Skip to main content

Table 12 Comparison of classifiers: classification accuracies (acc), mean precision (pre), and mean recall (rec) for classification on AllInst test data with the NoLyr feature set for the V3 and A3 tasks. Considered alternatives to SVM are Random Forests (RF, with 250 trees found optimal and minor differences in the range between 100–250), a K2 hill climbing structure-learnt Bayesian Network (BN), and k Nearest Neighbours with Euclidean distance (kNN, with k being 5 found optimal). Feature set NoLyr.

From: Determination of Nonprototypical Valence and Arousal in Popular Music: Features and Performances

Type

Valence

Arousal

%

acc

pre

rec

acc

pre

rec

SVM

58.5

57.6

58.8

53.3

52.6

54.1

RF

61.0

60.4

58.3

58.7

56.5

56.2

BN

51.6

51.0

53.1

52.9

51.3

54.0

kNN

45.4

46.8

47.0

44.9

45.3

46.0