Hybrid Human and Machine Learning Algorithms to Forecast the European Stock Market

Table 4

Community membership of the network of classifiers based on the top 5% of common news topics for 3 hours return according to Clauset et al. [86]. NB, RF, LR, SVM, and EW are naïve Bayes, random forests, logistic regression, support vector machine, and expert weighting algorithm, respectively. MCC is the community average of the Matthews correlation coefficient. Modularity = 0.08.

I.D.	Components	MCC	Ranking of main topics by categories
1	NB, CART, RF, AdaBoost	0.07	(1) Geopolitical unit, (2) Events (results forecasts)
2	Expert, trained, LR, SVM	0.09	(1) Events (results forecasts, mergers, and acquisitions), (2) Geopolitical unit
3	EW, EWh, EWa, best crowd	0.12	(1) Geopolitical unit, (2) Business sectors