Find Missing and Incorrect Labels
Some deliveries are labeled late even though arrival records are missing.
Find Missing and Incorrect Labels explained visually
Some deliveries are labeled late even though arrival records are missing. English captions are optional and start off. The complete transcript follows the written lesson.
What you will learn
Some deliveries are labeled late even though arrival records are missing.
A practical question
The delivery table marks some trips late even though their arrival time is blank. A missing value is not the same as a zero, and a filled label is not automatically correct.
The useful idea
Trace a sample of records back to dispatch logs. Check duplicate trips, impossible timestamps, and whether 'late' means the same threshold across depots. Separate values unavailable because they were never recorded from values that appear only after the trip. Each pattern can affect a model differently.
Work through the example
Count how many labels are uncertain by route and month. If missing logs cluster around difficult late trips, quietly dropping those rows can change the learning question. Document the review rule and preserve the original records.
Limits and next step
Clean evidence makes a useful model possible. Next, we'll ask whether the cleaned sample represents the routes on which the system will run.
Worked example
Audit a sample of uncertain delivery records against dispatch logs.
Quick check: test your understanding
Try it yourself: In the situation described above, explain the method and one reason its output might be misleading.
Check the reasoning
Check label definitions, missingness, duplicates, and impossible values before fitting. Automatically dropping rows can bias the remaining population.
What is the main idea behind find missing and incorrect labels?
Check label definitions, missingness, duplicates, and impossible values before fitting.
How is it used in the example?
Audit a sample of uncertain delivery records against dispatch logs.
What limitation should be checked?
Automatically dropping rows can bias the remaining population.
Further reading
Machine Learning — Tom M. Mitchell.
Original ToolsInVision teaching examples. Editorial approach · Suggest a correction
Accessible referenceLesson 21 transcript137 words · Video transcript · Matches the published narration and captionsOpen transcript
The delivery table marks some trips late even though their arrival time is blank. A missing value is not the same as a zero, and a filled label is not automatically correct.
Trace a sample of records back to dispatch logs. Check duplicate trips, impossible timestamps, and whether 'late' means the same threshold across depots. Separate values unavailable because they were never recorded from values that appear only after the trip. Each pattern can affect a model differently.
Count how many labels are uncertain by route and month. If missing logs cluster around difficult late trips, quietly dropping those rows can change the learning question. Document the review rule and preserve the original records.
Clean evidence makes a useful model possible. Next, we'll ask whether the cleaned sample represents the routes on which the system will run.