Concept: Labor Optimization Machine Learning Process

The Labor Demand Forecasting Engine uses a structured, multi-step process to convert your raw historical data into reliable labor demand forecasts. Your data is used to build a robust statistical understanding of your business's needs, leading directly to optimized staffing decisions. Your data never leaves the Workday environment.

Step	Description	Data Handling
Data Intake & Analysis	Your data is securely stored in a central data hub where the ML Engine picks it up and initiates an Exploratory Data Analysis (EDA).	We analyze the raw data to understand its attributes, trends, and fundamental characteristics.
Preprocessing & Cleaning	This step ensures the data is clean and consistent for training the model. The data is processed through several automated actions.	We consolidate data, remove duplicates, fill missing intervals with zeros, perform outlier detection and replacement, clean inconsistencies, and add scheduling events (e.g., holidays or promotions).
Model Training	The ML Engine then trains the model using the cleaned, preprocessed data.	The model looks for patterns and trends based on daily metric data across all of your defined scheduling organizations. This is where the model learns the historical relationship between your data and demand.
Prediction Generation	Once trained, the ML Engine generates the daily forecasts. Predictions are generated using either a Trend-based Forecasting model or a Gradient-Boosted Tree model, depending on the best fit as determined at onboarding.	The ML Engine generates daily predictions for each Scheduling Organization using the scheduling organization data and the trained model.
Prediction Distribution	The daily predictions are then broken down into smaller intervals to match your operational needs.	The ML Engine uses the scheduling organization's historical patterns to redistribute the total daily prediction into your configured interval (e.g., hourly, 30-minute, or 15-minute segments).
Output & Application	The final, interval-specific predictions leave the ML Engine and are stored for reporting.	The labor mapping rules then use these predictions to calculate the required labor demand, which is finally used by the Labor Optimization Engine to create best-fit schedules for your teams.