Kohavi R. and John G., The Wrapper Approach, In Feature Extraction, Construction and Selection: A Data Mining Perspective, H. Liu and H. Motoda (eds.), Kluwer Academic Publishers, 1998.  Home/Search   Document Details and Download   Summary   Related Articles   Check

....to be more accurate than its constituent members depends on the diversity among its members [8] In this work we consider two approaches to feature selection. We consider a situation where we select the first n features based on one of the ranking criteria. We also consider a wrapper like [11] forward sequential search that takes a ranked set of feature as starting point. A key issue in a forward sequential search is the order in which to test the attributes. It is important to start with the more promising attributes. In the following, we present the ranking algorithms we applied. ....

R. Kohavi, G.H. John, "The Wrapper Approach", in Feature Selection for Knowledge Discovery and Data Mining, H. Liu & H. Motoda (Ed.), Kluwer, 33-50, 1998.

....is to analyze the impact of representation changes on learning. The main question is related to the choice of one operator and its parameters. In Machine Learning, the abundant literature on feature selection shows that approaches fall in two broad categories: the wrapper and the filter approach [20]. Intuitively, the wrapper approach uses the performance of the learning algorithm as a heuristic to guide the abstraction. In the following, we present how the wrapper approach can be used to choose the most tted abstraction. As it is an approach that attempt to learn from the learning process ....

R. Kohavi, G. John, The wrapper approach, in: Feature Selection for Knowledge Discovery and Data Mining, H. Liu and H. Motoda (eds.), Kluwer Academic Publishers, pp33-50., 1998.

....we have combined in the PLIC system two approaches, one based on a priori consideratyions, and one using the learning results. From a Machine Learning point of view, this 16 architecture corresponds to the combination of the two widely used approaches in feature selection: the Wrapper model [23] and the Filter model [24] The set of experiments that have been conducted show that both operators do impact on the learning accuracy. It is interesting to notice that the best resolution and structure (sort of coordinates in the abstract space) found by the system depends of the concept. Since ....

Kohavi, R., John, G.: The wrapper approach. In: Feature Selection for Knowledge Discovery and Data Mining, H. Liu and H. Motoda (eds.), Kluwer Academic Publishers, pp33-50. (1998)

.... the training set to calibrate (tune) the algorithm, run the calibrated versions of the algorithm on functions (problem instances) from the test set, report the outcomes of the experiments on the test set, This kind of techniques are widely used in machine learning and are called wrapping [7]. The point is that the evaluation is done on the test set and not the training set, which is not common practice in the EA community. As a result of the application of these techniques we gain much more reliable indicators of a given algorithm on a given class. This idea can be generalized to ....

Ron Kohavi and George H. John, "The wrapper approach," in Feature Extraction, Construction and Selection: A Data Mining Perspective, Huan Liu and Hiroshi Motoda, Eds., vol. 453 of The Kluwer International Series in Engineering and Computer Science. Kluwer, 1998.

....of features is involved but needs to be monitored for problems with smaller numbers of features say less than 25. 1. Introduction Recent research supports the view that the Wrapper approach to feature subset selection produces better results than alternative approaches (Aha Bankert, 1994; Kohavi John, 1998). In the Wrapper approach the induction algorithm for which the feature subset is required is itself the evaluation mechanism in the feature selection process the induction algorithm is wrapped in the search process. This has the obvious advantage that the inductive bias of the selection ....

....inductive bias of the selection process is the same as that of the target induction process. This advantage comes at a considerable computational cost since the induction mechanism must be built and tested at each step in the search process. In advocating that this approach to feature selection Kohavi and John (1998) introduce the caveat the Wrapper approach may overfit the data used in the evaluation mechanism. They say that this is less likely to be a problem when there is plenty of data available for use in evaluation. This may be a considerable problem in practice since in many realworld scenarios there ....

[Article contains additional citation context not shown here]

Kohavi, R. & John, G.H., (1998) The Wrapper Approach, in Feature Selection for Knowledge Discovery and Data Mining, H. Liu & H. Motoda (eds.), Kluwer Academic Publishers, pp33-50.

No context found.

Kohavi R. and John G., The Wrapper Approach, In Feature Extraction, Construction and Selection: A Data Mining Perspective, H. Liu and H. Motoda (eds.), Kluwer Academic Publishers, 1998.

No context found.

Kohavi, R., & John, G. H. (1998). The wrapper approach.

No context found.

R. Kohavi and G. H. John. The wrapper approach. In H. Liu and H. Motoda, editors, Feature Extraction, Construction, and Selection: A Data Mining Perspective, pages 33--50. Kluwer, 1998.

No context found.

R. Kohavi and G. John. The wrapper approach. In Feature Selection for Knowledge Discovery and Data Mining, H. Liu and H. Motoda (eds.), Kluwer Academic Publishers, pp33-50., 1998.

No context found.

R. Kohavi and G. John. The wrapper approach. In Feature Selection for Knowledge Discovery and Data Mining, H. Liu and H. Motoda (eds.), Kluwer Academic Publishers, pp3350. , 1998.

No context found.

R. Kohavi and G. John. The wrapper approach. In H. Liu and H. Motoda, editors, Feature Extraction, Construction and Selection: A Data Mining Perspective. Springer Verlag, 1998.

Online articles have much greater impact   More about CiteSeer.IST   Add search form to your site   Submit documents   Feedback  

CiteSeer.IST - Copyright Penn State and NEC