YOURPACEAI WORKSHOP · Beginner+

Data & machine learning

Build a coding foundation, explore datasets, and learn how models make predictions.

75–120 minutes · Five modules, a portfolio workbook and automatic assessment.

A spreadsheet is enough for the workshop. Optional Python/Pandas extension requires basic coding. The supplied CSV is fictional.

Open interactive workshop

Read all lessons freely here. The interactive workshop lets you save your workbook, take the automatic assessment and earn a non-accredited certificate.

By the end, you can

Clean a small dataset with a documented rule.
Calculate a reproducible result and choose a chart.
Explain limitations and avoid leakage in a prediction design.

Practice materials

Download practice pack

Read the fictional source material

FICTIONAL SALES CSV
id,day,category,sales_gbp
1,Mon,Books,40
2,Tue,Books,60
2,Tue,Books,60
3,Wed,Books,
4,Thu,Books,80
5,Fri,Books,100

Question: What is the average sales value across unique records with a recorded amount?
Unit: GBP. Blank means missing, not zero. Repeated id 2 is an exact duplicate. No population sampling claim is made.

From a dataset to a useful answer

MODULE 1 OF 5

Start with a question

Choose a question before choosing a model. A spreadsheet and a chart may answer it. Use a public dataset with clear permission and definitions; record what each column means.

Worked example

Input

Question: average recorded sales for unique records; CSV has five unique IDs.

Reviewed result

Use id to identify the exact duplicate, and sales_gbp for the amount. Keep the raw CSV unchanged.

Define the question and units first so the cleaning rule is linked to the analysis.

Quick check: Should a missing amount automatically mean zero sales?

No. Missing and zero are different states.

Your activity

Pick a small public dataset. Write one question and identify the columns you need.

MODULE 2 OF 5

Check the data

Look for missing values, duplicates, inconsistent units and unrepresentative samples. AI can suggest cleaning code, but inspect and test it. Keep the original dataset and document changes.

Worked example

Input

Six raw rows; id 2 appears twice; id 3 has a blank amount.

Reviewed result

Drop the repeated exact record. Exclude id 3 only from this mean calculation and document it. Four recorded values remain: 40, 60, 80, 100.

This is a stated rule for this exercise. Real duplicates or missing values require context rather than automatic deletion.

Quick check: What is the mean under this rule?

(40 + 60 + 80 + 100) / 4 = £70.

Your activity

Count missing values and duplicates. Explain one cleaning decision and how it changes your answer.

MODULE 3 OF 5

Avoid misleading results

A chart shows a relationship, not necessarily a cause. For predictive models, keep test data separate from training and avoid using information unavailable at prediction time. Compare against a simple baseline.

Worked example

Input

A model predicts purchases using a field recorded after purchase.

Reviewed result

Remove post-purchase information from prediction inputs; keep a test set separate and compare with a simple baseline.

Information unavailable at prediction time creates leakage and misleading evaluation.

Quick check: Does a rising chart prove what caused the increase?

No. These values alone do not establish causation.

Your activity

Make one chart and write a finding plus two limitations. If predicting, define the test split before training.

MODULE 4 OF 5

Make your result reproducible

In a spreadsheet, keep Raw and Clean tabs. Document duplicate removal and the missing-value rule, then calculate count, sum and mean. In Python, inspect duplicated IDs and missing values before transforming the data.

Worked example

Reproduction checklist: 6 raw rows → 5 unique IDs → 4 recorded amounts → total £280 → mean £70.

Your activity

Create a bar chart of the four recorded day amounts. Label GBP and note the excluded missing record.

MODULE 5 OF 5

Write a finding with limits

Separate your arithmetic from broader claims. This is a tiny fictional set, not evidence about a business or population. A useful report includes what you did and what you cannot conclude.

Worked example

Finding: the four unique recorded amounts average £70. Limitations: one value is missing and this invented week cannot establish a trend or cause.

Your activity

Write a finding, the cleaning log and two limitations. Add a separate plan for how you would evaluate a predictive model.

A reproducible data analysis

Clean the fictional CSV, calculate the mean, design a chart and explain what the data cannot show.

Workbook sections

Question, column meanings and cleaning rules
Clean values, formula or code, count and mean
Chart description and evidence-backed finding
Missing-data limits and a leakage-free evaluation plan

Review rubric

The raw data and stated cleaning rules are retained.
The exact duplicate is handled and the missing amount is not silently replaced with zero.
The mean is reproducible and chart units are clear.
Findings do not claim causation or population generality.

Automatic assessment and certificate

Five knowledge questions and three applied scenario checks are marked immediately. Pass with at least 4/5 knowledge answers and all 3/3 applied checks correct. Feedback and retries are available. Certificates also require five completed activities and a four-section workbook. The portfolio is recorded, not independently graded; the assessment is open-book, unproctored and non-accredited.

Take the workshop assessment

Continue with external study

CS50’s Introduction to Python — Harvard
Free course materials · Suitable for first-time programmers
Python & Pandas lessons — Kaggle Learn
Practise programming and data handling in notebooks
Intro to Machine Learning — Kaggle Learn
Learn model training and validation

Go further with a portfolio project

Use a public dataset to answer one question. Make a chart, explain your findings, and record the limitations of the data.

External course access and fees are set by their providers.