Data and AI Training

Home | Power BI | Excel | Python | SQL | Generative AI | Visualising Data | Analysing Data

Gen AI for Data Analysis - Exercises

There are four group exercises to practice data analysis using Generative AI tools. Each group will have a different cases study: Strictly, Bank Churn, Titanic and Supercar.

There will be about 30 minutes for the exercise. Within your groups, please:

• introduce yourself: videos on!
• elect a spokesperson
• start with the suggested prompts

There will be about 10 minutes for each group’s presentation to the class. In the presentation, the spokesperson will

• explain the dataset and background (class will not be familiar with this)
• describe your results
• include any insights, conclusions, opinions

Here are the links to each dataset and some specific tips

Strictly

The dataset details and download links are here.

Here are some suggested prompts to start your analysis.

• Act as a data analyst. Analyse the data in the Week 1 table of this web page
• What was the total score given by each judge?
• Who is the meanest judge?
• Combine with the data in the Week 2 and Week 3 tables and describe the combined dataset.
• The tables from Week 2 onwards also show the result e.g, whether the couple was safe, in the bottom two or eliminated. (In week 1 there was no elimination.) Is it always the couple with the lowest score that is eliminated? (note: in my practice run, the AI tool made mistakes so please check its answers carefully)

Bank Churn

The dataset details and download links are here.

Titanic

The dataset details and download links are here.

Here are some suggested prompts to start your analysis. The first prompt provides the context. This is especially useful as some of the variable names are difficult to understand.

1. Act as an data analyst. The attached data has a partial list of passengers on the Titanic. Here is a description of the columns insert here
2. Describe the data
3. Are there any data quality issues in the data?
4. How many passengers were in each passenger class?
5. What percentage of passengers survived?
6. What factors made it more likely that a passenger would survive?
7. Can you provide the results of your analysis in a few charts.

Supercar

The dataset details and download links are here.

Here are some suggested prompts to start your analysis. The first prompt describes of the columns – this is useful as the variable names are difficult to understand

• Act as an data analyst. The attached data has a details of 32 cars from the 1970s. Here is the description of the columns. insert here
• To make the charts more readable use Transmission rather than am and Automatic / Manual labels for the values rather than 0 / 1
• Create similar scatter charts but use engine shape on the legend
• Is there a correlation between weight and horsepower?
• Add a column to the data indicating where each of these cars was manufactured: US, Europe or Asia.
• Analyse fuel efficiency by Origin (in my practice run, the response to the previous prompt created a column named Origin)
• Is it possible to buy any of these cars today?

Suggestions for getting the best from these exercises

Attendees are in teams of about 4 - 5 people and work in a breakout room. Each team will have a different exercise.

Teams then report back to the class. Here are some hints how to do this well:

• agree on a main spokesperson for each presentation
• start at the beginning; explain the background, progress, challenges as well as results and conclusions
• show and tell: for example, when presenting a conversation with an AI tool, give the audience time to read the conversation