Final Year Project Report Best Practices

Chapter 4 · Practice 2 of 4

Report every number with the conditions that produced it

From Testing, evaluation and results

Why it matters

A number without its conditions cannot be checked. "95% accuracy" means little until the reader knows the dataset, the split, the baseline and the metric.

For each result, state:

  • Data: which dataset, how many records, how it was split.
  • Setup: hardware, settings, number of runs.
  • Metric: accuracy, precision, recall, F1, response time, task completion, and why that metric fits.
  • Baseline: what you compared against, even if it is a simple rule.

Report results that did not go your way as well. A model that failed to beat the baseline, explained carefully, still demonstrates good research practice. Never adjust, round up or invent numbers; examiners often ask to rerun a test or show the raw output.

Do

  • State data, setup, metric and baseline for each result
  • Keep raw outputs and logs for every reported number

Avoid

  • Rounding results in your favour
  • Hiding results that did not support your hypothesis