Page 552 - AI for Good Innovate for Impact

AI for Good Innovate for Impact

(continued)

Item Details
Model Training Both models are implemented using Python and Rust. SQL queries are
and Fine-Tuning used to fetch and match relevant data from the operational systems. The
models are retrained and updated periodically to keep the recommen-
dations relevant and accurate.
Model 1: Haulier Recommendation Model
This model uses XGBoost, a gradient boosting algorithm, to recommend
the top 3 hauliers based on historical bookings and container specifica-
tions. The model pipeline includes:
• Data fetching and aggregation across multiple systems.
• Label encoding of categorical values (e.g., haulier names, container
types).
• Handling of data imbalance using SMOTE (Synthetic Minority
Over-sampling Technique).
• Model training using XGBoost, with iterative tuning until the F1 score
exceeds 0.6.
If the accuracy is acceptable, feature importance is extracted to update
attribute weights in the database. The model is retrained every 15 days
to maintain accuracy.
Model 2: Time Slot Recommendation Model
This model uses a ratio-based scoring algorithm to recommend the best
time slot for haulier execution. The process includes:
• Standardizing available slots by computing their proportion out of the
total daily slots.
• Standardizing haulier execution data similarly.
• Computing probability scores by multiplying standardized slot values
with execution ratios.
Recommending the slot with the highest weighted score. This approach
does not use a machine learning model but applies a mathematical
policy-driven method using ratio-proportion and weighted averages to
ensure fairness and slot efficiency.

Testbeds or Pilot Pilot deployments & Live: Uatcms.dubaitrad.ae (test environment)
Deployments live environment[1]

2 Use Case Description

2�1 Description

Cargowaves, an AI-driven inland transport platform, is transforming logistics by eliminating
inefficiencies and enhancing cost-effectiveness for shippers, transporters, and cargo depots.
Acting as a centralized digital marketplace, it streamlines truck booking, scheduling, and
execution using AI-powered recommendations. This optimizes trucking operations, minimizes
empty runs, and enhances supply chain visibility.

Key Features & Benefits

• AI-Powered Hauler Matching: Recommends transport providers based on experience
and price, improving cargo delivery efficiency.

516