This past summer, we were both exposed to the wealth of publicly available data at an internship at NASA. After speaking with scientist and engineerings working with LandSat and ICESat data, scientists doing field work and more, we saw a potential for outside uses. The World Health Organization states that Malaria accounts for hundreds of thousands of deaths per year. There are many efforts in place to help this cause, many of which are our inspirations, such as the Malaria Atlas Program (MAPS), ongoing research at NASA that uses LandSat data, and even more individual field research programs.
We want to use machine learning as a tool to optimize efforts to decrease Anopheles mosquito density with actions determined by our trained agent; the goal is to create a Q-learning Reinforcement Learning Model that uses keras and tensor-flow to achieve this.
This is a project that is important for both of us, as our interest lies at the intersections of healthcare, accessibility, and technological innovation.
What it does
At the moment, CAMP's capabilities are that it reads climate data, aggregated with Anopheles data. Uses ARIMA time-series algorithm approach to forecast future density of Anopheles in a specified area.
Using multiple compounded datasets to optimize the placing of mosquito traps to decrease the population affected by mosquito bites that lead to malaria infection.
An interactive tool to aid researchers in field work by visualizing the density of mosquito breeding habitats and their proximity to at-risk populations.
training our agents to do two actions: identify locations to place traps and use drones to find breeding sites.
-Evaluate and quantify the effectiveness of the malaria vector control methods over time.
How we built it
Python Jupyter notebook for the ARIMA model. Colab for visualizing the Arima Model and defining requirement functions for the Reinforcement learning environment.
Challenges we ran into
The format, size, and content of the datasets used. Time constraints for training model.
- assigning high-resolution data that wasn't a 5kx5k
- each pixel had multiple survey points that were difficult to understand.
- time constraints for picking compatible datasets.
** this will be taken care of once we have more time to complete in the coming weeks
Accomplishments that we're proud of
We are proud of what we have accomplished, as this project is more of a marathon than a (hackathon) sprint.
Stefany- Although I wasn't able to get the Front-end aspects to fully compile, I learned about API's and their uses, attempted data aggregation, learned about machine learning and how easy it is to explore and use open-source tools.
Melchi- Aggregating data and creating the preliminary environment for the project.
What's next for CAMP
We have big plans for CAMP, which include a submission to Call for Code, a published paper, and work to find partners to endorse camp for real-world applications. We strongly believe that CAMP has the potential to be a useful tool for researchers and scientist working on malaria prevention.