AI against Carbon Impact
Classify bank transactions and estimate their carbon impact to help people live more sustainably
Premium participation: Get 1-on-1 Mentoring & Guidance
Try a free mentoring sessionUsing NLP to classify bank transactions and estimate CO2 emissions
We’d all like to live the most sustainable life possible. But being mindful of our impact on the environment is only effective with a detailed understanding of the consequences of our actions. No measurement, no control. Luckily, we are living in a time where technology is rapidly evolving to help us gain that understanding—and you can play a crucial role developing it!
In AI against Carbon Impact we are partnering with Yayzy to build a machine learning model that estimates the carbon impact of people’s purchases through their bank transactions. Having such a tool means that, besides helping you make better financial decisions, your bank statements can also help you live more sustainably.
Yayzy has a carefully researched classification system to divide bank transactions up into about 90 categories. They also recognize transactions from over 30,000 retailers worldwide. Knowing the type of transaction, the amount of money transferred is multiplied with a weighing factor to determine carbon impact. The only trouble is: there’s much more than 30,000 retailers worldwide, so in some cases it’s necessary to estimate which category a transaction belongs to and you guessed it: that’s a task ideally suited to machine learning.
Info session
Our challenge partner
GOAL: Build an accurate and robust natural language model that can classify and predict the estimated carbon footprint of a single transaction.
By providing insight into consumer patterns we can help people make more sustainable choices. We will be using NLP to classify the type of transaction, eg. “Food & Dining”, “Shopping”. We use this information to predict the CO2 emissions.
In this challenge, we will
- Build language models for classification
- Use clustering methods to find new categories
- Analyse and enrich large datasets (>500.000 instances)
- Integrate models into the Yayzy application
Technologies we will use
- NLP
- Word2Vec
- Clustering/Classification
- BERT
- Huggingface 🤗
Who are we looking for?
We are looking for data science & AI engineers, previous work with language models would be of great benefit.
You will collaborate with a diverse team of up to 50 international collaborators in subteams. You can join as a contributor (8-12 hours per week commitment for 8 weeks) or coach (2-4 hours per week, only for experienced ML professionals)
We’ll organize a masterclass on using language models with Huggingface during the challenge to bring you up to speed.
Did you know
🇶🇦 Qatar has the highest CO2 emissions per capita at 37.29 tons a year (world average is 4.79 tons)
🥵 In order to have a reasonable chance of limiting global heating to 1.5°C we need to cut our emissions to 2.3 tons of CO2 per person per year in 2030.
⛽️ 70% of all historical greenhouse gas emissions can be tranced back to 100 fossil fuels companies. If consumers can reduce their fossil fuel consumption we can decrease emissions drastically.
Timeline
Application Deadline: 10 Mar 2023
Challenge Kick-off: 14 Mar 2023
Midterm Presentations: 4 April 2023
Final Presentations: 9 May 2023