Predict Individuals’ Behaviors from Their Social Media Accounts, Different Approaches: A Survey

Abstract

Predicting individual behavior has been among the key objectives in the social sciences, helping derive important insights, such as the individuals to target with specific marketing material. For many years, rudiments behavioral, geographic, and demographic methods were used for prediction. However, different researchers have developed advanced, computerized methods to quantify emotional and sentimental intensity from social media posts. Therefore, the research employed a literature review methodology to determine the main approaches proposed in the last five years. IEEE Xplore was used to select reliable research articles about different prediction methods. Four techniques were identified: the Lexicon approach, the Louvain algorithm, Naïve Bayes classification, and MCDM. Based on collected information, the Lexicon approach can be used to arrange data into neutral, negative, and positive labels that depict prevailing sentiments. It can also be combined with Multi-Criteria Decision Making to detect emotions. Conversely, the Louvain method is a clustering algorithm that can be employed for topic modeling, the process of extracting a group of words from a set of documents that best represent the contained information. The Naïve Bayes approach can also predict personality and emotions from typical social media posts. The best results are attained when the method is combined with statistical tests.