Skip to main content

Table 5 Evaluation on the youth development data set of SVM classifiers trained with topic features

From: Supporting systematic reviews using LDA-based document representations

Topic densitya Precision Recall F1-score Accuracy ROC PRC
Linear
2 0.15659 0.76767 0.28124 0.55268 0.685362 0.16959
5 0.16389 0.76986 0.28025 0.61253 0.7485 0.215196
10 0.21661 0.775616 0.334683 0.706321 0.782012 0.239912
20 0.22839 0.7767123 0.31605 0.66244 0.77806 0.276948
30 0.235857 0.772465 0.357992 0.732592 0.816773 0.288951
40 0.239157 0.7730136 0.36417129 0.73246 0.820795 0.320774
50 0.232558 0.7671232 0.35691523 0.72242 0.818062 0.371289
60 0.2494141 0.7771232 0.35320 0.717881 0.811705 0.338449
70 0.283706 0.7719178 0.407421 0.789108 0.841333 0.342748
80 0.27956 0.782191 0.3548788 0.782191 0.84254 0.359227
90 0.28068 0.77479 0.376366 0.748830 0.832757 0.345683
100 0.28137 0.786575 0.376121 0.751306 0.831486 0.358541
150 0.29082 0.79178 0.379514 0.740751 0.836747 0.367825
200 0.2949 0.79123 0.423232 0.77078 0.850254 0.40842
300 0.3224 0.72054 0.4558 0.9588086 0.847479 0.389575
500 0.3059 0.7082 0.4272 0.80935 0.844137 0.39549
RBF
2 0.151121 0.812328 0.2548345 0.5229 0.694288 0.168685
5 0.159186 0.826027 0.26693 0.54436 0.719956 0.194878
10 0.189766 0.802739 0.306966 0.635988 0.775362 0.201232
20 0.19948 0.7452 0.314723 0.674095 0.774715 0.253942
30 0.257261 0.679452 0.373213 0.77082 0.816608 0.312387
40 0.264912 0.6205 0.37131 0.78896 0.799286 0.301266
50 0.246453 0.641098 0.356534 0.767093 0.779354 0.250867
60 0.23598 0.57671 0.33492 0.76998 0.77882 0.250866
70 0.255531 0.49041 0.33532 0.80533 0.773743 0.237212
80 0.39523 0.39041 0.38255 0.873435 0.806185 0.318034
90 0.4092 0.219178 0.285459 0.889806 0.801959 0.312336
100 0.368421 0.019178 0.03645 0.898197 0.817434 0.319278
150 0 0 0 0.877579 0.812314 0.297883
POLY
2 0.153 0.82602 0.25818 0.5233 0.70262 0.170105
5 0.17164 0.14315 0.156498 0.843513 0.70445 0.166452
10 0 0 0 0.899574 0.285556 0.06007
  1. Items in italics refer to the highest scores obtained in a column
  2. aResults are reported according to different values of the topic density