How machine learning is helping to predict critical events among thousands of shipment routes

e-Core • August 12, 2024

The challenge

The client, a leading company in risk management for the transportation and logistics sector, faced the challenge of efficiently managing and prioritizing a vast number of workflows. Their system handled approximately 25 million events daily, with about 2 million workflows per month and 350,000 shipments tracked monthly.

Operators managed 100 to 200 shipments simultaneously, but only 1 to 2% of these shipments resulted in the incidents they aimed to prevent. The challenge was to use machine learning to predict which shipments were at the highest risk and prioritize them accordingly, thereby improving operational efficiency and reducing the workload on human operators.

The solution: machine learning for risk prediction

Machine learning techniques were applied to predict risks and prioritize workflows based on the likelihood of critical events. The solution involved identifying the types of incidents that should be prioritized and selecting appropriate algorithms considering factors like model complexity, problem-solving capacity, and predictive effectiveness.

The problem was addressed using an ensemble method — a machine learning technique that builds a sequence of decision tree models and combines their outputs to create a robust and accurate model. Given the naturally imbalanced nature of the problem, machine learning techniques were used to correct data distribution and minimize the impact on predictions.

The strategic focus was on minimizing false negatives (shipments with critical risks that the model incorrectly predicts as safe), as untreated incidents could lead to financial losses. The approach allowed for an increase in false positives (shipments without critical risks that the model predicts as risky), ensuring that more potential incidents were captured in the workflow management system.

The Results

The implementation of the machine learning model significantly improved the client’s ability to predict and prioritize high-risk shipments. The model demonstrated a high capacity to distinguish between shipments with and without critical risks, achieving an AUC of 94%, indicating reliability and effectiveness in risk prediction.

The false negative rate was 11.5%, a manageable figure given the challenge of identifying incidents. The model correctly classified 88.5% of shipments with critical risks, proving its efficiency in the task.

These results reflect the model’s effectiveness in making precise risk predictions, enhancing the reliability of its applications in logistics risk management.

3. Results

  • High Predictive Accuracy: The model achieved a 94% AUC, indicating strong discriminatory power in predicting risks.
  • Reduction of False Negatives: The false negative rate was reduced to 11.5%, allowing better focus on high-risk shipments.
  • Operational Efficiency: Prioritizing shipments with predicted risks reduced the operational burden on human operators, enabling them to concentrate on the most critical cases.

LET'S CONNECT

Ready to unlock your team's potential?

e-Core

We combine global expertise with emerging technologies to help companies like yours create innovative digital products, modernize technology platforms, and improve efficiency in digital operations.

You may also be interested in:

Transforming healthcare with data and AI portfolio alignment

How a health insurer brought structure to scattered AI efforts and turned them into results

By e-Core • August 14, 2025
A top insurer aligned data and AI to cut waste, speed up delivery, and turn preauthorization into a fast, accurate, real-time process.

Major Brazilian energy provider boosts scalability and efficiency with AWS

By e-Core • July 31, 2025
Energisa modernized 70 systems by migrating to AWS with e-Core’s support, gaining scalability, availability, and performance in the energy sector.

Best service desk tool: A practical comparison for smarter IT operations

By e-Core • July 22, 2025
Introduction In today’s digital economy, IT Service Management (ITSM) platforms are more than just support tools, they are strategic enablers that impact operational efficiency, user satisfaction, and business agility. With an increasing focus on automation, AI, and user-centric service delivery, selecting the right service desk solution has become a critical decision for CIOs, IT leaders, and digital transformation teams. Why choosing the right service desk tool matters The service desk is the heart of IT operations. Beyond resolving tickets, it supports governance, ensures compliance, and drives consistent service experiences. A well-chosen ITSM platform enables faster resolution, proactive problem management, better user engagement, and alignment with digital transformation goals. On the other hand, a misaligned tool can lead to inefficiencies, user dissatisfaction, and mounting operational costs.  When to use each tool Choosing the ideal tool depends on your organization’s size, IT maturity, integration needs, and strategic objectives: Jira Service Management (JSM): Best suited for organizations already using the Atlassian ecosystem or looking for agility and fast ROI. Strong in DevOps integration and automation via Rovo. ServiceNow : Preferred for large enterprises needing robust enterprise-grade features, deep AI capabilities, and high levels of customization. Ideal for those with higher budgets and complex service architectures. Freshservice: A favorite among mid-sized companies that need a full-featured ITSM without the complexity. Known for ease of use and rapid implementation. Zendesk: Strong contender for businesses that need multichannel support, especially customer-facing teams. Good choice for aligning IT support with customer experience. ManageEngine ServiceDesk Plus: Ideal for organizations needing strong ITIL compliance, asset management, and on-premise or hybrid deployment options.