LSTM and Word Embedding : Classification and Prediction of Puskesmas Reviews Via Twitter Tukino 1,a*) , Agustia Hananto2,b), Rizki Aulia Nanda3,c) Elfina Novalia 4,d) , Eko Sediyono 5,e), Jabar Sanjaya 6,f)
An initial level of community service for the benefit of the entire community is provided by the presence of a Community Health Center in a subdistrict or village. Therefore, in an attempt to improve service quality, patient feedback regarding service performance is required. patient input or reviews on social media platforms like WhatsApp, Facebook, Instagram, and Twitter. However, opinions expressed on social media are unstructured texts with a high volume- this makes it difficult to analyze the text and prevents one from comparing the quality of care provided by different community health centers. Aside from that, some community health centers do not have a single website that ranks health centers according to user interest, aesthetic appeal, and operational effectiveness. Consequently, the goal of this study is to categorize and display sentiment analysis from Twitter at Community Health Centers. Five factors are the focus of the scope: health worker skills, employee friendliness, finance, mechanisms, and administrative services. For text mining, word embedding makes use of word2v, fastex, the Bi-LSTM model, and the adadelta and adamax optimizers. The model created is assessed using a confusion matrix to determine the model^s level of accuracy in patient review classification and prediction.
Abstract
An initial level of community service for the benefit of the entire community is provided by the presence of a Community Health Center in a subdistrict or village. Therefore, in an attempt to improve service quality, patient feedback regarding service performance is required. patient input or reviews on social media platforms like WhatsApp, Facebook, Instagram, and Twitter. However, opinions expressed on social media are unstructured texts with a high volume- this makes it difficult to analyze the text and prevents one from comparing the quality of care provided by different community health centers. Aside from that, some community health centers do not have a single website that ranks health centers according to user interest, aesthetic appeal, and operational effectiveness. Consequently, the goal of this study is to categorize and display sentiment analysis from Twitter at Community Health Centers. Five factors are the focus of the scope: health worker skills, employee friendliness, finance, mechanisms, and administrative services. For text mining, word embedding makes use of word2v, fastex, the Bi-LSTM model, and the adadelta and adamax optimizers. The model created is assessed using a confusion matrix to determine the model^s level of accuracy in patient review classification and prediction.
