Python Programming for Biology: Bioinformatics and Beyond

Figure 24.7.  Example output for an SVM classifier with two data dimensions. The

input SVM training data is shown with black and grey circles representing the two data

categories (left). The support vectors are indicated as the circled points. The prediction

result is displayed over the whole data range, to illustrate the classification boundaries

(right). The category for each point is illustrated by being a black or grey shade.

The following Python code tests the trained SVM on a whole range of different points

in the plane between 0 and 1, thus showing the shape of the categorisation boundaries. The

distinction between the categories is made according to whether the prediction is greater

than zero (positive) or not. Note that the x and y values for a given query are appended to

lists  so  that  the  data  can  be  displayed  as  graphical  scatter  plots,  using  the  helpful  pyplot

library (from MatplotLib).

from matplotlib import pyplot

ds1x = []

ds1y = []

ds2x = []

ds2y = []

x = 0.0

while x < 1.0:

y = 0.0

while y < 1.0:

query = array( (x,y) )

prediction = svmPredict(query, data, knowns, supports,

kernelGauss, params)

if prediction > 0:

ds1x.append(x)

ds1y.append(y)

else:

ds2x.append(x)

---

ds2y.append(y)

y += 0.02

x += 0.02

pyplot.scatter( ds1x, ds1y, color='grey' )

pyplot.scatter( ds2x, ds2y, color='black' )

pyplot.show()