geeksessions-streamlit, May 22, 2024

streamlit dashboard kickstarter

• Minimalistic way how to create streamlit interactive dashboard using exemplary ML Iris dataset.
• With 130 lines of code, you train model, visualize results, and create interactive environment for the user to look into your KMeans clustering.

Before running the app

• to investigate how the codes work, play with the jupyter NB
• Btw. Colab has free GPU support, if you ever need it for ML testing.
• This can be done using the links below, no python needed.

	Run	Run	View
Iris dataset

To run the app

1. Clone the repository

Running locally

• pip install -r requirements.txt
• streamlit run app.py
• In the upper right hand corner is option to deploy the app.

No python

1. You can still deploy and edit with codespaces without much testing though.
2. Go to streamlit
3. Sign in with GH, gmail or other
   • If you want to run dashboard from your organization, follow instructions here
4. Create app
   • repo: Py-ualg/geeksessions-streamlit
   • branch: main
   • Main file-path: app.py
   • app URL: UP-TO-YOU.
5. You should have the app functional and running

Two ways to continue

Expand the dashboard capabilities individually

1. For each input, in addition to classification, show also the probabilities of classes
   • KMeans are not good for that, perhaps try XBoost classifier.
2. Model tab: How to visualize the 4D data:
   • pairplot (seaborn library)
   • 3D + color hue?
   • Confusion matrix (for train and test data), docs.
3. Prediction tab: Visualize input values against testing data using the plotClusters function

Group work: Given that not all must be interested in streamlit

1. Let's make interactive dashboard which is useful for GeekSessions community
2. List of active tech people in Algarve
3. Events
4. Contacts and calendar
5. Skills for collaborating

We need

• Database (tables in pandas for simplicity)
• visualization (maps, calendar, links)
• search