We were initially inspired by the AGCO talk at the start of hackathon, where we met Lance, a farmer, who had numerous issues with labor, machinery, and sustainability. Lance provided a human face for the otherwise vague and foreign problems presented by AGCO. In addition, one of our team members was personally motivated by the plight of Indian farmers that do not have many of the same resources that we can provide in the US today- the idea being that AGCO could possibly extend our solutions with their global reach.
What it does
The website allows farm owners to easily view the power consumption of their devices and find anomalies in the amounts of current that each draws over time. The intuitive dashboard allows for easy access and use in analyzing equipment.
How we built it
We used AWS SageMaker in order to run our Jupyter Notebook to analyze and work with the data. Using a random cut forest, we were able to find the anomalies in our dataset, and based upon that we can generate alerts. Using EC2, we were able to host our website
Challenges we ran into
We ran into numerous challenges with data acquisition and sorting, as some of the poultry fan data was simply missing at the start. It was also challenging to sort and merge information from multiple different sources. Our unfamiliarity with farm design and machinery also added to the difficulty of deriving a conclusion from our various sources of information.
Accomplishments that we're proud of
We are proud of having succeeded in building a model out of the limited and less organized data in the poultry set [as opposed to the machine set] and of our use of AWS for hosting and running our analyses.
What we learned
We have learned quite a lot regarding working with missing or imperfect data sets; mainly making something out of less than ideal conditions. We have also learned a bit about front end and presentation when it comes to making a picture or claim out of seemingly random data points.
What's next for AnomalyAlerts
In the future, we would like to scale the site such that we could add multiple users. In addition, we would also like to provide these additional users with increased functionality including real time data on the dashboard and the ability to upload their own data sets.