Technical Analysis of Twitter Data in Preparation of Prediction using Multilayer Perceptron Algorithm
Pilita A. Amahan, Mia V. Villarica, Albert Alcause Vinluan · 2021
Social networking sites have been the partner of everyone today to express feelings and emotions, especially that physical communication has been a bit prohibited caused by the pandemic. The platforms like Twitter, Facebook, etc., has been the choice to express words relating to stress, anxieties, and depressions to which the World Health Organization states that once not paid with enough attention may lead to mental health issues. A depression today stands out to be the leading problem known as mental health disorders that if not to anticipate earlier may also lead to this so called self-harming. In this regard, this study wants to illustrate the technical analysis of twitter data in preparation of prediction using the Multilayer Perceptron (MLP) in Weka algorithms to help the data mining community to dig knowledge from the stored historical data. The technical analysis has been made through the process of extraction, validation, and preparation of the model ready for interpretation and evaluation of predicted model. The legality of data has been permitted by twitter developer's account that permits the study to extract 1000 tweets and eventually validated 931 tweets necessary for exploration. Optimization was made by setting the epoch to 200 and by testing the eight attributes to 70:30 split of test. The initial result of 79.9141% was optimized to 82.0789% accuracy rate and found reliable based on the result of kappa statistics. However, the study still suggests to explore on varied parameters to increase the reliability of data.