1. Home
  2. K Nearest Neighbours Options

K Nearest Neighbours Options

Use this to select different options to be used in constructing the K Nearest Neighbours and the displayed output.

Display

Specifies which items of output are to be displayed in the Output window.

Similarities The symmetric matrix of similarities between observations. This may be very large.
Abbreviated similarities This reduces the printing of the similarity matrix to just the first decimal digit (available only when Similarities are selected).
Data summary The list of data variables and their means, minima, maxima and test types.
Cross-validation errors The cross-validation error for all combination of options provided. If the Data to predict is a factor, this is a mean squared error. If is a factor, it is the percentage of observations for which the predictions and observed values do not match.
Confusion matrix The percentage of observations for each observed group allocated to the predicted groups using the optimal combination of options.
Predictions The predicted values for the observations from the rest of the observations using the optimal combination of options.

Calculate predictions of variates using

This setting controls how the values of the neighbours are summarized when the Data to predict is a variate.

Medians The median of the neighbour’s values will be used.
Means The mean of the neighbour’s values will be used.

Number of simulations

Specifies the number of times the data is split into random cross-validation groups. Increasing this will increase precision but slow down the analysis.

Number of cross-validation groups

Specifies the number of cross-validation groups into which the data are randomly split. Values between 5 and 10 (the default) are reasonable. If this is too low, the cross-validation error may be lower than could be achieved using a full training set, but if set too high the cross-validation error may not reflect the variation in the data set.

Seed for randomization

This gives a seed to initialize the random number generation used for the random selections of variates and units. Using zero initializes this from the computer’s clock, but specifying an nonzero value gives a repeatable analysis.

Defaults

Reset the settings in the dialog to what they were on first opening the dialog.

Action Icons

Clear Clear all fields and list boxes.
Help Open the Help topic for this dialog.

See also

Updated on February 14, 2023

Was this article helpful?