Ingeniería de características basada en análisis exploratorio de datos para el modelado y la previsión de gastos en turismo

This pilot study introduces a feature engineering (FE) methodology based on exploratory data analysis (EDA) to augment and enrich an initially limited dataset. The primary objective is to establish a systematic framework for accurately predicting tourism-related expenditures at an operational resort in Esquel, Argentina. Data collection practices in such tourism facilities are often suboptimal: reservations are typically recorded manually or stored in simple spreadsheets, then examined informally to identify apparent trends. Under these conditions, EDA is a valuable tool for investigating relationships among the original variables and for constructing derived variables that more effectively capture their association with the studys target variable. A multivariate linear regression model was specified to predict Tourist Spending. The newly engineered predictors derived through FE informed by EDA, namely Interaction between guests and days, Spending per person, and Staydays, enabled a transparent and interpretable decomposition of the variance in the dependent variable. Model performance, assessed using RMSE, MAE, MSE, and the F-test, suggests that the model explains approximately 66.6% of the variability in Tourist Spending at the resort. These results are promising and provide further evidence of the critical role of FE and EDA in developing enterprise-level predictive models for tourism-related expenditures.