Novel Algorithm to Predict Electoral Trends, Case in Mexico

The purpose of this research work is to investigate whether the natural language expressed on social networks, in addition to traditional surveys, can be a factor in anticipating the winner of a presidential election in Mexico. To achieve this, our research team has developed an algorithm that aims to differentiate the explicit and implicit manifestations of users’ electoral preferences based on the grammatical structure of Twitter comments. After conducting a systematic search for argument structures, we identified a set of verbal words that allowed us to infer voting intentions and program the algorithm accordingly. Subsequently, we applied the tool to a sample of 10,000 tweets during the 2018 presidential elections. The results obtained by the algorithm proposed in this research show similarity in trend with the results obtained by the National Electoral Institute of Mexico.