Harvest the classification tree

Description

Aimed at applying the Harvest classification tree algorithm, modified algorithm of classic classification tree.The harvested tree has advantage of deleting redundant rules in trees, leading to a simplify and more efficient tree model.It was firstly used in drug discovery field, but it also performs well in other kinds of data, especially when the region of a class is disconnected. This package also improves the basic harvest classification tree algorithm by extending the field of data of algorithm to both continuous and categorical variables.

To learn more about the harvest classification tree algorithm, you can go to http://www.stat.ubc.ca/Research/TechReports/techreports/220.pdf for more information.

Details

The main function of package called 'harvest', it can be used to analyze the data which is stored in a data frame, where first column stores the class of response data, and the second to last column stores explantory variables accordingly.The 'predict' funciton offers function to predict the unclassified data based on training model. The 'harfunc' function is the fundemental part of 'harvest', which can be used to analyze the data which has already been classified by rpart function(traditional classification tree). Please check the help file of these three functions for more information.

Author(s)

Bingyuan Liu \ Yan Yuan \ Qian Shi

Maintainer: Bingyuan Liu<adler1016@gmail.com>

Bound of rules

Description

This function takes in a ruleset and output the lower and upper bounds of each rule.

Usage

extrule(myrules, varname)

Arguments

Value

A p*2 matrix, p is the length of varname. The first column is the lower bound, the second column is the upper bound. The default lower bound is "-Inf",the default upper bound is "Inf". row corresponse to x variables ordered in the data matrix given to rpart.

A harvested classification tree

Description

Basic function to apply the harvest algorithm to the training data set, computing whether we can harvest any nodes based on the classic classification tree algorithm.

Usage

harfunc(rpart.object, data, varname, sig = 0.95)

Arguments

Value

the list of orginial result of classification, likelihood improvment and harvested classification result.

A harvested classification tree

Description

The main function of the package, aiming at develop the harvest classification tree. Training data input and

Usage

harvest(training, num.var, numeric.info, sig = 0.95)

Arguments

Details

The function will return the harvested tree model. Missing values are allowed, and they will be treated accordingly. To use the trained tree model to predict, you can use predict function in this package.

Value

An object of class "harvest", which is the result of algorithm with the following elements for each nodes(nodes are ordered in sequence of harvesting):

rule constriants of the node

total total number of data points in the node

'1' the number of data points belonging to class 1 in the node

'logchange' the improvement of log likelihood of deleting the redundent rules by the algorithm for the node

Examples

data(training)
harvest(training,4,3)

Predictions from a harvested tree

Description

The function predict computes the prediction of membership from a new data set classified by harvested classification model of training data.

Usage

predict(harfunc.object, data, num.var)

Arguments

Details

To run the predict function, a trained harvested classification tree formed by harvest function is required.

Value

pred.mat is a data frame stored the information of result of prediction with the following columns:

belong the node that data point belongs to

possibility the probability of point being in class 1

predict the simple perdict based on whether probability is larger than 0.5.

Ranking of nodes

Description

Rank harvested node by lower p value

Usage

rank.nodes(harfunc.object)

Arguments

Value

the ranked harvest nodes

A logical matrix for a terminal node

Description

Return a logical matrix of the rule sets which define a terminal node

Usage

rulesets(noden, newsim, varn, nodenumb)

Arguments

Value

A nxnn logical matrix, n=number of data points to be harvested, nn=number of rules defining a terminal node. Each column of the matrix corresponding to a node that is defined by one variable/rule, its name corresponds to that variable. Note the original terminal node is just the intersection of these nodes.

training

Description

A simulated data set of symptoms of breast cancer patients

Usage

data(training)

Format

A data frame with 300 observations on the following 5 variables.

y

a numeric vector

x1

a factor with levels 21 22 23 24 25 26 27 28 29

x2

a factor with levels 39- 40-49 50-69 70-74 75+

x3

a numeric vector

x4

a factor with levels 2004 2005 2006 2007 2008 2009 2010

Source

simulated data for breast cancer diagnosis