Results file format for tsml/sktime

The experiments files in the tsml and sktime toolboxes classification output results a specifically formatted csv. Below we provide an example of an results file in this format and describe each of the compontents.

1. Line 1 of the results file provides descriptive meta information about the experiment. The values in this line are:
   1. The dataset name
   2. The classifier name
   3. Whether these results are on the test data or a performance estimate using the train data
   4. The train/test split fold id
   5. The time unit for any recorded timings
2. Line 2 provides classifier specific training and parameter information. This could be the seed, parameters such as number of trees or accuracy found during an internal tuning process.
3. Line 3 provides simple summary performance statictics. These are:
   1. Accuracy
   2. Train time
   3. Sum prediction time
   4. Time to perform a set of benchmark operations for normalising over different hardware
   5. Maximum runtime memory
   6. Number of classes
   7. Train set accuracy estimate method i.e. cross-validation or out of bag error
   8. Train set estimate time
   9. Train time plus estimate time
4. All remaining lines are details about each prediction made. The first prediction line will correspond to the first series in the data file, the second line to the second series etc. The components of each prediction line are:
   1. The series true class value
   2. The series predicted class value
   3. The predictions probabilities for each class value
   4. Time to make the prediction