We were fascinated about the toolkit provided by IBM, i.e. its cloud platform and the given climate dataset. Moreover, we found that the insurance setting provided by SwissRe enabled us to tackle a real world challenge - albeit in the outset it was not fully clear to us which specific task to focus on.

What it does

Our webapp is used by farmers anywhere in an arable setting to get an overview of climatic risks potentially impacting them within the next weeks or months. Using time series analysis, we generate insights from past weather data, specifically using a US-based dataset. Moreover, the goal is to provide each farm its own sensors to further refine their data accuracy. Ultimately, these insights help reinsurers assess upcoming natural disasters and mitigate risks before the fact by providing financial support for prevention (e.g. watering to prevent drought, harvesting before hail, etc.)

How we built it

We built the webapp using Dash by Plotly to visualize the results of our data analysis. Moreover, we integrated an IBM cloud assistant to provide farmers with a more personal interface, dealing with suggestions and serving as a steady connection to the insurance. We also integrated a map to show the current weather trends. Furthermore, our predictions are aggregated into a risk score, which can serve to identify crops and times prone to be adversely affected by natural disasters.

Challenges we ran into

The PAIRS dataset: We discovered inconsistencies in spatial and temporal dimensions. Furthermore, the API crashed on Saturday, forcing us to seek alternative sources of climate data. During this process, which consumed a lot of time and made us reflect on our overall strategy, we realized that a huge challenge in big data endeavors is being able to acquire a valuable and consistent dataset to train on, even more so than finding a suitable ML model.

Accomplishments that we're proud of

Team. We really had a lot of fun together :) Also, we all got to experience new areas of coding that we hadn't been exposed to before, such as front-end development and IBM tools.

What we learned

Using IBMs REST APIs. front-end development, finding datasets, UX wireframing, properties influencing crops-growth and insurance markets, how to composite a multi-dimensional metric into a one dimensional indicator.

What's next for FarmHero

Deploying IoT devices to monitor crop yield and environmental parameters. Distributing the software to farmers.

Share this project: