# Tag Info

21

The problem addressed by One Class SVM, as the documentation says, is novelty detection. The original paper describing how to use SVMs for this task is "Support Vector Method for Novelty Detection". The idea of novelty detection is to detect rare events, i.e. events that happen rarely, and hence, of which you have very little samples. The problem is then, ...

6

I will assume you understand how a standard SVM works. To summarise, it separates two classes using a hyperplane with the largest possible margin. One-Class SVM is similar, but instead of using a hyperplane to separate two classes of instances, it uses a hypersphere to encompass all of the instances. Now think of the "margin" as referring to the outside of ...

2

According to the paper One-class document classification via Neural Networks of Manevitz and Yousef it seems to be possible to construct a one-class Naive Bayes classifier, even without a standard deviation. I cite the relevant passage where the authors mention how to implement the core of the classifier: We calculate $p(d|E)$ as the product of $p(w|E)$ for ...

2

The C parameter is used for multiple classification. The nu parameter is used in one-class classification, so as long as you optimize nu you are fine. nu is the upper bound in the fraction of training points. Please check this link Depending on the kernel function you choose, you will have to optimize gamma too. The hyperplane varies depending on both ...

2

This has, of course, been tried before. In fact, there is an implementation in ELKI, based on k-means. The problem is, this isn't as easy as it sounds. k-means itself is sensitive to outliers. the points inbetween clusters aren't necessarily outliers, but may have a high score the points away from all clusters can be more easily found by computing the ...

1

If your entire data only has the "0" class, then life is easy: just classify everything as "0". Any tool or method will do so, too. (If a method, upon seeing only "0" instances, classifies something as "rhubarb", I would question its sanity.) If you classify everything as "0", and everything is in fact "0", then every instance is a true positive. There are ...

1

nu and C are alternative constants from alternative formulations of the SVM. You have to choose one or the other. If you are using the svm function of e1071 please check the type parameter in http://www.rdocumentation.org/packages/e1071/functions/svm type="C-classification" which is the default will use the C-SVM formulation which requires the C parameter, ...

1

I used JMP. I know the temp spike is an outlier. Approach: normalize the variables because the magnitudes make relative error less meaningful. make 3 lagged variables (you have a very high sample rate). First is one-row lag of sensor, second is 10-row lag, third is 100-row lag. exclude the known bad rows (~325 of 14857 rows) use NN tool with 5 ...

1

I think you can see the nature of the problem if you calculate how many "strange" data points you need, given that you've observed a long run of "not-strange" data points (using the language of Ho and Wechsler here). Let's do some back-of-the-envelope calculations: for a stream of $k$ data points with constant $p_i = p$ then $M_k \sim 10^{k(\text{log}_{10}(\... 1 Your implementation is correct. The power martingale tends to get very small (closer to 0) when p-values are uniformly distributed. To avoid that, you just need to restart your martingale from 1 as soon as it gets smaller than 1. So just add: Mtest[i] = max(Mtest[i], 1) This will keep your martingale small (but not less than 1) when the p-values are ... 1 Let's first clarify the terminology a little. The data you have, according to the description, is a multi-class data rather than a multi-label one. In both cases, the number of possible classes/labels of the data set is equal or larger than 2, i.e.,$\lvert Y \rvert \ge 2$. The difference is, in a multi-class data set, one instance$\mathbf{x}\$ is associated ...

Only top voted, non community-wiki answers of a minimum length are eligible