Analisis Sentimen Ulasan Fitur Music Aplikasi Instagram pada Google Play Store Menggunakan Metode Convolutional Neural Network (CNN)

Instagram application is a social media (social media) to communicate from various circles that has many features, especially music features. Users have the right to pour criticism or suggestions in the form of text reviews regarding how the Instagram application works on the Google Play Store. From this problem, the problem of identifying reviewsabout the music features of the Instagram application that contain negative and positive sentiments taken on the Google Play Store with the web screpping technique can be formulated. Reviews taken to be a dataset of 2,260. The dataset is collected and manually labeled with 2,042 negative
data and 218 positive data. In addition, this study aims to measure the performance
of the model using the Convolutional Neural Network (CNN) method. Measuring
the performance level of the model by using 2 experiments, each of which has 2
model scenarios. The thing that distinguishes the two experiments lies in the
hyperparameters used in the model. The first experiment of the first scenario has
filter values 32,16,8 and the second scenario 256,128,32 using the value of 32 as
the value of the random state (K-Fold), 50 epochs and 20 batch sizes. The second
experiment of the first scenario has a filter value of 64,32,16 and 128,64,32 for the
second scenario filter value, a value of 42 random states (K-Fold), 80 epochs and
32 batch sizes. The results of the first experiment of the first and second models are
93% and 92%, while the second experiment of the first and second models are 97%
and 95%.