<![CDATA[Accuracy in analyst sentiment classification is very important so that the trained model can be implemented well to make business decisions. Researchers proposed a method for configuring neural network models arranged in parallel to improve classification accuracy. The results of the first stage, a bidirectional long short-term memory (Bi-LSTM) algorithm with Keras embedding with a sequential layer configuration, produced the best accuracy of 80.20%. The results of this first stage served as the baseline to be used as a reference for the combination in the second stage of the experiment. In the second stage of the experiment, a combination of the Bi-LSTM algorithm with other algorithms was carried out in parallel, such as gated recurrent unit (GRU), recurrent neural network (RNN), and Simple RNN with Keras embedding. It was found that the combination of three parallel layers of GRU-BiLSTM-RNN with Keras Embedding produced the highest accuracy for sentiment analysis of three classes, with a value of 88%. A statistical test of the t-test method was carried out with a critical p-value of 0.05 to prove the accuracy that has been produced between the sequential and the parallel configuration. The results of the t-test between the sequential configuration and the parallel configuration obtained a p-value of 0.5e-9 which is much smaller than the critical p-value of 0.05 so that in statistical testing the average accuracy produced from the two configurations is significantly different.]]>
Copyrights © 2025
