All price prediction is done in a web browser. Use a web server to serve the page: ex. python -m http.server [port]. The page will be loaded to localhost:[port].
To modify the trained model used to predict prices, replace model.json with a Keras trained model. See TensorFlow.js for more information.
Note: The file containing the weights of the associated json model must also be included when using a new model.
The TensorFlow models were trained in python before being ported to TensorFlow.js. To retrain a model, use the program found in the python folder.
This code can be modified and reused to train with new datasets. The trained model can be saved as model.json and used to make predictions (see above for more details).