Definition
To predict how many students will want lunch tomorrow, what would you need to know? Past sales? The day of the week? Whether there is an exam? The weather? Where would you get each of these?
Data acquisition is the stage of the AI project cycle where the team identifies and collects the data needed to solve the problem, from sources such as surveys, sensors, records, cameras and open data. A system map is a diagram that shows the parts of a situation and how they affect each other, which helps the team decide what data matters.
In simple words
Before cooking, you check which ingredients you need and where to get each one: vegetables from the market, spices from the kitchen shelf, milk from the dairy. Data acquisition is the same shopping trip, for information instead of food.
Why it matters
An AI model can only learn from the data it is given. Missing, biased or wrongly collected data leads to poor or unfair results. Choosing and collecting data carefully, and legally, is one of the most important steps in any AI project.
Key terms
- Data acquisition
- Identifying and collecting the data a project needs.
- Data source
- Where data comes from, such as a survey, a sensor or a government website.
- Feature
- One piece of information used to make a prediction, such as the day of the week.
- Training and testing data
- Examples used to teach the model, and separate examples kept aside to check it.
- System map
- A diagram of the parts of a situation and the cause-and-effect links between them.
- Feedback loop
- A chain of effects that circles back, so a change strengthens or weakens itself.
Explained in detail
Drawing a system map first
A system map helps you see what affects the problem before you collect anything. Write the problem in the middle, such as "lunches sold". Around it, write the things that might affect it: number of students present, day of the week, exams, sports practice, weather, menu, price. Draw arrows showing cause and effect. Mark an arrow with + if more of one thing leads to more of the other (more students present, more lunches sold) and with − if more leads to less (higher prices, fewer lunches sold). The elements with arrows into the problem become candidate features to collect.
Feedback loops
Some arrows form circles. If popular food runs out, students stop coming to the canteen; fewer students means the canteen cooks less; cooking less means food runs out sooner. This is a feedback loop that makes the problem worse over time. Spotting loops like this on a system map shows where a solution could break the cycle, and warns you that past sales data may already be shaped by the problem itself.
Where data comes from
| Source | Example | Things to check |
|---|---|---|
| Surveys and interviews | Asking students what they plan to eat | Fair questions, consent, enough people |
| Records | The canteen's daily sales register | Complete? Accurate? Permission to use? |
| Sensors and cameras | A counter at the canteen door | Privacy, accuracy, placement |
| Open data | Weather records from a public website | Licence, date, how it was collected |
| Observation | Counting the queue every ten minutes | Consistent method, enough days |
Collecting data responsibly
- Collect only what the problem needs: names are not needed to predict lunch numbers.
- Ask for consent and explain how the data will be used.
- Make sure the data represents everyone affected, including small groups.
- Respect website terms and copyright: not all data online is free to use.
- Keep training and testing data separate so the model is tested fairly.
India's Open Government Data platform publishes many free data sets, on weather, agriculture, health, education and transport, which school and college teams often use for projects. Always read the description to understand what each column means and how the data was gathered.
How it works
- Draw a system map with the problem in the centre.
- Mark cause-and-effect arrows with + or −, and look for loops.
- Choose the features that most affect the problem.
- Find a fair, legal source for each feature.
- Collect the data with consent, and set some aside for testing.
Try it yourself
Mapping the canteen system
The canteen team draws a system map before collecting any data.
Good data source or not?
For predicting daily canteen demand, is each source useful and fair?
The school's daily attendance count
Photos of students secretly taken in the queue
The sports and exam calendar
Students' home addresses
Past daily sales from the canteen register
Common mistakes
- Collecting data before deciding what the problem needs.
- Collecting personal data that is not needed.
- Testing the model on the same data it learned from.
Check yourself
Summary
- Data acquisition collects the data a project needs from fair, legal sources.
- System maps show causes, effects and feedback loops, and guide feature choice.
- Collect only what is needed, with consent, and keep test data separate.
Finished reading?
Sources: Open Government Data Platform India · CBSE Artificial Intelligence curriculum resources