A comparative study of statistical and intelligent classification models for predicting airlines passenger management satisfaction

This paper compares Statistical and Intelligent classification models in predicting passenger satisfaction with airlines. It seeks to identify the most accurate and reliable model amongst GLM, Robust Regression, MARS, KNN and Neural Networks. Methodology. These models were analyzed using a set of performance metrics: MSE, MAE, R-squared, RMSE, COV, CC, EC, and overall accuracy. Data set comes from KaGGle website. The results obtained from these models were analyzed with the help of descriptive statistics, statistical tests of ANOVA, and the Wilcoxon signed rank test to establish significant differences in the performance of these models. Key results are that MARS is significantly better than others, with the least MSE (0.0717), MAE (0.0717), RMSE (0.2679) and the highest R² (0.7078) and CC (0.8537) among them. The ANOVA of the test results showed significant differences between the models, as the P-value calculated was less than 0.0001. This confirmed that the performance variations obtained are of statistical significance. Similar results are confirmed with the Wilcoxon Signed Rank Test, with P-values of 0.002 for all models. This shows significant differences between the theoretical and actual medians. Conclusions. MARS is the most effective model for predicting passenger satisfaction with Airlines as it is more accurate and reliable. The critical potential research domain is to develop Hybrid models by blending the best of multiple approaches for better accuracy and robustness in passenger satisfaction with airline prediction.