Logistics Data Analytics and Delay Prediction

Abstract

With transportation delays, the logistics sector has many challenges in ensuring the timely delivery of goods. In order to predict transportation delays, this study investigates the use of predictive modelling and logistics analytics. A dataset consisting of 10,000 randomly selected records from a larger Kaggle dataset which use for study intends to create efficient models for forecasting delays based on previous logistical data by applying machine learning methods such as Naïve Bayes, Decision Trees, and Logistic Regression. The CRISP-DM framework is used in the study to prepare and analyze data in a methodical manner, create prediction models, and assess how well they operate. The Decision Tree model achieved the highest accuracy (74.75%) otherwise Logistic Regression had the highest recall (83.31%) which making it the best at detecting delays. The Naïve Bayes classifier achieved 71.58% accuracy, 38.89% precision, 13.39% recall, and an F1-score of 0.63. The Decision Tree model had 74.75% accuracy, 40.00% precision, 0.33% recall, and an F1-score of 0.68, whereas Logistic Regression had 54.37% accuracy, 33.64% precision, 83.31% recall, and an F1-score of 0.68. The findings suggest that Logistic Regression is the best at identifying delays with a higher recall, even though the Decision Tree model has the maximum accuracy. By implementing predictive models and advanced analytics, logistics providers can reduce delays, optimize resources, and improve customer satisfaction. In order to increase operational effectiveness and proactively control delays, this paper offers suggestions for combining real-time tracking systems with predictive analytics.