I'm looking for a way of generating vehicle speed over time datasets for specific traffic scenarios, e.g. heavy city traffic, highway, urban area, etc...

I have a dataset of 8000 individual trips recorded from selected vehicles in a fleet. Unfortunately these do not include any GPS location information. Some of these are complete trips, some are only fragments of data, which should be filtered out. Rather than manually stepping through all 8000 sets, or programatically filtering with strict rules, I want to try using an NN or Fuzzy Inference System, which is trained on selected data to group the data into appropriate categories as well as label any incoplete or improbable dataset (e.g. extremely large distance over a short period of time).