Mid-Curve Recommendation System: a Stacking Approach Through Neural Networks

摘要: Derivative traders are usually required to scan through hundreds, even thousands of possible trades on a daily-basis; concrete case is the so-called Mid-Curve Calendar Spread (MCCS). The actual procedure in place full pitfalls and more systematic approach where information at hand crossed aggregated find good trading picks can be highly useful undoubtedly increase trader’s productivity. Therefore, this work we propose an MCCS Recommendation System based stacking Neural Networks. In order suggest that such methodologically computationally feasible, used list 15 different types US Dollar MCCSs regarding expiration, forward swap tenure. For each MCCS, 10 years historical data ranging weekly from Sep/06 Sep/16. Then, started modelling stage by: (i) fitting base learners using as input sensitivity metrics linked with time t, its subsequent annualized returns output; (ii) feeding prediction model particular stacker; (iii) making predictions comparing methodologies by set performance benchmarks. After establishing backtesting engine setting metrics, our results proposed Network stacker compared favourably other combination procedures.