Remember, healthy predictions need healthy data!
Healthy
dataset
Finds
correct
patterns
Healthy
prediction!
Correct actions or decisions!
But what makes data healthy data?
Can you make a list?
Did you think of any of these in your list? (click to see)
Healthy data
Lots of data
Accurate
Matches your problem and solution
Different examples of data
The right kind of data
You have permission to use it
Healthy prediction!
Better actions or decisions
THERE ARE 3 WAYS YOU CAN COLLECT HEALTHY DATA
Collect your own training data from
Gather data with
Use data from
What path is right for you?
COMMUNITY
if the information you need is available from your community, and you can get permission to use it.
SENSORS
if you want your app to collect data on its own .
PUBLIC DATASETS
if you need lots and lots of data or you’re solving a problem that is more global.
COMMUNITY DATA
Plan: What type of data do you need from your community? How will you gather it?
SENSOR DATA
Voice assistants like Siri or Alexa have a microphone, which is a sound sensor, to gather data (like what a person is saying)
This smart crosswalk has a light sensor to gather information about how light and dark it is and a motion sensor to detect movement.
Different sensors gather different types of data!
HARDWARE OPTIONS
You will need devices to help you capture your data. Consider using:
Phone
Your phone has many sensors that can be used to gather information... Think mobile app!
Arduino
Small circuit board that can gather data and is built to use AI.
micro:bit
Small circuit board that can gather data and is built to use AI.
Raspberry Pi
Mini-computer (board) as powerful as a laptop!
For more, check out this big list of sensors.
This video from our World Summit 2020 has lots of great information on hardware options!
PUBLIC DATASETS
You might want to use a public dataset if …
You want more complete data.
You want to use a lot of data.
You want accurate data.
WHERE TO FIND PUBLIC DATASETS
Here are some good places to start. (Click images to visit sites.)
Here are some links to sample datasets according to problem types:
- Water issues: World WASH Data
- Life on land issues:
- Climate issues
- Gender inequality: UNDP Gender inequality dataset
Here are some links to sample datasets according to problem types:
- Water issues: World WASH Data
- Life on land issues:
- Climate issues
- Gender inequality: UNDP Gender inequality dataset
WHERE ELSE COULD YOU LOOK?
You can try:
- local organizations
- local schools
- local libraries
- where else?
It’s not easy finding the perfect dataset. Remember to ask for help!
The activity below is practice in finding and using a public dataset and training a model using Machine Learning For Kids. The dataset has information about iris flowers, and the model will predict the type of iris based on length and width of flower parts.
ACTIVITY: TRAIN A MODEL
In the next unit, we’ll use the iris AI model in a Scratch project!
DO THE BEST YOU CAN!
You may not be able to gather the perfect dataset, but gather what you can.
Proxy data is data that is a placeholder for a later version of your dataset.
JUMP IN AND STARTING PUTTING YOUR DATASET TOGETHER!
Check off these items to make sure you create a strong dataset!
