Home > News > Tea Brašanac: Managing concept drift in machine learning models on financial data.

Tea Brašanac: Managing concept drift in machine learning models on financial data.

Date: 11. 12. 2017
Source: Seminar for probability, statistics, and financial mathematics
Četrtek, 14. decembra 2017, ob 14:15 v predavalnici 2.03 na FMF, Jadranska 21, Ljubljana.
V četrtek, 14. decembra 2017, ob 14:15 bo v predavalnici 2.03 Fakultete za matematiko in fiziko Univerze v Ljubljani na Jadranski ulici 21 v Ljubljani potekalo predavanje Tee Brašanac z naslovom Managing concept drift in machine learning models on financial data.
Abstract: Concept drift primarily refers to changes in data where the relation between the input variables and the target changes over time. With its dynamic nature and drifting target concepts, predicting the stock market is one of the most challenging tasks in machine learning. With the presentation I will shed light on the necessity for the concept drift recognition, furthermore we will explore the predictive power of several adaptive learning approaches on the U.S. stock market dataset. The machine learning model used will be random forest, which is a powerful ensemble classifier that combines the results of a series of decision trees. The predictive power of the proposed learning methods will be compared to the non-adaptive learning method. For the comparison purposes results of several backtests will be shown. Lastly, some recommendations for the further research will be given.