Skip to contents
Open the chapter list
  1. Chapter 1. Start Here
  2. Chapter 2. Set Up Your Workspace
  3. Chapter 3. Prepare Accelerometer Files
  4. Chapter 4. Record Observations
  5. Chapter 5. Build Your First Model
  6. Chapter 6. Predict with a Model
  7. Chapter 7. Understand Your Results
  8. Chapter 8. Optimise for Deployment
  9. Chapter 9. Scripted Reruns
  10. Chapter 10. Troubleshooting

Troubleshooting and Glossary

Source: vignettes/troubleshooting-and-glossary.Rmd
troubleshooting-and-glossary.Rmd

This is Chapter 10 of 10 in the beginner path.

If something goes wrong, that does not mean you are “bad at R” or that the project is failing. Most problems in an early moover workflow are ordinary setup issues: a path is wrong, a timestamp is being interpreted incorrectly, or ids do not line up across files.

This chapter collects the most common issues in one place.

When the import preview looks wrong

If the import preview does not look sensible, stop there and fix the import before doing anything else.

Typical causes are:

  • the wrong timestamp format was chosen
  • the wrong columns were mapped in a generic file
  • the ids in the raw data are not the ids you expected

The preview stage exists precisely so you can catch those problems early.

When training fails or behaves strangely

If training runs but the results look suspicious, the most common causes are usually in the inputs rather than the Random Forest settings.

Things to check first:

  • Do the observation ids match the raw-data ids?
  • Are the observation times in the right timezone and format?
  • Are the behaviour labels spelled consistently?
  • Do you actually have enough labelled examples for each behaviour?

These checks are usually more productive than changing model settings at random.

When prediction gives unexpected results

If a prediction run finishes but the outputs do not make sense, ask whether the new data really matches the conditions the model expects.

For example:

  • Is it the same type of accelerometer data?
  • Are the ids being mapped correctly?
  • Can the same features be rebuilt from the new files?

Unexpected predictions do not automatically mean the model is broken. Often they mean the new dataset is not aligned with the training setup.

A short glossary

Accelerometer file: the raw movement data recorded by the device.

Canonical format: the standard 5-column layout moover uses internally: id, t_unix_ms, x, y, z.

Epoch: a fixed time block. moover builds features for each block.

Feature: a numeric summary calculated from the movement data within an epoch.

Model bundle: the exported folder containing the fitted model and supporting files.

LOCO: leave-one-cow-out validation, meaning the model is tested on animals not used for training.

Spec: a saved set of instructions for a run.

When to ask for help

A good moment to ask for help is after you have answered three questions:

  1. Which chapter were you following?
  2. Which file or step seems to be causing trouble?
  3. What did the preview or output look like when it went wrong?

Those details usually make it much easier for someone else to help you quickly.

Move through the tutorial
Previous chapter: Chapter 9. Scripted Reruns and Saved Specs
Back to the start: Chapter 1. Start Here