ai4gooders

Discovering hidden insights and creating value from the 211 calls and related datasets.

Comment

AI4Gooders

The project is based on using data indicators to create an accurate and weighed scoring structure that will assist in providing a strategic placement system for predictions of risk assessment areas for certain categories. Our system will compile key-environmental datasets such as Covid-19 number of cases in the area, the income of the population in the area, age and gender of 211 callers, etc. The information is used to create a system that prioritizes sub-categories of people in sub-areas that have a higher need for emergency help, in order to improve the local life. The scoring system acts as a visualized map for users to compare each part. Each individual dataset or factor will have an effect on the overall score of a location and will represent which site has a greater demand for life quality improvement.

Inspiration

We are simply looking for learning opportunities and fun, while dedicating some brain cycles to helping out Centraide. The Covid-19 is currently a global issue that needs any help it can get. Obviously, there are issues for people who need help that increase the dangers of the pandemic for them. These issues would be related to gender, age, newcomers, low-income, etc. Everything adds up, and we believe that even one person can make an impactful difference, just like how our project may be able to make a difference. Using that logic, we thought about how we could potentially help our local community first. We developed a system that uses key centraide datasets to help identify geographical locations that include people under higher risks due to the pandemic, or weather challenges for new immigrants, to boost the quality of life. Through centraide’s data analysis, it can essentially generate a report that will be converted into a scoring system that can help users visualize where helping people would have the greatest positive impact on the local area.

Significance

This is important because strategically placing new helpers based upon needs of people rather than geographical purpose has the potential to help local life, which can grow to a larger scale and have a beneficial impact on the global health.

What it does

The system uses python to retrieve Covid-19 data and 211 Calls by Week data, longitude, and latitude of organizations specified by the area. We have looked at this hackaton from 3 different angles:

  1. We implemented some visualizations about the evolution of the number of declared cases of Covid by borrow through time
  2. We tried to characterize the profiles of callers to the 211 with some unsupervised clustering algorithms. Our clusters show the frequency of callers in each gender, and their age range. This informs us to predict the future calls to identify under-need people are from which category.
  3. We trained a model to recommend which agencies each caller should be referred to based on their profile and their needs.

How we built it

Our first target was to find relevant datasets provided by the organizers, specifically, important centraid data. After finding the necessary datasets, we set our minds to being able to understand what information we want to use from them and how we would go about using it. We used a collection of Python, Mapbox API, Python Pandas, and Numpy, Plotly, and Sklearn libraries, Google Colaboratory notebooks in python. At the top of the page is a visualization of our work.

Challenges we ran into

Our main problem that we faced was the constraint of time. We have big plans, however it is difficult to integrate everything into the final product because that takes a significantly longer amount of time. In terms of victories, we were able to fully incorporate key datasets into the our system. All that data would come together to show the points system. Also, we could not find polygons for the municipalities in greater Montreal that are not on Montreal island. We wanted to combine the two visualizations on Covid19 cases and 211 calls together that proved to be tricky. We also wanted to explore graph databases to visualize the relations between callers and their needs.

Accomplishments that we are proud of

We discovered a flaw in the 211 datasets with the taxonomy codes no one else had noticed. (It will be corrected).

What we learned

During this hackathon, we learned how to split the problem into smaller sub-problems to have a better understanding of each part of it. This helps us to tackle the main problem from different perspectives as a group which enables us coming up with a better and more comprehensive solution. We got better at manipulating datasets with pandas.

What's next for ai4gooders

The next steps toward our main goal--helping society and Centraide for a better life-- we will continue using more related data from other agencies to be able to cover more areas with helpers. By identifying problems that each category of people who need help are struggling with, we would be able to provide specific need that they need.

Built With

Share this project:

Updates