2.5. Experiment 2: Label Terrain Resolution

In this experiment, the appropriate level of terrain resolution is determined. There was an expected trade-off between the resolution of detail in the activity and the corresponding classifier performance. This experiment aimed to validate this theory, and to show at what specific label of terrain resolution can an acceptable level of classifier accuracy be obtained. Using the best performing classifiers from experiment 1, the terrain resolution was increased in stages from level 1 to level 3 as per the criteria described in Section 2.3. Each level of resolution is validated using 5-fold cross validation of the entire dataset.