OCI Forecasting includes processes to automatically address preprocessing, a cumbersome prerequisite for dealing with algorithms. It handles missing values through machine learning–based estimates, identifies and treats outliers, transforms data to improve quality, and aggregates temporal data to the required forecasting horizon (such as converting days to weeks and months).
OCI Forecasting includes a wide range of training algorithms, from commonly used statistical methods to complex algorithms for machine learning and deep learning. It automatically selects the best one based on the fewest rolling over cross-validation errors. With scheduling capacity, OCI Forecasting can include all the latest data points each time it runs to capture even small changes in the business. OCI Forecasting also provides the three best-performing model results for end users to observe.
Unlike its competitors, OCI Forecasting provides explainability as an output, which describes influential features at global and local levels in your data and brings transparency to forecasted results. It also provides confidence intervals, which are a range of variations in predicted values at any time. This aids in making decisions, taking variations into account.