Batch Model Serving

The batch model serving pattern is more common than real-time serving in today’s world of large-scale data. With batch model serving, newly arrived data will not be immediately available in the model. Depending on the batching criteria, the impact of new data may become apparent after a certain period of time, ranging from an hour to a year. Batch model serving uses a large amount of data to build a model. This gives us a more robust and accurate model. On the other hand, as the batch model does not use live data, it cannot provide up-to-date information during prediction. For example, let’s consider a product recommendation model that recommends shirts, and the model is retrained at the end of every month. You will get a very accurate shirt recommendation based on the available shirts up to the last month, but if you are interested in shirts that have recently appeared on the market, this model will frustrate you, as it doesn’t have information...