Data & Analytics

Exploratory data analysis in Python, on generated test data

• Role: I wrote the notebook.
• When: Jan 2025
• Context: The question was simple: what does this generated dataset look like? The notebook states its aim in one comment, "Create charts based on the dataset." The data is generated, not customer data.
• What I did: I loaded the JSON file with pandas and drew five charts with matplotlib: a 15-bin histogram of age, the top 10 currencies by total random float, the true and false split of a boolean column, the top 10 companies by total random integer, and a latitude and longitude scatter plot.

• Example decision (part of the analysis exercise, not a delivered result): In this analysis, I chose to chart only the top 10 currencies and the top 10 companies, ranked by their totals, instead of a bar for every value.

  • The options: a bar for every currency and every company, or a top 10 cut ranked by total.
  • What I weighed: currency and company are category fields, and a generated set can hold a lot of different values. A bar for each one turns into a wall of labels nobody can read. Age, the true and false column, and latitude and longitude didn't have that problem, so they got a histogram, a pie and a scatter plot.
  • What I chose and why: add up each group, sort it largest first and keep the top 10. That puts the biggest groups up front. The trade-off is that everything below the cut drops out of the chart.
  • What I'd watch: how much of the total sits outside the top 10, and whether the ranking holds when the data is generated again.
• Artifact: SampleData.ipynb on GitHub
• Stack/tools: Python, pandas, matplotlib