Current ongoing projects
AI for demand forecasting
In this project I work with several companies to improve their demand forecasting process. Major players like Amazon or Walmart are aware of the potential of AI for demand forecasting and have large teams to do it. A major challenge for SMEs when deploying AI is the limited availability of historical data. On the one hand, companies don’t keep up with years of sales data and on the other hand, markets are changing rapidly making old data less and less relevant. Nevertheless, there are many techniques to use the hidden structures of the business dataset to make AI, neural networks and machine learning relevant here.
AI for students
In this project, we built a Convolutional Neural Network model that predicts a student’s pass or fail and exact exam score early in the semester. This allows the system to intervene, inform the lecturer (anonymously) or give the student a helping hand by e-mail.
Software packages
Grafton
Grafton is a python package that can easily anonymise very large files with an anonymisation key and a informed consent list. Light weight and speed is key. This helps us to do research with GDPR compliance. Check it out here
Greybox
This R package has many functions and instruments for regression model building. It focusses on variable selection for time series data, including promotional modelling, selection between different dynamic regressions, selection based on cross validation, solutions to the fat regression model problem and more. Check it out here
Leading indicators
An R package that identifies the most relevant leading indicators for your time series. It is tailored for demand forecasting and works well with a small training set and a very large database of exogenous indicators. This has been applied in business and reported a 15\% improvement in Mean Absolute Percentage Error (MAPE) in forecast accuracy. Contact me here to get started.
Inventory
The improvement in demand forecasting does not stop at the forecast accuracy level. The forecast is a central input for the decision-making process and supply chain management. In this inventory package, we can evaluate the impact of an inventory-level forecasting model. This is done through inventory simulations and also works for complex hierarchical product families.