Next generation demand forecasting

Major multinational retailer

Proof-of-concept development of a machine learning-based demand forecasting application for retail transformation

2014

IBM Global Business Services

Client's supply chain organization was building a proof-of-concept application leveraging machine learning to enhance their demand forecasting operations. The goal was to be able to accurately predict demand for products at each of their retail locations using a combination of unstructured and structured data sources such as social media feeds, weather forecasts, market disruptions, and historical trends. We supported the data aggregation, preparation, and analysis of the machine learning model that powered the application. Work resulted in successful proof-of-concept for a select retail location, with improved demand forecasts from model outputs compared to baseline forecasting practices.

Natural language processing ontologies

Created ontologies to create associations between common keywords and specific retail products for natural language processing model. Ontologies were used to convert web scraped unstructured social media and weather data into quantifiable metrics for model input.

Data analysis

Analysis of outputs from Watson Analytics to assess accuracy of machine learning-based demand forecasts against realtime consumer sales. Continuous adjustments made to ontologies used for text sentiment analysis and language classification.