Amazon Web Services Inc. today announced the release of the Amazon Forecast Weather Index, which will help developers increase predictive accuracy of machine learning models by including local weather information into demand forecasts for retailers and service providers at no extra cost.

Weather conditions can affect many industries, especially consumer buying patterns, such as product merchandising decisions, staffing requirements, energy consumption needs and numerous other decisions made by businesses. It’s an area that machine learning can be applied to great effect.

For example, although it might feel like common wisdom that rainy weather would drive customers indoors, it does tend to make people feel a greater need than on sunny days, by 12%, to buy more furniture, wholesale and clothing online.

For retail inventory management, these day-to-day variations in foot traffic can be significant when it comes to the supply chain because a single snag can lead to massive losses and returns. By taking weather conditions and forecasts into account, and therefore the need for showroom floor, warehouse space or fleet management.

Amazon Forecast Weather Index combines many weather metrics from historical weather events and current forecasts to create as accurate a model as possible. It then provides the hooks and a dashboard that requires no prior ML experience. Forecast is a fully managed service that doesn’t require rebuilding entire systems to take advantage of its capabilities.

Predicting the weather for businesses is not a new thing in the industry. IBM Corp. acquired the Weather Co. in 2015 to roll it into its own machine learning-focused unit, IBM Watson. IBM currently provides an AI-based tool that allows companies to predict how fluctuations in weather will impact business, similar to Amazon Weather Index, called IBM Weather Signals.

“At Peak, we work with retail, CPG [consumer packaged goods] and manufacturing customers who all know that weather plays a strong role in dictating consumer buying habits,” said Tom Summerfield, the director of Retail at Peak.AI, an accessible artificial intelligence system to improve business efficiency. “Variation in weather ultimately impacts their product demand and product basket mix. Our customers frequently ask us to include weather in their demand forecasts. With Amazon Forecast adding a weather feature, we are now able to seamlessly integrate these insights and improve the accuracy of our demand planning models.”

The demand for weather data across the industry has been present for years and the need for historical data can be overwhelming. However, acquiring and actually managing that historical data can also be beyond the reach of small businesses.

As a subscription service, Forecast will train predictive models that will tap historical models and provide hyperlocal 14-day weather forecasts that are influenced by day-to-day variations and provide accurate demand forecasts can be modeled alongside data provided by data sourced from company ledgers. As a result, ML models can be produced that combine weather with consumer behavior.

Developers interested in getting started with the Amazon Forecast Weather Index can see the Weather Index portal and go through the notebook in Amazon’s GitHub repo that walks through the application programming interface. There’s more information about regional availability on the AWS Regional Services website.

Photo: Pixabay