Contact us | Terms of use | Privacy Policy | |


A strong quantitative expertise... but not quant fundamentalists!

TAC has a very strong « quantitative bias », meaning that we collect, organize, mine and treat data on a large scale with tools that range from the most traditional to the most advanced and sophisticated. However, we often say that we are not « quant fundamentalists ».

We keep a wide-open eye on results from our modeling and data-mining exercises and stand ready to challenge them and explore alternative routes and methods. But we also believe that the data analysis and modeling techniques give us an enviable unbiased view on (often hotly debated) issues and they ensure a strong consistency in our analytical process. The models also allow us running different simulations and alternatives, while our top-of-the-market non-linear techniques are the only ones able to consistently deliver better advance signals of major breaks or systemic shocks.

Economic Modeling & Data Mining

A large range of modeling techniques are usually involved in the different quantitative exercises at TAC. We are definitely familiar with traditional/well-known techniques, such as time series modeling, panel estimates, volatility models, cointegration, structural VAR,... as well as more advanced/sophisticated approaches: GMM, Error Correction Models, Kalman filters, markov switching models, difference in differences estimations, wavelets analyses, etc...

TAC has also developed strong skills in classification methods, data mining and text mining techniques, from simple principal component analyses or discriminant analyses to the more complex recursive partitioning algorithms (CART), genetic algorithms, simulated annealing, random forest, neural networks, self organizing maps, support vector machines (SVM), etc...

Large databases & web plateforms

We construct, organize and use very large databases on economic, financial and industrial indicators across all continents and countries. The largest part of this work is for our own internal use, but we have made some of the databases public, usually associated with a strong analytical content: