Table 3.

Accuracy of the SVM prediction

Target value	Total actuals	Correctly predicted (%)	Cost
Insoluble	549	56.50	239
In PDB	290	74.80	73

The attributes identified in Table 2 were used to calculate properties for the MCSG-INSOLUBLE and the MCSG-PDB data sets. The generated matrix was used in a clustering using a SVM approach splitting the entries 60–40%. The 60% portion was used for training while the 40% portion for evaluating the accuracy of the prediction. The SVM model building and validation were repeated five times. The insoluble proteins were predicted at 56% while the PDB deposits were predicted at 75% accuracy