class static
analytics.OneVsAll

Source: analyticsdoc.js:3082

One vs All model for multiclass prediction. Builds binary model for each category and predicts the one with the highest score. Binary model is provided as part of the constructor.

Methods

decisionFunction(X)
fit(X, y)
getParams()
predict(X)
setParams(params)

new OneVsAll([arg])

Example 

// import analytics module
var analytics = require('qminer').analytics;
// create a new OneVsAll object with the model analytics.SVC
var onevsall = new analytics.OneVsAll({ model: analytics.SVC, modelParam: { c: 10, maxTime: 1000 }, cats: 2 });

Parameter

Name Type Optional Description

arg

module:analytics~oneVsAllParam

Yes

Construction arguments.

Methods

decisionFunction(X) → (module:la.Vector or module:la.Matrix)

Apply all models to the given vector and returns a vector of scores, one for each category. Semantic of scores depend on the provided binary model.

Example 

// import modules
var analytics = require('qminer').analytics;
var la = require('qminer').la;
// create a new OneVsAll object with the model analytics.SVC
var onevsall = new analytics.OneVsAll({ model: analytics.SVC, modelParam: { c: 10, maxTime: 1000 }, cats: 2 });
// create the data (matrix and vector) used to fit the model
var matrix = new la.Matrix([[1, 2, 1, 1], [2, 1, -3, -4]]);
var vector = new la.Vector([0, 0, 1, 1]);
// fit the model
onevsall.fit(matrix, vector);
// create the vector for the decisionFunction
var test = new la.Vector([1, 2]);
// give the vector to the decision function
var prediction = onevsall.decisionFunction(test); // returns the vector of scores

Parameter

Name Type Optional Description

X

(module:la.Vector, module:la.SparseVector, module:la.Matrix, or module:la.SparseMatrix)

 

Feature vector or matrix with feature vectors as columns.
Returns

(module:la.Vector or module:la.Matrix) The score and label of the input X:
1. module:la.Vector of scores, if X is of type module:la.Vector or module:la.SparseVector.
2. module:la.Matrix with columns corresponding to instances, and rows corresponding to labels, if X is of type module:la.Matrix or module:la.SparseMatrix.

fit(X, y) → module:analytics.OneVsAll

Apply all models to the given vector and returns category with the highest score.

Example 

// import modules
var analytics = require('qminer').analytics;
var la = require('qminer').la;
// create a new OneVsAll object with the model analytics.SVC
var onevsall = new analytics.OneVsAll({ model: analytics.SVC, modelParam: { c: 10, maxTime: 1000 }, cats: 2 });
// create the data (matrix and vector) used to fit the model
var matrix = new la.Matrix([[1, 2, 1, 1], [2, 1, -3, -4]]);
var vector = new la.Vector([0, 0, 1, 1]);
// fit the model
onevsall.fit(matrix, vector);

Parameters

Name Type Optional Description

X

(module:la.Matrix or module:la.SparseMatrix)

 

training instance feature vectors.

y

module:la.Vector

 

target category for each training instance. Categories must be integer numbers between 0 and oneVsAllParam.categories-1.
Returns

module:analytics.OneVsAll Self. The models have been fitted.

getParams() → module:analytics~oneVsAllParam

Gets the parameters.

Example 

// import analytics module
var analytics = require('qminer').analytics;
// create a new OneVsAll object with the model analytics.SVC
var onevsall = new analytics.OneVsAll({ model: analytics.SVC, modelParam: { c: 10, maxTime: 1000 }, cats: 2 });
// get the parameters
// returns the JSon object
// { model: analytics.SVC, modelParam: { c: 10, maxTime: 1000 }, cats: 2, models: [] }
var params = onevsall.getParams();
Returns

module:analytics~oneVsAllParam The constructor parameters.

predict(X) → (number or module:la.IntVector)

Apply all models to the given vector and returns category with the highest score.

Example 

// import modules
var analytics = require('qminer').analytics;
var la = require('qminer').la;
// create a new OneVsAll object with the model analytics.SVC
var onevsall = new analytics.OneVsAll({ model: analytics.SVC, modelParam: { c: 10, maxTime: 1000 }, cats: 2 });
// create the data (matrix and vector) used to fit the model
var matrix = new la.Matrix([[1, 2, 1, 1], [2, 1, -3, -4]]);
var vector = new la.Vector([0, 0, 1, 1]);
// fit the model
onevsall.fit(matrix, vector);
// create the vector for the prediction
var test = new la.Vector([1, 2]);
// get the prediction of the vector
var prediction = onevsall.predict(test); // returns 0

Parameter

Name Type Optional Description

X

(module:la.Vector, module:la.SparseVector, module:la.Matrix, or module:la.SparseMatrix)

 

Feature vector or matrix with feature vectors as columns.
Returns

(number or module:la.IntVector) 
1. number of the category with the higher score, if X is module:la.Vector or module:la.SparseVector.
2. module:la.IntVector of categories with the higher score for each column of X, if X is module:la.Matrix or module:la.SparseMatrix.

setParams(params) → module:analytics.OneVsAll

Sets the parameters.

Example 

// import analytics module
var analytics = require('qminer').analytics;
// create a new OneVsAll object with the model analytics.SVC
var onevsall = new analytics.OneVsAll({ model: analytics.SVC, modelParam: { c: 10, maxTime: 1000 }, cats: 2 });
// set the parameters
var params = onevsall.setParams({ model: analytics.SVR, modelParam: { c: 12, maxTime: 10000}, cats: 3, verbose: true });

Parameter

Name Type Optional Description

params

module:analytics~OneVsAllParam

 

The constructor parameters.
Returns

module:analytics.OneVsAll Self. The parameters are changed.