The training and test are automatically generated for a cross validation, so you can just train and test with no data source. In the training phase you need to open your training data in the explorer (not recommended) and use Create and select to train your model. When the training is finished, you can close and continue to the next step. Once the training is finished, the test set is automatically generated for you. To visualize you can open your Test set in the explorer and you can find the visualized data in the Exploring Data section. This process is simplified in the Explorer. The prediction method is executed for the test set. Kaggle provides a tool for offline analysis of your model, including accuracy, F1-Score, ROC, and precision/recall. For real time analysis and making predictions based on users' query, you can use the model, a trained model, or a public model. This process is called a real-time prediction. To do real-time prediction you need to store a trained model, and the data of the prediction and query in a database. You can learn more about this from the section on Prediction based on data sources.
Model Versioning and deployment
Model Versioning and deployment are also important concepts in Kaggle Notebooks. A model version refers to the combination of a notebook and a trained model. In the Exploring Data section, you can see the version of the notebook and model you are using. If you want to deploy the model in production, you need to create a new version that includes a trained model and a notebook. A trained model means a machine learning model that is already trained using a dataset you provided to Kaggle. A notebook refers to an example of how to train and run a machine learning model. For example, in the Exploring Data section, there are 4 notebooks to explore and learn how to run machine learning models. Kaggle provides a method to deploy models in production. A deployed model in production means a model that has been trained and then is made available in production. Using the process and technology of a deployed model in production, you can collect real-time data from a device in your network.
Other Kaggle Notebooks
A Kaggle Notebook is not a single model. There are other Kaggle Notebooks that are based on other models. In the Explorer, you can see the list of Kaggle Notebooks in the Explore section. If you want to learn how to use these
Related links:
Commenti