Here are techniques for creating fun predictive tools using spatial data.

• Load NYC Open Data
• Prepare Your Training Dataset
• Train Your Model
• Add Visualizations
• Create and Deploy Your Tool

Instructions¶

1. Use the links at the right to explore the notebooks.
2. Sign in using your Google account.
3. Choose a NUMERICAL environment and TINY memory for a fast startup. (WARNING! If you choose any other configuration, your notebook session may take as long as five minutes to begin).
4. Start a 60 minute session to experiment with the code.

Steps to Creating Your Fun Predictive Tool¶

1. Choose your team.
2. Form your hypothesis.
3. Find and download datasets.
4. Explore variables.
5. Choose the target variable you want to predict.
6. Choose the feature variables you will use to predict the target variable.
7. Decide what constitutes one sample in your example training dataset. If your dataset is a table, what does one row represent?
8. Design an example training dataset you will use to train your model.
9. Prepare your training dataset.
10. Train your model.
11. Compare the performance of different sets of features and different models.
12. Design your tool using a tool template.
13. Create your tool and check that it works in preview mode.
14. Publish your tool and check that it works on crosscompute.com.
15. Set the visibility of your tool to PUBLIC.

Resources¶

Here are some video tutorials:

• Prepare and Fit Spatial Regression Models 20190117

Here are additional references:

• Scikit-Learn Documentation
• Spatial Data Science with PyData