Skip to content
LessonItem 1 of 36

The Data Science Workflow from Question to Conclusion

You are viewing a preview. Enroll to save your progress.

Your school produces data every day: attendance lists, test scores, canteen sales, library loans. Most of it is simply stored. Data science is a way of working that turns raw data into answers that can be used to make decisions.

Example: the school canteen

The school cooperative wants to know which menu items deserve more stock. A data analyst would work through the following cycle:

  1. Frame the question. The question must be specific and answerable with data, for example "Which menu item sold best from Monday to Friday during August?"
  2. Collect the data. The cashier's record of each transaction: date, time, item, quantity, price.
  3. Clean the data. Fix item names written in different ways ("Nasgor", "nasi goreng"), remove duplicate rows, handle empty cells.
  4. Explore. Compute totals and averages and compare groups, for example sales per day.
  5. Visualise. Draw a bar chart of sales per item so the pattern is visible at a glance.
  6. Conclude and communicate. "Chicken noodles sold 40% more on Fridays; stock should be increased on that day." A conclusion also states its limits, for example that the data covers only one month.

The cycle often repeats: exploration can raise new questions.

The shape of the data we will work with

Almost all data in this course is a table. Each row is one observation (one student, one transaction), and each column is one variable (name, class, score). Variables come in two main kinds:

  • Numerical: numbers whose average makes sense, such as a maths score or hours of study.
  • Categorical: labels or groups, such as class, gender, or extracurricular club.

The kind of variable decides how you process it. An average score makes sense; an average class name does not.

The tools we use

We use Python with two main libraries: pandas for working with tables and matplotlib for drawing charts. Everything runs in a notebook (Jupyter or Google Colab), a document that combines code, results, charts, and notes in one place, so your analysis is easy for others to reread.