Feature breakdown
Looking at this from a feature standpoint, we need to see this as consisting of three major parts.
- Data ingestion and training: this needs a separate interface, maybe it’s done without a user interface and it’s just static data being fed into code capable of training a model from it. With that understanding, we can outline the steps like so:
- Load data
- Clean data
- Create features
- Train model
- Evaluate model
- Run predictions
- Consuming the model: Once the model is trained, it needs to be exposed, preferably through a web endpoint. To get there, we think we need these set of steps:
- Convert the model to suitable format if needed
- Build a Web API
- Expose model through Web API
- Deploy model, there’s a step here where we need to bring the API online
- Prediction: For the prediction part, this is a functionality...
