Using this package
Two Simple Steps:
1. Build the model on training data
2. Test the model on test data
Step 1:
The function you need to call is aode.
It accepts a lot of parameters.
aode(train, mestimate = 1, weighted = FALSE,subsumption = FALSE, S = 100)
train - training data , should be a data frame. AODE works only discretized data. It would be better to discreetize the data frame before passing it to this function.
However, aode discretizes the data if not done before hand. It uses an R package called discretization for the purpose. It uses the well known MDL discretization technique.
(It might fail sometimes)
mestimate(optional) - the constant to be used for m-estimation. This is used for smoothening of probabilities.
weighted(optional) - TRUE for a weighted aode , FALSE for simple aode.
subsumption(optional) - TRUE for doing subsumtion resolution , FALSE for ignoring subsumption resolution.
S(optional) - subsumtion constant. Default 100. Will be considered only if subsumtion is set TRUE.
return value : an object of class AODE
Step 2:
predict(object,test)
object - object returned by aode function above.
test - testing data frame. If the training data was discretized, then the same cut points shall be used to
discretize the test data.
comments :
predict is a generic function. It is generally used in all machine learning packages in the same way.
It is therefore compatible and simple to use.
If interested in this method , here is some information about this algorithm
AODE achieves highly accurate classification by averaging over all of a small space
of alternative naive-Bayes-like models that have weaker (and hence less detrimental)
independence assumptions than naive Bayes. The resulting algorithm is computationally
efficient while delivering highly accurate classification on many learning tasks.
For more information, see<br/>
G. Webb, J. Boughton, Z. Wang (2005).
Not So Naive Bayes: Aggregating One-Dependence Estimators. Machine Learning. 58(1):5-24.
Further papers are available at<br/> http://www.csse.monash.edu.au/~webb/.
Default frequency limit set to 1.
BibTeX:
article{Webb2005,
* author = {G. Webb and J. Boughton and Z. Wang},
* journal = {Machine Learning},
* number = {1},
* pages = {5-24},
* title = {Not So Naive Bayes: Aggregating One-Dependence Estimators},
* volume = {58},
* year = {2005}
* }
* Specify critical value of specialization-generalization for
* Subsumption Resolution (default is false).
* Results in lowering bias and increasing variance of classification.
* Recommended for large training data.
* See: in proceedings{Zheng2006,
* author = {Fei Zheng and Geoffrey I. Webb},
* booktitle = {Proceedings of the Twenty-third International Conference on Machine Learning (ICML 2006)},
* pages = {1113-1120},
* publisher = {ACM Press},
* title = {Efficient Lazy Elimination for Averaged-One Dependence Estimators},
* year = {2006},
* ISBN = {1-59593-383-2}
* }
*
-W (Optional)
* Weighted AODE. Uses mutual information between attribute and the class as weight of
* each SPODE (default is false).
* Results in lowering bias and increasing variance of classification.
* Recommended for large training data.
* Can not use weighting for A1DEUpdateable classifier.
* See:
* in proceedings{Jiang2006,
* author = {L. Jiang and H. Zhang},
* booktitle = {Proceedings of the 9th Biennial Pacific Rim International Conference on Artificial Intelligence, PRICAI 2006},
* pages = {970-974},
* series = {LNAI},
* title = {Weightily Averaged One-Dependence Estimators},
* volume = {4099},
* year = {2006}
* }