Mini-blog 

How do you sift for anomalous real estate valuations from a pile of pdfs?

We ran an exercise on Kronofogdens publicly available foreclosure property descriptions offered as pdfs, were we ingested a series of 'värderingsdokument' (valuation document) and mapped different metrics in this document (old school NLP algorithms that ranks the sentimental context; positive/negative for property for valuation). The algorithm does not valuate itself but maps to detect discrepancies. So the purpose is to flag the properties with valuations anomalous with respect to the rest of the dataset, and hence worthy of an investment manager's attention.

Learn more

Let underwriters drive.

Even with a scenario modeling engine, underwriters and investors are *less* likely to trust your scenario model when raising capital over theirs, given natural biases and resistance to novelty. So a seemingly radical alternative is give them access to the engine and slide the controls over to them. Give them the deal data, your financial data, and the engine and let them make themselves home.

That way they adjust the variables *they* deem crucial - things like valuation uncertainties, interest rates, performance and liquidity risk. They also choose the bottom line which may not be the same as yours; they maybe modelling for illiquidity risk, DCS and you are more into cash-on-cash or RoE.

Your risk assessment model is not always aligned with theirs. Give them the engine wheel (not the report) and let them drive.

Learn more

Therapy

Paragraphs are the main building blocks of web pages. To change what this one says, just double-click here or hit Edit text. You can change the style here, too. Of course, dragging the text will move the paragraph on the page.

Learn more