Analyzing the data

Analyzing the data

The raw data might contain errors or outliers, so it was necessary to preprocess it before further processing. The raw data was directly obtained from the smartphone memory and shifted to the computer for preliminary analysis.

After analyzing the data, some outliers were detected, i.e., some touch features were equal to zero or missing, and some values were not in the normal range of values. That was maybe because of an inappropriate touch event, e.g., rather than swipe movement, the user might mistakenly touch the screen. In this case, values of length and absolute length will be out of the normal range because these values do not belong to the swipe movement, so it is necessary to delete the whole vector to avoid lousy performance. In the first phase, the outliers were deleted manually. Later in phase 2, this process of discarding was handled directly from the mobile device.

From the raw data, some features were taken as it is, like starting and ending coordinates, pressure, and are values. Some features were extracted later from the raw data, such as direction, absolute length, the rectangular area under the swipe.