Page 256 - 《软件学报》2020年第10期

3232 Journal of Software 软件学报 Vol.31, No.10, October 2020

Tabel 4 Accuracies of the feature dictionaries built under eight different conditions (%) (Continued)
表 4 8 种不同条件下建立的特征字典对应的模型分类精度(%)(续)
特征评价指标 tf-idf tf-idf tf-idf Chi tf-idf Chi Chi Chi
阈值 Dynamic 0.3 0.3 3 0 3 2 2
单词长度 Variable Variable Variable Variable Variable Fixed Fixed Fixed
n 0 语法模型 Unigram Unigram Bigrams Unigram Unigram Unigram Unigram Bigrams
Dataset TDVU T0.3VU T0.3VB C3VU T0VU C3FU C2FU C2FB
ProximalPOAG 84.6 85.6 85.8 86 84.4 84.4 83.9 83.9
ProximalPOC 89.9 89.3 88.7 89.6 90.1 86.9 88.3 88
ProximalPTW 80.7 78.2 82.1 80.2 79.4 81 81.5 78
RefrigerationD 57.8 51.4 53.2 52.5 51.6 51.2 51.6 52.8
ScreenType 52.6 55.1 53.9 53.8 55.8 56.8 58.5 53.3
ShapeletSim 100 100 100 98.8 76.5 100 96.1 100
SmallKitchenA 76.3 77 76.7 76.5 78.7 76.3 75.2 75.7
SonyAIBORobotS1 84.1 82.7 78.9 82.8 82.7 79.5 85 81.9
SonyAIBORobotS2 95.6 94.6 94.4 91.3 95.6 89.2 93.7 94.8
Strawberry 97.3 97.8 97.6 97.5 97.3 98.1 97.8 97.6
SwedishLeaf 96.4 96.3 96.2 96.3 96.5 95.8 96.3 96.4
Symbols 95.9 95.7 96.7 95.8 96.6 95.5 95.5 96.1
SyntheticControl 98.8 99 98.7 98.8 98.6 99.3 99.3 99
ToeSegmentation1 96.5 94.3 94.8 94.3 94 91.2 89.5 95.2
ToeSegmentation2 89.5 85.9 89.4 83.2 90.2 78.5 81.5 82.3
Trace 100 100 100 100 100 100 100 100
TwoLeadECG 99.9 99.9 99.9 99.9 99.8 92.2 95.6 99.4
TwoPatterns 99.3 99.2 99.3 99.3 99.3 99.1 99.1 98.9
Wafer 100 100 100 100 100 100 100 100
Wine 86.4 89.6 89.3 85.9 87 70.4 75.9 85.6
WordsSynonyms 70.7 71 70.8 71.5 70.3 71.3 70.8 71.6
Yoga 91.6 90.7 92.2 87.2 90.9 84 85.8 90.4
平均值 84.6 83.9 84.1 83.5 84.3 82.2 82.8 83.7
最大值个数 22 14 15 13 18 12 13 17
上面表 4 中我们将各特征字典构建模型符号表示中的 TfIdf 简记为 T,Chi 简记为 C,Dynamic 简记为 D,
VLWEA_X 和 FLWEA_X 分别简记为 VX 和 FX,X 表示使用的 n 0 元特征生成模型,U 表示 Unigram,B 表示
Bigrams.例如,模型 TfIDfDynamicVLWEA_U 记作 TDVU.表 4 中每行加粗数值表示对应行的最优值.

Fig.6 Comparison of accuracy between VLWEA and 6 BOP models on 65 datasets
图 6 VLWEA 和 6 个 BOP 模型在 65 个数据集上的分类精度比较