## Combining Instance-Based and Model-Based Learning (1993)

Citations: | 115 - 0 self |

### BibTeX

@INPROCEEDINGS{Quinlan93combininginstance-based,

author = {J. R. Quinlan},

title = {Combining Instance-Based and Model-Based Learning},

booktitle = {},

year = {1993},

pages = {236--243},

publisher = {Morgan Kaufmann}

}

### Years of Citing Articles

### OpenURL

### Abstract

This paper concerns learning tasks that require the prediction of a continuous value rather than a discrete class. A general method is presented that allows predictions to use both instance-based and model-based learning. Results with three approaches to constructing models and with eight datasets demonstrate improvements due to the composite method.

### Citations

3909 |
Classification and Regression Trees
- Breiman, Friedman, et al.
- 1984
(Show Context)
Citation Context ...af of a decision tree contains just a class name, the corresponding leaf of a model tree is a linear model relating the class values of the training cases to their attribute values. Regression trees [=-=Breiman et al, 1984-=-] are based on a similar divide-and-conquer strategy, but have values rather than linear models at the leaves. Consider a set T of training cases for which a model tree is to be constructed. Unless T ... |

1055 | Instance-based learning algorithms - Aha, Kibler, et al. - 1991 |

777 |
C4.5: Programs for
- Quinlan
- 1993
(Show Context)
Citation Context ...pwlinearsdomain, the use of local information might introduce additional error. I have also experimented with a similar composite approach in which the model attempts to predict differences directly [=-=Quinlan, 1993-=-b]. For domains in which such a difference model can be found, including servo and lhrh-def among the present datasets, this approach produces predictors that are better still. Acknowledgements This r... |

222 | Learning with continuous classes
- Quinlan
- 1992
(Show Context)
Citation Context ...proposed. This method, which can be used with any form of model, is illustrated here using three of them -- familiar linear regression [Press, Flannery, Teukolsky, and Vetterling, 1988], model trees [=-=Quinlan, 1992-=-], and neural networks [Hinton, 1986; McClelland and Rumelhart, 1988; Hinton, 1992]. Over a representative collection of datasets, the composite methods often produce better predictions than either in... |

173 |
Learning distributed representations of concepts
- Hinton
- 1986
(Show Context)
Citation Context ...sed with any form of model, is illustrated here using three of them -- familiar linear regression [Press, Flannery, Teukolsky, and Vetterling, 1988], model trees [Quinlan, 1992], and neural networks [=-=Hinton, 1986-=-; McClelland and Rumelhart, 1988; Hinton, 1992]. Over a representative collection of datasets, the composite methods often produce better predictions than either instance-based or model-based approach... |

37 |
Instance-based prediction of real-valued attributes
- Kibler, Aha, et al.
- 1989
(Show Context)
Citation Context ...aradigm, three different forms of model-based learning, and eight learning tasks. Short descriptions of each follow. 3.1 INSTANCE-BASED LEARNING The instance-based approach used here is based on ibl [=-=Kibler, Aha, and Albert, 1988-=-]. Prototypes consist of all training cases without any modification. The dissimilarity of two cases is computed by summing their normalized differences over each attribute, defined as ffl for discret... |

25 |
How neural networks learn from experience
- Hinton
- 1992
(Show Context)
Citation Context ...e using three of them -- familiar linear regression [Press, Flannery, Teukolsky, and Vetterling, 1988], model trees [Quinlan, 1992], and neural networks [Hinton, 1986; McClelland and Rumelhart, 1988; =-=Hinton, 1992-=-]. Over a representative collection of datasets, the composite methods often produce better predictions than either instance-based or model-based approaches. 2 USING MODELS AND INSTANCES We assume som... |

19 | Attributes of the performance of central processing units: a relative performance prediction model - Ein-Dor, Feldmesser - 1987 |

4 | A case study in machine learning
- Quinlan
- 1993
(Show Context)
Citation Context ...pwlinearsdomain, the use of local information might introduce additional error. I have also experimented with a similar composite approach in which the model attempts to predict differences directly [=-=Quinlan, 1993-=-b]. For domains in which such a difference model can be found, including servo and lhrh-def among the present datasets, this approach produces predictors that are better still. Acknowledgements This r... |