You have learned about the AI Project Cycle and maybe even completed your 4W Canvas for the problem statement. Now you can start building your AI solution. But before you dive into modelling or coding, there’s one crucial step you must not skip. Or rather two. Data acquisition and data preparation,
Data acquisition and preparation stage can make or break your AI project. Because in any AI system, data is the “fuel” that powers your project. AI models feed off “data” to become intelligent.
If your data is wrong, incomplete or biased, even the smartest model will fail. In fact, most real-world AI failures happen because of bad or biased data. Not because the AI itself was poorly built.
This is why Data Acquisition and Data Preparation — the second and third stages of the AI Project Cycle — are so important.
In this article, you will dive deeper into:
- How to collect the right kind of data
- How to clean and organize it so your AI can learn
- How small mistakes in data can create big problems later
What Is Data Acquisition?
Data acquisition means gathering the information your AI project needs to solve the problem you have chosen. Choosing the right type and source of data helps you build a strong foundation for your AI model.
Think of it like collecting all the right ingredients before you cook a meal. If you miss an ingredient or pick the wrong one, your final dish won’t taste right. Similarly, if you collect poor or incomplete data, your AI solution may give wrong results.
When collecting or selecting data, it’s important to make sure it represents the full scope of your problem. For example, if you are studying lunch wastage and you collect data only from Grade 6 students, your project may not work well for the whole school.
Depending on your problem, you can collect two types of data:
Primary Data
This is data you collect yourself. Examples include:
- Surveys and questionnaires (using Google Forms or direct interviews)
- Observations and recordings (like counting tap leaks in school washrooms)
- Data captured through sensors, cameras or mobile apps
Secondary Data
This is data that already exists, collected by others. Reliable sources include:
- Government websites like data.gov.in (India’s national open data portal)
- World Bank Open Data (global statistics on education, economy, health)
- UNESCO Open Data (data related to education, science and culture)
- Kaggle (an online community offering a wide variety of public datasets)
Important Note: When using secondary data, always confirm it is from a reliable source. Also, ensure you are allowed to use it under an open-data license. An open data license is a legally binding agreement that allows anyone to use, share, and sometimes modify data freely, as long as they follow specific conditions set by the data provider.
What are Kaggle datasets?
Kaggle (www.kaggle.com) is an online platform owned by Google where people participate in machine learning competitions, share datasets, and work on AI projects. It has thousands of open-source datasets that students, researchers and developers can download and use freely for learning and project work.
I included Kaggle because it is:
- Very beginner-friendly: You can simply click and download datasets without complex logins or payments.
- Widely used even by college students and beginners in AI fields.
- Safe for practice, offering clean datasets on topics students can relate to — like movies, sports, shopping habits, health, etc.
Key Questions to Ask Before Collecting Data
Before you start collecting data for your AI project, it’s important to think carefully about what you need. Collecting data without a clear plan can waste time and cause confusion later. Also, collecting and storing data is expensive. So collecting and storing data that may not be useful is a waste of resources too.
To make sure you gather the right kind of data, ask yourself these key questions:
1. What problem am I trying to solve?
Be specific. A clear problem statement will guide what data you need.
Example: Predicting how much food will be left over in the canteen each day.
2. What type of data will help solve this problem?
Decide whether you need numbers, text, images or something else.
Example: Number of students present, menu items served, amount of food left each day.
3. Where will I get the data from?
Identify whether you will collect new data (primary) or use existing data (secondary). Choose sources that are reliable and relevant.
Example: Record leftover food daily (primary) or use past canteen logs and attendance data (secondary).
4. Is the data accurate, recent, and representative?
Outdated or biased data can lead to wrong results. Try to ensure that your data:
- Covers different groups involved
- Reflects current trends
- Is collected carefully without errors
Example: Collecting data only during exam week won’t reflect normal lunch patterns across the school.
5. Is the data collected ethically?
Respect privacy. If you are surveying people, take their permission. Make sure the data you use is openly available if you are downloading it from the internet.
Example: When surveying students, always ask for permission and keep their answers anonymous.
What Is Data Preparation?
Raw data that you collect is often messy, incomplete or inconsistent.You cannot use it as it is. Before using it as training data for your AI model, you need to clean and organise the data.
This cleaning and organising of raw data to make it fit for training AI models is called Data Preparation.
Data preparation involves performing one or more of these tasks for each piece of data:
- Removing duplicates: Sometimes, the same data entry might appear more than once. Duplicate entries must be deleted.
- Handling missing values: If some entries have missing information (like no value recorded for a day), you must decide whether to remove them or fill them carefully.
- Correcting inconsistencies: Data might have typing errors (like “yes” written as “yess” or “YES”). You need to make the formats uniform.
- Changing data types if necessary: Sometimes, you may have to convert data, for indtance text labels into numbers, so that the machine can understand them.
Preparing your data properly is like cleaning your room before starting a study session. A clean, organised space makes it easier to focus and the same is true for your AI model.
