AI & data science research starter

Original investigation outline

• Frame a question: How does the prediction threshold affect a model’s false positives and false negatives?
• Identify the variables: sensor score, true class, and model prediction. The sample dataset is fictional.
• Record the method: vary the threshold from 35 to 85 in steps of 10 and capture the four outcome counts.
• Analyse the trade-off: describe why accuracy alone does not measure all possible harms.
• Evaluate: identify what additional data you would need before applying such a model to real decisions.
• Reflect: decide what changed from your first plan and explain why.

Practical files and worked starter

Use the browser activity to explore the idea first, then download the corresponding CSV dataset and commented Python program to run locally.

Suggested research log