Page 237 - AI for Good Innovate for Impact

AI for Good Innovate for Impact

TV broadcasting system, IoT, etc.); upon the provision of additional weather and particular
transmission requirements.

2�3 Future Work Change 4.2-Climate

• Data Collection: The project utilises live weather events, therefore continuous
collaboration with relevant meteorological agencies is necessary for continued capturing
of high-quality weather events. The collaboration ensures that the model is continuously
train using up-to-date and consistence data. Partnership with regulatory bodies can be
established to strengthen the model’s generality.
• Model’s Development: In the project, an NNR-based ML model analyses the impact
of weather events on PLC-based system regarding data rate in Malaysia and Nigeria.
The model’s performance can further be evaluated with other traditional ML models
(such as Polynomial Regression, Support Vector Regression, Random Forest Regression,
Gradient Boosting Regression, k-NN Regression, Gaussian Process Regression, etc.).
The ML training and hyperparameters can be maximised and optimised respectively to
improve accuracy and efficiency.
• Use Case Difference: The NNR-based ML model is developed to analyse impact of
extreme weather events on PLC system. The PLC is a wireline-based technology that
support power and data transmission using similar power cabling infrastructure. By
providing additional weather and specific transmission requirements, the model’s
capabilities can be extended to analyse other regions and communication technologies.

3 Use Case Requirements

• REQ-01: The Use Case domain is to analyse impact of climate change-inducing extreme
weather on data transmission of PLC-supported power lines in Malaysia and Nigeria.
Other communication technologies and regions are not captured.
• REQ-02: It is critical for power industries to deploy or maintain PLC systems based on the
ITU-T G.9960 recommended architecture and performance. Most regional PLC-based
power systems use different parameters. It is also important maintain partnership with
stakeholders (power industries, weather monitoring agencies, and regulatory bodies) for
effective prediction of current and future data performances.
• REQ-03: It is critical to continue sourcing for up-to-date weather data, as climate change
is continuously disrupting weather patterns. Datasets and scripts utilised in the project
are publicly available at https:// github .com/ adams714040/ ITU -Project .git .[1]
• REQ-04: It is critical to explore other ML-based regression models that are suitable in
analysing weather patterns and impact such as polynomial regression, support vector
regression, random forest, Gradient Boosting, k-NN, gaussian process, etc.) for accuracy,
efficiency, and reliability.
• REQ-05: It is critical to align prediction carried out in this project with other relevant
standards like IEEE 1901 (2010).

4 Sequence Diagram

The figure below shows development process of the NNR-based ML model. Dataset is collected
and validated. It is then preprocess, and process onto ML for training. The ML processing is
computed using 2-layer structure; where the first layer takes the input (dataset), and the second
layerenables the NNR to learns and trains the data for prediction (output).

201