Opinion mining toward work from office policies on post-pandemic covid-19 by using supervised learning

Abstract

Post-pandemic COVID-19, many companies have re-implemented work from office policies for their employees. However, the policy has been controversial among social media activists, especially in Twitter. The sentiment arose according to them, during the pandemic COVID-19 they believe that working from home has many advantages over working in an office. The emergence of 'work from office' sentiments is an interesting target for opinion-mining research. Opinion mining or sentiment analysis is a general research area of ​​data mining that helps to explore and analyze existing views and opinions to obtain useful information. The analysis process involves the use of machine learning with several supporting algorithms. This study used four classification algorithm models of supervised learning, including naive Bayes, support vector machines, k-nearest neighbors, and random forests.  The selection of those algorithms also aims to find out which model produced a good performance for the results. The performance results of each model were evaluated by the confusion matrix and the k-nearest neighbor algorithm model with an accuracy value of 96.62% was found to give the best results and to be the most used model in the classification process. On the other hand, the algorithm model that obtains the lowest accuracy is a random forest with 72.08%.