As opposed to defining a single best classification model (or algorithm) for all formations, it was better to use the best classifier for each formation. Using F-1 score as an evaluation metric, the algorithm with the highest F-1 score for a given formation – based on cross validation error – can be used to classify SSI for that formation. Adopting this method will result in a formation dependent classifier as shown in Table 2, Table 5; this was the bag-of-models approach, where the best classification algorithm among many was used for a specific formation based on some evaluation criteria. This method follows the no free lunch theorem in statistics.