jsat.io

Class LIBSVMLoader

java.lang.Object

jsat.io.LIBSVMLoader

public class LIBSVMLoader
extends Object

Loads a LIBSVM data file into a DataSet. LIVSM files do not indicate whether or not the target variable is supposed to be numerical or categorical, so two different loading methods are provided. For a LIBSVM file to be loaded correctly, it must match the LIBSVM spec without extensions.

Each line should begin with a numeric value. This is either a regression target or a class label.
Then, for each non zero value in the data set, a space should precede an integer value index starting from 1 followed by a colon ":" followed by a numeric feature value.
The single space at the beginning should be the only space. There should be no double spaces in the file.

LIBSVM files do not explicitly specify the length of data vectors. This can be problematic if loading a testing and training data set, if the data sets do not include the same highest index as a non-zero value, the data sets will have incompatible vector lengths. To resolve this issue, use the loading methods that include the optional vectorLength parameter to specify the length before hand.

Author:

Edward Raff

Method Summary

Modifier and Type	Method and Description
static ClassificationDataSet	loadC(File file) Loads a new classification data set from a LIBSVM file, assuming the label is a nominal target value
static ClassificationDataSet	loadC(File file, double sparseRatio) Loads a new classification data set from a LIBSVM file, assuming the label is a nominal target value
static ClassificationDataSet	loadC(File file, double sparseRatio, int vectorLength) Loads a new classification data set from a LIBSVM file, assuming the label is a nominal target value
static ClassificationDataSet	loadC(InputStreamReader isr, double sparseRatio) Loads a new classification data set from a LIBSVM file, assuming the label is a nominal target value
static ClassificationDataSet	loadC(Reader reader, double sparseRatio, int vectorLength) Loads a new classification data set from a LIBSVM file, assuming the label is a nominal target value
static RegressionDataSet	loadR(File file) Loads a new regression data set from a LIBSVM file, assuming the label is a numeric target value to predict
static RegressionDataSet	loadR(File file, double sparseRatio) Loads a new regression data set from a LIBSVM file, assuming the label is a numeric target value to predict
static RegressionDataSet	loadR(File file, double sparseRatio, int vectorLength) Loads a new regression data set from a LIBSVM file, assuming the label is a numeric target value to predict
static RegressionDataSet	loadR(InputStreamReader isr, double sparseRatio) Loads a new regression data set from a LIBSVM file, assuming the label is a numeric target value to predict
static RegressionDataSet	loadR(Reader reader, double sparseRatio, int vectorLength) Loads a new regression data set from a LIBSVM file, assuming the label is a numeric target value to predict.
static void	write(ClassificationDataSet data, OutputStream os) Writes out the given classification data set as a LIBSVM data file
static void	write(RegressionDataSet data, OutputStream os) Writes out the given regression data set as a LIBSVM data file

Methods inherited from class java.lang.Object

clone, equals, finalize, getClass, hashCode, notify, notifyAll, toString, wait, wait, wait

Method Detail

loadR

public static RegressionDataSet loadR(File file)
                               throws FileNotFoundException,
                                      IOException

Loads a new regression data set from a LIBSVM file, assuming the label is a numeric target value to predict

Parameters:

file - the file to load

Returns:

a regression data set

Throws:

FileNotFoundException - if the file was not found

IOException - if an error occurred reading the input stream

loadR

public static RegressionDataSet loadR(File file,
                                      double sparseRatio)
                               throws FileNotFoundException,
                                      IOException

Loads a new regression data set from a LIBSVM file, assuming the label is a numeric target value to predict

Parameters:

file - the file to load

sparseRatio - the fraction of non zero values to qualify a data point as sparse

Returns:

a regression data set

Throws:

FileNotFoundException - if the file was not found

IOException - if an error occurred reading the input stream

loadR

public static RegressionDataSet loadR(File file,
                                      double sparseRatio,
                                      int vectorLength)
                               throws FileNotFoundException,
                                      IOException

Loads a new regression data set from a LIBSVM file, assuming the label is a numeric target value to predict

Parameters:

file - the file to load

sparseRatio - the fraction of non zero values to qualify a data point as sparse

vectorLength - the pre-determined length of each vector. If given a negative value, the largest non-zero index observed in the data will be used as the length.

