Abstract
More and more applications have come with large data sets in the past decade. However, existing algorithms cannot guarantee to scale well on large data. Averaged n-Dependence Estimators (AnDE) allows for flexible learning from out-of-core data, by varying the value of n (number of super parents). Hence, AnDE is especially appropriate for large data learning. In this paper, we propose a sample-based attribute selection technique for AnDE. It needs one more pass through the training data, in which a multitude of approximate AnDE models are built and efficiently assessed by leave-one-out cross validation. The use of a sample reduces the training time. Experiments on 15 large data sets demonstrate that the proposed technique significantly reduces AnDE's error at the cost of a modest increase in training time. This efficient and scalable out-of-core approach delivers superior or comparable performance to typical in-core Bayesian network classifiers.
| Original language | English |
|---|---|
| Article number | 7565579 |
| Pages (from-to) | 172-185 |
| Number of pages | 14 |
| Journal | IEEE Transactions on Knowledge and Data Engineering |
| Volume | 29 |
| Issue number | 1 |
| DOIs | |
| Publication status | Published - 1 Jan 2017 |
Keywords
- Bayesian network classifiers
- Large data
- Classification learning
- Attribute selection
- Averaged n-dependence estimators (AnDE)
- Leave-one-out cross validation
Cite this
}
Sample-based attribute selective An DE for large data. / Chen, Shenglei; Martinez, Ana M.; Webb, Geoffrey I.; Wang, Limin.
In: IEEE Transactions on Knowledge and Data Engineering, Vol. 29, No. 1, 7565579, 01.01.2017, p. 172-185.Research output: Contribution to journal › Article › Research › peer-review
TY - JOUR
T1 - Sample-based attribute selective An DE for large data
AU - Chen, Shenglei
AU - Martinez, Ana M.
AU - Webb, Geoffrey I.
AU - Wang, Limin
PY - 2017/1/1
Y1 - 2017/1/1
N2 - More and more applications have come with large data sets in the past decade. However, existing algorithms cannot guarantee to scale well on large data. Averaged n-Dependence Estimators (AnDE) allows for flexible learning from out-of-core data, by varying the value of n (number of super parents). Hence, AnDE is especially appropriate for large data learning. In this paper, we propose a sample-based attribute selection technique for AnDE. It needs one more pass through the training data, in which a multitude of approximate AnDE models are built and efficiently assessed by leave-one-out cross validation. The use of a sample reduces the training time. Experiments on 15 large data sets demonstrate that the proposed technique significantly reduces AnDE's error at the cost of a modest increase in training time. This efficient and scalable out-of-core approach delivers superior or comparable performance to typical in-core Bayesian network classifiers.
AB - More and more applications have come with large data sets in the past decade. However, existing algorithms cannot guarantee to scale well on large data. Averaged n-Dependence Estimators (AnDE) allows for flexible learning from out-of-core data, by varying the value of n (number of super parents). Hence, AnDE is especially appropriate for large data learning. In this paper, we propose a sample-based attribute selection technique for AnDE. It needs one more pass through the training data, in which a multitude of approximate AnDE models are built and efficiently assessed by leave-one-out cross validation. The use of a sample reduces the training time. Experiments on 15 large data sets demonstrate that the proposed technique significantly reduces AnDE's error at the cost of a modest increase in training time. This efficient and scalable out-of-core approach delivers superior or comparable performance to typical in-core Bayesian network classifiers.
KW - Bayesian network classifiers
KW - Large data
KW - Classification learning
KW - Attribute selection
KW - Averaged n-dependence estimators (AnDE)
KW - Leave-one-out cross validation
UR - http://www.scopus.com/inward/record.url?scp=85027493553&partnerID=8YFLogxK
U2 - 10.1109/TKDE.2016.2608881
DO - 10.1109/TKDE.2016.2608881
M3 - Article
VL - 29
SP - 172
EP - 185
JO - IEEE Transactions on Knowledge and Data Engineering
JF - IEEE Transactions on Knowledge and Data Engineering
SN - 1041-4347
IS - 1
M1 - 7565579
ER -