With the widespread use of data today, the sky’s the limit with data analytics. That’s how our data science team aggregated flight data and analysed over 50 factors to generate insights and boost flight safety for the RSAF.
By analysing the impact of factors such as weather conditions, pilot experience and flight path profile, the team developed a predictive modelling application that computes a risk score for each training flight undertaken by RSAF pilots.
“The idea is to help RSAF trainers evaluate training plans ahead of time, and allow necessary adjustments to be made for increased training safety for our pilots,” said Senior Data Scientist (Digital Hub) Chong Shi Kai.
The team leveraged a vast amount of data from various sources such as the aircraft’s Safety Information System, which records flight incidents and their probable causes, as well as the Flight Administration System, which records more comprehensive flight data including important flight and pilot attributes.
However, the large variety of data sources meant a huge challenge in integrating them for analysis. To overcome this, our data scientists explored various methods and eventually developed a series of natural language processing techniques to extract key information from unstructured text. Going one step further, they even automated a data engineering pipeline to comb through the datasets effectively and efficiently.
To better understand the root causes of safety risk during training, the team employed a combination of statistical tests and econometric methods, and developed an ensemble predictive model that leveraged the strength of multiple types of machine learning models. This enabled a risk score for each training flight to be generated quickly, allowing high-risk flights to be flagged out and mitigating measures to be considered.
This pre-flight risk assessment tool could not just boost flight safety, but also enhance the RSAF’s capability to better leverage data for more informed decision-making.