A great advantage of the traditional econometric models1 at Ortec Finance, is the explainability of individual predictions, e.g. your house is worth more than your neighbor’s house, because it is 10m2 larger. This makes the valuation consistent and transparent, which is essential for legal issues.
Machine learning (ML) has taken a flight over the last few years, and Ortec Finance has introduced ML models in a variety of its applications. In the case of real estate valuation, an advantage is that a specification of the model structure is not required, which simplifies the addition of variables to the model.
In terms of accuracy of real estate valuation, ML models are currently comparable to traditional models. A drawback, however, is their lack of transparency regarding the variables’ influence on the valuation, especially on the level of individual predictions. Recent developments in the literature on ML model explainability seem promising23. However, as the number of variables increases, or feature engineering becomes more complex, proposed methods start to fail to comply with the properties4 of a good explanation.
The aim of the project is therefore to:
- Obtain an overview, from academic literature, on methods for ML model explainability,
- Assess these methods in the case of individual real estate valuations,
- Critically evaluate the limitations of the methods and identify the causes,
- Propose solutions.
Requirements for this project are:
- Familiarity with basic machine learning methods,
- Affinity with advanced statistical analysis and/or data science,
- A critical mind and strong research capabilities,
- Preferably knowledge of advanced regression modelling techniques.
- 1. M. Francke. The hierarchical trend model. Mass Appraisal Methods. An International Perspective for Property Valuers, pp. 164-180, 2008.
- 2. M. T. Ribeiro, S. Singh & C. Guestrin. “Why Should I Trust You?” - Explaining the Predictions of Any Classifier. arXiv, 2016.
- 3. S. M. Lundberg, G. G. Erion & S.-I. Lee. Consistent Individualized Feature Attribution for Tree Ensembles. arXiv, 2019.
- 4. C. Molnar. Interpretable Machine Learning. 2020.
What we offer
We offer a challenging and inspiring work environment with excellent career opportunities in both specialized positions and management. The exact scope of the thesis will be discussed before the thesis starts - we are flexible when it comes to the exact scope and topic. Ortec Finance gives you the opportunity to combine IT, mathematical models and specific market knowledge in your work.
In addition, we will provide:
- Required devices and tools,
- Office-space in Rotterdam or Amsterdam,
- Guidance from Ortec Finance R&D Labs.
To consider your application please send us:
- An up-to-date CV,
- A cover letter with your motivation to work on this topic at Ortec Finance,
- Information about requirements in terms of start date, duration of project, etc.,
- Grades list.
More information or to apply?
For more information regarding this position please contact HR via HR@ortec-finance.com. To apply, please use the application button to send us your CV, cover letter and additional information.
We are looking forward to your application!