Natural Language Processing for Marketing

Promotion is one effort to increase consumers. The use of the Internet as a promotional medium has been widely used by companies because it can save costs in reaching consumers. Social media can be a platform for marketing, one of the social media that is widely used today is Twitter. However, promotions often fail if they are given to the wrong consumers. So it is necessary to identify consumers correctly when carrying out promotions. This research aims to identify users who are looking for a place to study, by analyzing messages sent via social media. In this research, the dataset from Twitter is processed first so that it is free from noise. The initial stages carried out in the research were deleting messages in the form of advertisements, filtering so that the data was clean of hashtags, links and symbols other than letter characters; tokenizing; stemming; remove stop words so that the data is clean of words that are felt to be unrelated to the topic. As a result of the initial stage, the dataset was then subjected to feature extraction using TF-IDF. The results of data set feature extraction are fed into machine learning to determine which users are looking for campuses and which users are just talking about campuses. The final result of this research is to get usernames from social media which will be promotional targets. The testing carried out in this research is more about testing the accuracy of machine learning to determine the right users who will be potential customers from campus. In this research, cross validation is used. The accuracy of Support Vector Machine (SVM) and Naive Bayes (NBC) reaches 96%. Even though in terms of accuracy it is very good, the recall and precision values are 0%. So the research objectives were not achieved. This happens because the dataset is unbalanced. Therefore, in future research it is necessary to consider the imbalance dataset problem.