Returns:

a regression data set

Throws:

FileNotFoundException - if the file was not found

IOException - if an error occurred reading the input stream

loadR

public static RegressionDataSet loadR(InputStreamReader isr,
                                      double sparseRatio)
                               throws IOException

Loads a new regression data set from a LIBSVM file, assuming the label is a numeric target value to predict

Parameters:

isr - the input stream for the file to load

sparseRatio - the fraction of non zero values to qualify a data point as sparse

Returns:

a regression data set

Throws:

IOException - if an error occurred reading the input stream

loadR

public static RegressionDataSet loadR(Reader reader,
                                      double sparseRatio,
                                      int vectorLength)
                               throws IOException

Loads a new regression data set from a LIBSVM file, assuming the label is a numeric target value to predict.

Parameters:

reader - the reader for the file to load

sparseRatio - the fraction of non zero values to qualify a data point as sparse

vectorLength - the pre-determined length of each vector. If given a negative value, the largest non-zero index observed in the data will be used as the length.

Returns:

a regression data set

Throws:

IOException

loadC

public static ClassificationDataSet loadC(File file)
                                   throws FileNotFoundException,
                                          IOException

Loads a new classification data set from a LIBSVM file, assuming the label is a nominal target value

Parameters:

file - the file to load

Returns:

a classification data set

Throws:

FileNotFoundException - if the file was not found

IOException - if an error occurred reading the input stream

loadC

public static ClassificationDataSet loadC(File file,
                                          double sparseRatio)
                                   throws FileNotFoundException,
                                          IOException

Loads a new classification data set from a LIBSVM file, assuming the label is a nominal target value

Parameters:

file - the file to load

sparseRatio - the fraction of non zero values to qualify a data point as sparse

Returns:

a classification data set

Throws:

FileNotFoundException - if the file was not found

IOException - if an error occurred reading the input stream

loadC

public static ClassificationDataSet loadC(File file,
                                          double sparseRatio,
                                          int vectorLength)
                                   throws FileNotFoundException,
                                          IOException

Loads a new classification data set from a LIBSVM file, assuming the label is a nominal target value

Parameters:

file - the file to load

sparseRatio - the fraction of non zero values to qualify a data point as sparse

vectorLength - the pre-determined length of each vector. If given a negative value, the largest non-zero index observed in the data will be used as the length.

Returns:

a classification data set

Throws:

FileNotFoundException - if the file was not found

IOException - if an error occurred reading the input stream

loadC

public static ClassificationDataSet loadC(InputStreamReader isr,
                                          double sparseRatio)
                                   throws IOException

Loads a new classification data set from a LIBSVM file, assuming the label is a nominal target value

Parameters:

isr - the input stream for the file to load

sparseRatio - the fraction of non zero values to qualify a data point as sparse

Returns:

a classification data set

Throws:

IOException - if an error occurred reading the input stream

loadC

public static ClassificationDataSet loadC(Reader reader,
                                          double sparseRatio,
                                          int vectorLength)
                                   throws IOException

Loads a new classification data set from a LIBSVM file, assuming the label is a nominal target value

Parameters:

reader - the input stream for the file to load

sparseRatio - the fraction of non zero values to qualify a data point as sparse

vectorLength - the pre-determined length of each vector. If given a negative value, the largest non-zero index observed in the data will be used as the length.

Returns:

a classification data set

Throws:

IOException - if an error occurred reading the input stream

write

public static void write(ClassificationDataSet data,
                         OutputStream os)

Writes out the given classification data set as a LIBSVM data file

Parameters:

data - the data set to write to a file

os - the output stream to write to. The stream will not be closed or flushed by this method

write

public static void write(RegressionDataSet data,
                         OutputStream os)

Writes out the given regression data set as a LIBSVM data file

Parameters:

data - the data set to write to a file

os - the output stream to write to. The stream will not be closed or flushed by this method