Classification of Train Ticket Prices in Spain Using Machine Learning Models

Current public transport timetable systems often focus solely on travel time, neglecting other crucial factors such as ticket cost. This paper explores the classification of train ticket prices in Spain, particularly with Renfe’s complex fare system, which is influenced by numerous variables, including departure and destination, train type, and travel date. The study investigates the impact of these factors on ticket prices using various machine learning classification models, including Logistic Regression, Bayes Classification, One-vs-Rest, Random Forest, and Decision Tree. An extensive dataset of Spanish railway fares from Kaggle, containing 38.8 million rows (with 8 million rows selected after cleaning), was utilized for model training and testing. The paper emphasizes the importance of data preprocessing, feature selection, and parameter tuning in achieving high accuracy in fare prediction. The results indicate that Random Forest and Logistic Regression models offer significant predictive accuracy, with optimization techniques applied to enhance performance. The study also compares its findings with existing research, highlighting the efficacy of Decision Trees and the differences in performance between models based on varying data characteristics and fare systems.