Auto Forecasting

Learn more about Moselle's AI forecasting and the various models you can utilize to automatically forecast and plan your inventory for the upcoming 12 months.

AI-Generated Unit Forecast

Moselle offers different machine learning forecasts to help you automatically predict the next 3 to 12 months, depending on your plan. These models are excellent starting points for your business and intelligently learn from your historical sales data. As you continue to use Moselle, the forecasts will improve over time.

PulseCast (New Model)

PulseCast is Moselle’s basic ensemble model. It blends three forecasting models—SARIMA, Holt-Winters (HW), and Prophet—into a single forecast using equal weighting.

This model is optimized for:

  • Steady sales from week-to-week or month-to-month.

  • Items with more than 2 years of sales history.

  • Capturing seasonal patterns (E.g. Holidays, End of season sales, etc).

  • Items with occasional zero-sales periods.

ARIMA Weighted Ensemble

This model blends multiple ARIMA forecasts, each capturing time-series structure via lag, trend, and seasonality. Deep learning is used to tune the combination for best accuracy.

This model is optimized for:

  • Items with at least 2 years of data with regular sales activity

  • Predictable patterns that repeat over time (E.g. Seasonal Trends)

  • Items with few long gpas with no sales

Chronos

Customer-Friendly Summary:

A lightweight model focused on speed and transparency. Combines fast ARIMA-style logic with trend and seasonality smoothing.

Best for:

• Short history — just 6 to 9 months of data

• Simple, stable patterns

• Fast results needed for quick decisions

Statistical Models

Ensemble 1 - ARIMA weighted model

This is a statistical model that uses lagged moving averages to predict future values. The model employs deep learning to automatically tune and blend various forecasts, providing the most accurate prediction for your business.

Ensemble 2 - Exponential Smoothing weighted model

This is a statistical model that uses lagged moving averages and seasonal trends to predict future values. The model employs deep learning to automatically adjust and combine various forecasts, offering the most precise prediction for your business.

Probabilistic Models

Tuned Probabilistic Model - Prophet-based model

A model for nonlinear trends forecasts yearly, weekly, and daily seasonality, as well as holiday effects. Examples of holidays that can be incorporated into this trend include Amazon Prime Day, Black Friday, and Boxing Day. The model uses deep learning to automatically tune and blend various forecasts, providing the most accurate forecast for your business.

The model can also be influenced by custom data, such as a marketing calendar or spending, which you can upload.

Deep Learning Models

Temporal Fusion Transformer (TFT)

This model is specialized in deep learning models for time series forecasting. It is designed to handle complex patterns by incorporating multiple types of data (historical trends, seasonality, events, and holidays). It can also provide insights into which variables or time periods drive forecasts.

Last updated

Was this helpful?