but I don't see any such option in elastic nets. I was getting the same results for svm but there is a setting 'balanced weight' that assigns weight inversely proportional to the frequency, after which the results were usable.
The other reason might be the data is simply not seperaperble linearly in the 4d space. On Dec 7, 2015 2:01 PM, "dcockerham" notifications@github.com wrote:
I've been playing around with these methods and trying to get them to produce usable results, but with zero success. As best I can figure, they simply aren't suited to a classification problem like this. After fitting, they all seem to end up predicting a zero for any input (this does, amusingly, give them good accuracy due to the sparseness of RR Lyrae stars, but is obviously a useless result).
— Reply to this email directly or view it on GitHub https://github.com/NobleKennamer/astro_porject/issues/19.
I've been playing around with these methods and trying to get them to produce usable results, but with zero success. As best I can figure, they simply aren't suited to a classification problem like this. After fitting, they all seem to end up predicting a zero for any input (this does, amusingly, give them good accuracy due to the sparseness of RR Lyrae stars, but is obviously a useless result).