Description

https://www.medoid.ai/ai-solutions/forecasting/

Our multi-variate machine learning approach allows the consolidation of predictors from many different sources, from standard industry metrics to external data. Moreover, our expertise in Reinforcement Learning allows us to transform forecasts to actual decision policies, dynamic and adaptive by their nature. Our forecasts have been successfully deployed to the following industries and processes:

• Retail Demand
  We develop multi-variate machine learning forecasting systems that forecast in daily store/outlet level. Such a forecasting solution – build precisely for your needs and any underlying DB technology you have – gives you the competitive advantage over generic and proprietary, “lock-in”, solutions. Our take on retail forecasting is based on innovative thinking in every part of the process, from how we treat new items to how we include assortment information, campaign and promotion effects and integrating them with external data. Moreover, if you choose to implement some of our Customer Success models too – such as cart or customer segmentation – these models can work jointly, providing valuable information to forecasting.
• Energy Demand
  The analysis and exploitation of energy datasets using modern machine learning techniques leads to models that can accurately forecast future energy production by renewable energy sources, as well as future energy load demand. In addition, such models can form the basis for personalised energy consumption management. Machine learning has proven critical for optimising energy savings and reducing costs. Medoid AI has strong experience in building electricity and gas demand forecasting models using state-of-the-art machine learning algorithms.
• Stock Trading Metrics
  We develop models that forecast over high frequency trading market data, identifying critical time-series’ properties and forecasting return and mid-price in bid/ask prices per stock. We have worked on high granularity/high frequency order book data and we have developed advanced machine learning models capable of either segmenting or forecasting multi-variate time-series, also providing confidence levels.
• Transportation Demand
  We have successfully deployed models that forecast demand for public transportations buses using advanced machine learning methods. These multi-variate forecast models integrate several data sources, from smart sensor networks onboard to external weather data and smart data representations of routes. Moreover an adaptive scheduler using Reinforcement Learning transforms static forecasts to dynamic vehicle dispatch strategies!