Inspiration
The logistics sector is a price competitive sector with small margins. Small derivations or delays in delivery of packages can lead to losses.
What it does
Our model predict whether a delivery from point A to B with different types of trucks will be on time or delayed. By more accurately determining the probability of on time or delayed so that dispatchers can act upon it.
How we built it
We started from an open-source dataset and notebook from Kaggle. We then improved our dataset and quickly checked this with other Classification libraries to decide what dataset had the best results. We then used our optimal dataset to train a decision tree model with adaptive ASHA using determined AI's platform.
Challenges we ran into
Running an experiment from within the determined AI cloud was a challenge. We kept getting token expiration errors. At the end we decided to use determined client locally on our pc and run the experiments from there.
Accomplishments that we're proud of
We did a great job in data cleaning and feature selection compared to the earlier attempts of people using this dataset.
What we learned
We learned a lot of how to optimally use the determind AI platform and how to experiments in different modes
What's next for Mean Lean Learning Machine
We want to further improve current model and implement the complete pipeline by also using pachyderm for our data pipeline and Seldon for inference.
Built With
- kaggle
- python
- tonic
- tpot
Log in or sign up for Devpost to join the conversation.