Ph.D. candidate Tiyasha (India) in Civil Engineering successfully defended a doctoral dissertation at the University- level

On August 12, 2023, Ton Duc Thang University (TDTU) held an examining committee for the doctoral dissertation at the University-level for Ph.D. candidate Tiyasha, in Civil Engineering (Program code: 9580201), with the topic “Evaluating surface water quality using artificial intelligence models for water management in concrete engineering” under the supervision of Dr. Zaher Mundher Yaseen (TDTU) and  Dr. Tran Minh Tung (TDTU). 

Tiyasha was recruited at TDTU in 2018. Before defending the doctoral dissertation, she completed 12 credits of specialized courses, 06 credits of doctoral research topics, 04 credits of literature review, completed the faculty-level doctoral dissertation examining committee, and 02 scientific papers published in reputable journals.

The academic contribution of this dissertation is stated as follows:

- The dissertation designed a classification prediction model for WQI prediction using Rapid miner software while targeting two types of scenarios: Klang River Basin (large catchment) and Klang River data (small catchment) to have more information about seasonal variation and a point source of pollution.

- The deep learning model performed very well in both scenarios and was able to handle both smaller and larger catchment datasets (S-I and S-II: Accuracy 98.6% and 87.7%, precision 80% and 64%, classification error 13% and 10%, f measure 90% and 90%). 

- The remotensing data was procured using Google Earth Engine coder, i.e., image collections ERA5 ("ECMWF/ERA5/DAILY") and TerraClimate ("IDAHO_EPSCOR/TERRACLIMATE") and QGIS software.

- The study also applied Rstudio software for regression modeling. Feature selection algorithms, i.e., Boruta, genetic algorithm (GA), multivariate adaptive regression splines (MARS), and extreme gradient boosting (XGBoost) were implemented. The selected features were then applied using random forest (RF), conditional random forests (cForest), RANdom forest GEneRator (Ranger) and XGBoost models to predict DO.

- It was found WQ data and hydrological data obtained from satellites, which are characterized by a high degree of non-linearity, randomness, outliers, and low correlation values, can also be used for DO prediction using AI (XGBoost and RF) models. The MARS and XGBoost algorithms as feature selection methods are an efficient method to reduce the unnecessary application of input variables and decrease the computational cost.

The doctoral dissertation was approved with 07 votes from 07 committee members.

There are some photos of the doctoral dissertation examining committee on August 12, 2023:

Dr. Thi Ngoc Bao Dung - Assistant Head of the Department of Graduate Studies (TDTU) announced the Decision to establish the committee.

Assoc.Prof. Vo Le Phu (Ho Chi Minh City University of Technology), Chairman of the Committee.

Ph.D. candidate defends her doctoral dissertation

The Committee took souvenir photos with Ph.D. candidate