CAE Challenge - HackerBoises

Inspiration

The aviation industry is on the rise and pilot’s jobs are expected to grow at a rate of about 12 percent per year. The increase in demand of pilots means that there is an added responsibility on the instructors to make sure that flight rules are followed to ensure the safety of passengers. This project prevents mistakes caused by human errors affecting the outcome of the training session and therefore, makes the process more transparent.

What it does

CAE trains more than 120,000 people in civil aviation, defense and security every year. Due to the safety risks associated with the training of the pilots the company checks whether a pilot owns, or not, a specific competency after every 6 months. During a training session, the instructor tags the competency if the pilot does not comply with all requirements after each maneuver. Since manual labor is involved there are chances that some maneuvers are mislabelled. This project aims to use machine learning algorithms to predict the relationship between the features and the labels. It then uses the predicted values using a given dataset to return if the competency has to be flagged.

How we built it

The given problem was solved using two methods. The first solution of the program grouped the data by id while the second solution used each independent row of the data to predict competency. It was ensured throughout the course of the project that the workload was divided equally between the members of the team to maximize efficiency. Timeline: Team strategy and division of workload - 2 hours Research and Mathematical Analysis - 6 hours Data Visualization and Preprocessing - 2 hours Choosing machine learning technique - 10 hours Time series analysis - 2 hours Reporting results - 2 hours Total time of project - 24 hours

Challenges we ran into

Some of the major challenges include time constraint and introduction to advanced machine learning techniques.

Accomplishments that we're proud of

We as a team learned a lot and are proud of what we have achieved over the course of the weekend regardless of the result produced.

What we learned

1. Data Visualization and Preprocessing ○ Visualization ■ Orange Biolab ○ Data cleaning ○ Data reduction ○ Data transformation
2. Choosing the machine learning technique ○ K-mean clustering algorithm ○ Mean shift algorithm ○ Density-based spatial clustering of applications with noise(DBSCAN) algorithm ○ K-nearest neighbors algorithm ○ Cross-validation
3. Time series analysis, visualization, and forecasting ○ Neural Networks (Long short-term memory networks) : multivariate time series for unsupervised learning with LSTM for predicting the flagged every time step
4. Reporting results ○ Competency prediction and graph plotting ## What's next for CAE Challenge - HackerBoises It was the first time for us as a team and we believe that the team performed above our expectations. We plan to work together in other hackathons as a team and bring more ideas to the table.

Built With

• orange
• python