Continue refining constraints and limitations, inputs and outputs of model
feature engineering and preprocessing (data analysis, outlier detections, etc.)
Start on training simple models
Feedback on week 2 report
Load of vessel: add season feature and draft (calculated by diff between torque and thrust)
Captain: clustering
Extreme cases like avoid collision with crafts: by clustering / outlier detection
Limitation of the inputs: speed, thrust, scheduling
Scheduling: compute time left (difference of average expected trip duration (120 min) and current trip time) as another feature. This in combination with the GPS location, can tell the model if there is any delay in the schedule.
Think about having thrust, torque and other dependent variables as the output of the model to increase robustness
Consider dynamic models: RNN, LSTM, GLU, transformers, etc.
Continue refining constraints and limitations, inputs and outputs of model feature engineering and preprocessing (data analysis, outlier detections, etc.) Start on training simple models Feedback on week 2 report Load of vessel: add season feature and draft (calculated by diff between torque and thrust) Captain: clustering Extreme cases like avoid collision with crafts: by clustering / outlier detection Limitation of the inputs: speed, thrust, scheduling Scheduling: compute time left (difference of average expected trip duration (120 min) and current trip time) as another feature. This in combination with the GPS location, can tell the model if there is any delay in the schedule. Think about having thrust, torque and other dependent variables as the output of the model to increase robustness Consider dynamic models: RNN, LSTM, GLU, transformers, etc.