International Core Journal of Engineering 2020-26 | Page 60

performance of the model, the accuracy rate, the recall rate,
and the F1 value to evaluate the model.
From Table Ⅰ, it can be seen that Graph LSTM is superior
to SVM and LSTM in machine learning in terms of accuracy,
recall rate and F value. In terms of runtime, the Graph LSTM
runtime is close to that of the SVM. If the LSTM is to achieve
better accuracy, the number of iterations will be higher, so the
time is the longest.
x TP: Number of cases with positive predicted values
and positive values
x TN: Number of cases with negative predicted values
and negative actual values
R EFERENCES
x FP: Number of cases with positive predicted values
and negative actual values
[1]
x FN: The number of cases with a negative predicted
value and a positive value [2]
x Accuracy: A (Accuracy) = (TP + TN) / (TP + TN + FN + [3]
FP)
x Recall rate: R (Recall) = TP / (TP + FN)
[4]
x F1 value: F1 = (2P * R) / (P + R)
B. Experimental results
The experiment uses intel I3, and the comparative machine
learning methods include SVM (Support Vector Machine),
LSTM. SVM is implemented by the pycharrm installation tool
sklearn, which is implemented by the python installation tool
keras. The LSTM is trained in small batch random gradients
with a batch size of 15 and a maximum of 30 iterations.
[5]
[6]
TABLE I. C OMPARISON TEST RESULTS
Accuracy Recall rate F value
Graph LSTM 85% 0.94 0.90
SVM 79% 0.90 0.84
LSTM 75% 0.82 0.78
38
Tai KS, Socher R, Manning CD. Improved Semantic Representations
From Tree-Structured Long Short-Term Memory Networks. In
Proceedings of the 53rd Annual Meeting of the Association for
Computational Linguistics (ACL 2015), 1556-1566, 2015.
Brueckner R, Schulter B. Social Signal Classification Using Deep
BLSTM Recurrent Neural Networks[C].ICASSP.2014:4823-4827.
Ghosh S, Vinyals O, Strope B et al. Contextual LSTM (CLSTM)
models for large scale NLP tasks. Neural Networks.2016, 2(12):1127-
1136.
Turney P, Littman ML. Measuring praise and criticism: Inference of
semantic orientation from Association. ACM Trans. On Information
Systems, 2013, 21(4):315−346.
Read J. Using emoticons to reduce dependency in machine learning
techniques for sentiment classification[C]. In Proceedings of the ACL
Student Research Workshop. Association for Computational
Linguistics, 2015: 43-48.
Brueckner R, Schulter B. Social Signal Classification Using Deep
BLSTM Recurrent Neural Networks[C].ICASSP.2014:4823-4827.