Example AI Project: Predicting Water Wastage in School Bathrooms
Problem: Students want to predict how much water might be wasted each day.
Data acquired: Daily counts of taps left running, water flow sensor readings.
Data preparation steps:
- Removed extra readings that were recorded on holidays.
- Corrected spelling errors in the logbook entries.
- Filled missing timestamps by checking nearby entries.
- Combined survey and sensor data into a single table.
After preparation, the dataset became much more reliable for the AI model to work with.
| Messy Data | Cleaned Data |
|---|---|
| Tap running, Tap runing, tap Running | Tap running |
| Monday, Monda, Mon | Monday |
| Blank entries for 2pm data | Filled with estimated values |
| Duplicate sensor entries for same time | One correct record kept |
Good data preparation improves the quality of your project and reduces errors during modelling. It also saves a lot of time later when you test and evaluate your model.
Tips for Choosing and Preparing Good Data
As you collect and prepare your data, it’s important to focus not just on quantity but also on quality. Good data is like a strong foundation for a building. If the foundation is weak, the building will not stand tall.
A good AI project is built on clean, fair, and well-understood data. Spending enough time on choosing and preparing your data will make every other stage of your project much smoother.
Here are a few tips for choosing and preparing good data:
- Choose relevant data: Every piece of data you collect should help solve the problem you have identified. Extra or unrelated information can confuse your AI model.
- Use recent and reliable data: Outdated data may not reflect current trends. Always check when and where the data was collected to make sure it is still useful.
- Watch out for bias: Make sure your data covers all groups fairly. If you only collect responses from a small or similar group, your model will not work well for everyone. Example: If you survey only students from one class about favorite school activities, you will miss the interests of students from other classes.
- Respect Ethics and Privacy: When collecting primary data, always get permission from people if you are recording their information. When using secondary data, make sure it is allowed to be reused (open source).
- Document Your Steps: Keep a simple record of:
- Where your data came from? What changes you made (like deleting duplicates or correcting spellings) Any assumptions you made while filling missing values
When you get stuck at some stage, you have the documentation to help trace the path you took, and identify what could be changed for better outcomes. Documentation also helps you explain your project better during presentations or evaluations.
Putting It All Together
Data Acquisition and Data Preparation are the foundation on which your entire AI project stands.
If you collect the wrong data or prepare it poorly, even the best algorithms cannot give good results. But if you spend time gathering the right data, cleaning it carefully, and thinking about bias and fairness, your AI project will be much stronger.
When you work on your next AI project, remember:
- You can use primary or secondary data or both
- Ask the right questions collecting data to collect high-quality data
- Data preparation makes raw data usable by cleaning and organizing it
- Be aware of data bias, AI ethics and project documentation for all your projects
In the next step of the AI Project Cycle, we will move into exploring your cleaned data to find patterns and insights. But remember: exploration is only as good as the data you begin with.
Take your time with data. It is critical to your project’s success.
Frequently Asked Questions (FAQs)
1. What is data acquisition in the AI project cycle?
(From the CBSE Class 10 Handbook)
Data acquisition is the process of collecting relevant information required to solve the problem identified in an AI project. It can involve collecting new data (primary) or using existing data (secondary), depending on the goal of the project.
2. Why is it important to collect relevant and authentic data in AI projects?
(From the CBSE Class 10 Handbook)
Relevant and authentic data helps train your AI model accurately and prevents misleading results. If the data is incorrect, outdated, or biased, your AI solution may fail to solve the intended problem effectively.
3. What are data features, and how do they impact your AI model?
(From the CBSE Class 10 Handbook)
Data features are the specific variables or attributes used to train an AI model. Choosing the right features is critical because they directly influence how well your model learns patterns and makes predictions.
4. What are the reliable sources of data collection mentioned in the curriculum?
(From the CBSE Class 10 Handbook)
The curriculum suggests using open government portals like data.gov.in, trusted research databases, web APIs, and direct sources like surveys, sensors, and observations. These ensure that the data is legal to use and of high quality.
5. What are the steps involved in preparing data for an AI project?
Data preparation involves cleaning, formatting, and organizing your collected data so that it can be used by an AI model. This includes removing duplicates, handling missing values, correcting errors, and converting data types when needed.
6. What is the difference between primary and secondary data in AI?
Primary data is collected directly by the project team through surveys, sensors, or observations. Secondary data is already available from other sources like government databases, research studies, or online repositories.
7. How do you clean data before using it in machine learning or AI models?
Cleaning data involves checking for and correcting errors such as duplicates, missing entries, spelling mistakes, or inconsistent formats. It ensures your data is accurate, uniform, and ready for training AI models.
8. Why is data cleaning important in artificial intelligence?
AI models learn from data. If the data is messy or incorrect, the model will learn the wrong patterns, leading to poor or biased results. Data cleaning ensures that your model learns only from high-quality, trustworthy information.
