A Hybrid Approach for Outlier Detection in Pharmaceutical Cold Chain Logistics: A Case Study

Abstract

Pharmaceutical products are highly sensitive to environmental factors including temperature fluctuations, humidity, light and vibrations. Therefore, pharmaceutical cold chain disruptions pose significant human health, financial and regulatory risks, with logistics activities being the most critical point of vulnerability. This study proposes a hybrid approach called Route Detection-based Support Vector Regression (RD-SVR) algorithm for detecting temperature outliers during transportation of pharmaceutical products. The algorithm first conducts a preprocessing stage on the big data set where it detects the routes and prunes the unnecessary samples at the same time. Then, the cleaned data is used to train the SVR model for outlier detection. The automized route-based structure offers a real-time solution for cases involving a dynamic set of varying routes and vehicles. The proposed RD-SVR model is applied on one-year data logs of two vehicles gathered from an international logistics company, and tested against Random Forest and ANN. Test results and sensitivity analysis highlight the robustness and effectiveness of this innovative classification-based outlier detection model for identifying cold chain breakages and safeguarding product quality.