Predictive Methods for Event Revenues and Expenses to Support Convention Center Portfolio Management

Creating an efficient event portfolio is a multifaceted challenge for convention centers that involves selecting the right mix of events to maximize their economic impact and profitability. Accurate prediction of event revenues and expenses based on event characteristics is crucial in the evaluation and selection phases of a project portfolio management process. This study proposes a comparison of three prediction methods specifically applied to event revenues and expenses. Three of the evaluated methods employed Artificial Intelligence (AI) techniques (decision trees, random forest, and multilayer perceptron) alongside polynomial regression for its predictive capabilities. These approaches are evaluated with real data using the Mean Absolute Percentage Error (MAPE) to ensure the most reliable predictions are selected. By employing multiple techniques, our approach provides a sophisticated instrument for predicting financial results, enabling managers to make informed decisions that are in line with strategic goals and maximize the economic advantages of hosting events. In our case, with the appropriate features and number of trees, the Random Forest method stands out as the most effective among the tested methods. Incorporating AI-driven predictions into portfolio management represents a significant advancement in enhancing event outcomes and the overall performance of the center.