I&T Solution |
Aboveground Gas Pipes Health Condition Analysis
(REF: S-0954) |
Trial Project |
|
Solution Feature |
- Build a Big Data Analytics Platform, which will be used to store both the structured and unstructured data of Intelligence Aboveground Gas Pipes Health Condition Analysis System.
- Develop an ETL (Extract, Transform and Load) module for loading of gas pipes information into the data lake of the Big Data Analytics Platform.
- Build a AI module with machine learning algorithms (e.g. ) to generate a model to identify the factors & correlations as well as their importance affecting the health condition of gas pipes.
- Apply the model to the gas pipes information in the data lake to predict the health conditions of the individual gas pipes.
- Generate dashboards and alert & analysis reports for the monitoring of the gas pipes health conditions.
|
Trial Application and Expected Outcome |
- K-fold cross-validation testing will be used for the machine learning model validation, which 80% of the past incidents and maintenance records data for training the model.
- The historical data will be divided equally into, say, 4 parts. All parts but one part of the data will be used for training the model. The withheld piece will be used to test the model. This if the first fold of cross validation.
- Another fold of validation will then be performed, where a distinct combination of the data will be taken to train the model, and a withheld piece will be used to test the model.
- Upon completing all “k” of these k-fold cross validation, an average F1 score can be computed for each of the tests.
- This k-fold cross-validation testing will be applied to various machine learning model to find the best model to achieve the desired % accuracy for the system.
|
Info on I&T Solution Provider |
Solution Provider | : | NewTrek Systems Limited | Address | : | 1801 Westley Square, 48 Hoi Yuen Road, Kwun Tong, Kowloon, Hong Kong | Contact Person | : | SAMUEL POON |
Position | : | Chief Technology Officer | Tel | : | 2154 7189 | Email | : |
samuel.poon@gmail.com | Webpage | : | www.newtrek.com.hk |
|