Creating a forecasting pipeline with AWS forecast in JAVA
In this day, forecasting is now a basic function of most data-centric platforms and with so many tools now available, you get to spin up forecasts easily.
6 min readOct 2, 2023
This article will be focused on creating an end-to-end forecasting pipeline using AWS forecast.
We will be predicting a fictitious company cash flow, let's call the company “X”. Company X needs to be able to get forecasts for their cash flow daily and monthly, they would also like to be able to compare between three percentiles, 10, 50, and 90.
Company X has up to 5 years of data and would like the daily predictions to go up to 14 days and monthly predictions up to 4 months to serve different team needs.
We will start by getting the data to be used for the forecast, see the sample data format below:
metric_name | metric_value | timestamp
endOfDayBalance 460751.47 2023-07-27 00:00:00
endOfDayBalance 272732.92 2023-07-28 00:00:00
endOfDayBalance 460751.47 2023-07-29 00:00:00…