Integrated with your Supply Chain
Strategic planning
By analyzing future demand trends, your company can identify the key factors influencing demand. This enables better planning of production and purchases, leading to more informed decisions that maximize return on investment.
Use stock demand estimates to plan replenishments
Forecast demand can easily be entered into the material requirements engine to prepare your production processes and inventories. With this, when the actual demand arrives, your system will be ready to cope.
Configure regressors
Incorporate key variables that impact demand—whether from the ERP or external sources—such as product prices or marketing investments, and set their future values. This allows the models to detect patterns and predict how demand will evolve based on these future conditions.
Define your Forecasting Models
Account for seasonalities
Models can be configured to detect yearly, weekly, and daily seasonalities, enabling them to identify patterns and better understand the factors driving demand shifts.
Define special dates
Define special dates, such as holidays, where demand follows distinct patterns. This allows the models to recognize and learn from past trends, improving their ability to predict future demand shifts when these events occur again.
Set uncertainty intervals
Compared to point forecasting, uncertainty modeling estimates a range of possible values for the forecast rather than a single outcome. The confidence level determines the likelihood that the true value falls within this range. With this tool, when you introduce a confidence level you will obtain this interval in addition to the most probable value. This way, your company can plan and prepare for the range of possible outcomes that may occur.
Check the training process
The model training process is logged, allowing you to review metrics such as Mean Squared Error (MSE), Root Mean Squared Error (RMSE), and Mean Absolute Error (MAE) over epochs, both for training and validation. These insights help refine and improve model performance in future training sessions.
Choose between different frameworks
The forecasting tool integrates two different forecasting engines that make forecasts simpler and accessible to non-specialized users:
Prophet: A time series forecasting tool that uses an additive model to capture non-linear trends with yearly, weekly, and daily seasonality, along with holiday effects. It performs best on data with strong seasonal patterns and multiple seasons of historical records. Designed for robustness, Prophet effectively handles missing data, trend shifts, and outliers.
Neural Prophet: NeuralProphet combines traditional time series algorithms with deep learning techniques in a user-friendly package built on PyTorch. It supports global modeling of multiple time series, automates hyperparameter selection, and provides plotting utilities for visualizing forecasts and model components. NeuralProphet effectively manages changing trends, smooth seasonality, and the impacts of events and holidays, while offering customization options like regularization.