Clustering on ‘Z’ e-learning application’s reviews using self organizing maps and k-means methods

“Z” is one of the famous online learning applications in Indonesia which has an increasing number of users during the pandemic. Based on these conditions, “Z” is required to improve their services to keep user satisfaction. Reviews are great tool to get information about the user experience. The number of reviews obtained may be very large and contain various information. Reading all the reviews will take a long time and effort, so it needs a method to summarize the information, and one of the methods is to apply clustering analysis. This study groups the reviews taken from Google Play on rating 1 until 5, which aim is to obtain more detailed information about the weaknesses and strengths of the application. This study also compares performance of clustering using Self Organizing Maps (SOM) which is neural network-based algorithm and K-Means which is centered-based algorithm. Clustering is done using the bigram model to get a clearer meaning. Clustering reviews on each rating successfully getting much information regarding the good or bad user experience, and from those information can be given some recommendations for improvement. Based on Silhoutte and Davies Bouldin Index shows that K-Means gives better result than SOM.