Gama, J. (1998) Combining Classifiers by Constructive Induction. In Proceedings of the Ninth European Conference on Machine Learning.  Home/Search   Document Details and Download   Summary   Related Articles   Check

....re weighting of the misclassified training examples in boosting. Techniques for combining the predictions obtained from the multiple base level classifiers can be clustered in three combining frameworks: voting (used in bagging and boosting) stacked generalization or stacking [11] and cascading [4]. In voting, each baselevel classifier gives a vote for its prediction. The prediction receiving the most votes is the final prediction. In stacking, a learning algorithm is used to learn how to combine the predictions of the baselevel classifiers. The induced meta level classifier is then used to ....

Gama, J. (1998) Combining Classifiers by Constructive Induction. In Proceedings of the Ninth European Conference on Machine Learning.

....re weighting of the misclassified training examples in boosting. Techniques for combining the predictions obtained from the multiple baselevel classifiers can be clustered in three combining frameworks: voting (used in bagging and boosting) stacked generalization or stacking [14] and cascading [5]. In voting, each base level classifier gives a vote for its prediction. The prediction receiving the most votes is the final prediction. In stacking, a learning algorithm is used to learn how to combine the predictions of the base level classifiers. The induced meta level classifier is then used ....

Gama, J. (1998) Combining Classifiers by Constructive Induction. In Proceedings of the Ninth European Conference on Machine Learning.

....of landmarking amounts to investigating how well a landmark learner s performance hints at the location of the learning tasks in the expertise map. 2 i4 i1 i5 i6 i2 i7 i3 Figure 1: Example of map of areas of expertise 1 Work on stacked generalisation and related developments (e.g. see [17, 8]) served as an inspiration for the idea of landmarking. There, learners are invoked to redescribe the original input data for learning purposes. Here, the learners performances are used to describe properties of the dataset. 2 Arguably, different performance measures (predictive accuracy, ....

J. Gama. Combining classifiers by constructive induction. In Proceedings of the 10th European Conference on Machine Learning, pages 178--189, 1998.

....of the final classifier is the language used by the high level generalizer. This language uses terms that are expressions from the language of low level classifiers. In this sense, Cascade Generalization generates a unified theory from the base theories. Here we extend the work presented in (Gama 1998) by applying Cascade locally. In our implementation, Local Cascade Generalization generates a decision tree. The experimental study shows that this methodology usually improves both accuracy and theory size at statistically significant levels. In the next Section we review previous work in the ....

....We have also evaluated Stacked Generalization using C4.5 at top level. The version that we have used is somewhat better. Using C4.5 at top level the average mean of the error rate is 15.14. 3. This heuristic was suggested by Breiman et al. 4. Two different methods are presented in Ting (1997) and Gama (1998). 5. We have preferred C5.0Boosting (instead of Bagging) because it is available for us and allows cross checking of the results. There are some differences between our results and those previous published by Quinlan. We think that this may be due to the different methods used to estimate the ....

J. Gama. Combining classifiers by constructive induction. In C. Nedellec and C. Rouveirol, editors, Machine Learning ECML-98. Springer Verlag, 1998.

....of looking for methods that fit the data using a single representation language, we present a family of algorithms, under the generic name of Cascade Generalization, whose search space contains models that use different representation languages. Cascade generalization was first presented in [14]. It performs an iterative composition of classifiers. At each iteration a classifier is generated. The input space is extended by the addition of new attributes. These are in the form of a probability class distribution which are obtained, for each example, by the generated base classifier. The ....

....of the final classifier is the language used by the high level generalizer. This language uses terms that are expressions from the language of low level classifiers. In this sense, Cascade Generalization generates a unified theory from the base theories. Here we extend the work presented in [14], by applying Cascade locally. In our implementation, Local Cascade Generalization generates a decision tree. The experimental study shows that this methodology usually improves both accuracy and theory size with statistical significance levels. The next section presents the framework of Cascade ....

[Article contains additional citation context not shown here]

J. Gama. Combining classifiers by constructive induction. In C. Nedellec and C. Rouveirol, editors, Machine Learning ECML-98. Springer Verlag, 1998.

....of looking for methods that fit the data using a single representation language, we present a family of algorithms, under the generic name of Cascade Generalization, whose search space contains models that use different representation languages. Cascade generalization was first presented in [13]. It performs an iterative composition of classifiers. At each iteration a classifier is generated. The input space is extended by the addition of new attributes. Those new attributes are obtained in the form of a probability class distribution given, for each example, by the generated base ....

....The language of the final classifier is the language used by the high level generalizer. But it uses terms that are expressions from the language of low level classifiers. In this sense, Cascade Generalization generates a unified theory from the base theories. Here we extend the work presented in [13], by applying Cascade locally. In our implementation, Local Cascade Generalization generates a decision tree. The experimental study shows that this methodology usually improves both accuracy and theory size with statistical significance levels. The next section presents the framework of cascade ....

[Article contains additional citation context not shown here]

J. Gama. Combining classifiers by constructive induction. In C. Nedellec and C. Rouveirol, editors, Machine Learning ECML-98. 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