Build a Data-Driven Web App That Everyone Can Use
summary
Prototype web app turns a trained RandomForest (saved as model.sav via joblib/pickle) into an interactive demo that predicts crowdfunding amount and a funded/not-funded boolean from features like story word count and image count, reporting roughly 80%+ accuracy on the original laptop run. It uses Flask for routes and templating, loads features from a crawler or third‑party API, serves predictions without retraining, and renders dynamic charts by embedding base64-encoded matplotlib images into HTML. Covers model persistence pitfalls (trust, library versions) and simple deployment paths like Heroku.
AI-written summary · about
chapters (11)
- 0:00Introduction to web apps
- 1:11The case for data-driven apps
- 3:36Overcoming development myths
- 6:45Project overview and goals
- 9:33Flask framework basics
- 13:42Handling user input
- 17:09Integrating machine learning
- 23:22Pitfalls of model serving
- 25:33Dynamic data visualization
- 28:05Enhancing with Jinja2
- 34:35Deployment and next steps
related talks
01h 45mBuilding a Model Prediction Server
01h 18mBuild Data Apps by Deploying ML Models as API Services
01h 19mDatascience on the web
30mPython based predictive analytics with GraphLab Create
48mHow to make intelligent web-apps
48mHow to make intelligent web-apps
01h 17mOne workshop that data scientists don't want you to attend...
15mA Fiery approach to R web services