EcoMetrics

Inspiration

The inspiration for Portfolio Ecometrics stemmed from a collective commitment to fostering responsible and informed investments. Recognizing the growing importance of Environmental, Social, and Governance (ESG) factors in investment decisions, we sought to create a platform that empowers users to assess and enhance the sustainability of their portfolios.

What it Does

Portfolio Ecometrics is a comprehensive financial evaluation tool that allows users to upload their Yahoo Finance portfolios for in-depth analysis. The platform leverages a sophisticated algorithm to calculate ESG scores for each company within the portfolio, providing users with valuable insights into the ethical and sustainability practices of their investments.

How We Built It

The project was built using cutting-edge technologies and frameworks, including React and ExpressJS. The front-end, developed with React, Typescript, and MaterialUI provides a user-friendly interface for portfolio uploads and result visualization. The backend, powered by ExpressJS and NodeJS, processes and analyzes the portfolio data, generating ESG scores and ratings. NodeCache is utilized in our project to optimize performance by caching processed CSV file data. When a file is uploaded, it's parsed and the results are stored in the cache. Subsequent requests for the same file access the cached data, avoiding redundant parsing and hefty wait times. This results in faster response times and reduced server workload, making our application more efficient and responsive. ESG scores were calculated using the average sentiment score of news articles related to their impact regarding environment, social, and governance factors. These news articles were gathered from the GDELT Doc 2.0 API.

GDelt Doc API

We leveraged the GDELT Project's extensive database to acquire a large dataset of news articles relevant to ESG (Environmental, Social, and Governance) topics. GDELT, known for its vast collection of global news across various mediums and languages, updated every 15 minutes, served as an ideal source for our data needs. We utilized the GDELT Doc API Client to fetch the top matching articles over four years (2019-2023), focusing on ESG themes and company-specific information. To refine this dataset for predictive analysis of ESG ratings, we implemented a preprocessing strategy to filter out articles.

** Pre-processing strategy** To gather relevant new articles for certain companies, we queried articles relating to the company's name and a given theme. For example environment articles would fit into the "WB_1786_ENVIRONMENTAL_SUSTAINABILITY" theme of GDELT. To ensure that all articles relate to the company, we remove all queried articles that do not have the company's name in it.

Challenges We Ran Into

Throughout the development process, we encountered various challenges, such as [mention specific challenges, e.g., integrating third-party APIs, optimizing algorithm efficiency]. Overcoming these challenges required collaboration, problem-solving, and a dedication to delivering a robust and reliable solution.

Accomplishments That We're Proud Of

We take pride in successfully implementing an end to end platform that not only meets our initial vision but also exceeds user expectations. Achieving a seamless integration of complex algorithms and providing a visually appealing and intuitive user interface were major accomplishments for our team.

What We Learned

The development of Ecometrics provided us with invaluable learning experiences. From enhancing our technical skills to gaining a deeper understanding of ESG factors and sustainable investing, each team member grew both personally and professionally.

What's Next for Ecometrics

Looking ahead, we plan to continuously refine and expand Ecometrics. Future updates include:

Support for other investment portfolio formats such as Robinhood and Questrade.
Analyzing other datasets to calculate our ESG scores. For example we can analyze carbon emissions, yearly layoffs, and the number of lawsuits per year.
Improving the overall user experience.
Advanced Text Filtering and Processing: We plan to further refine our dataset with sophisticated text filtering techniques, enhancing the criteria for identifying relevant and high-quality articles. This step will involve more advanced NLP methods to assess article relevance to ESG topics comprehensively.
Integrating Beautiful Soup for Text Extraction: Although we haven't used Beautiful Soup in our initial phase, we see its potential for effective web scraping and text extraction in future developments. Beautiful Soup is known for its ability to parse HTML and XML documents efficiently. It can navigate and search the parse tree, which will be particularly useful for extracting meaningful content from various news website layouts, often complex and diverse in structure.
Data Significance Analysis: A deeper statistical analysis is on our roadmap to find the ideal data inclusion threshold. This will ensure a balance between the volume of data and its relevance for accurately predicting ESG ratings.
Enhanced Noise Reduction and Data Cleaning: We aim to improve our noise reduction and data cleaning processes, potentially employing machine learning algorithms to automatically filter out irrelevant sections, such as advertisements or unrelated content in news articles.
Content Summarization Techniques: Given BERT models' limitations, exploring advanced text summarization algorithms will be crucial. These algorithms will help in condensing articles while retaining essential ESG-related information, using NLP models trained specifically for summarizing ESG content.
Diverse Data Source Integration: Expanding our dataset to include a broader array of news sources is crucial. This will help mitigate any source-based bias and enhance the dataset's representativeness, especially from currently underrepresented regions and languages.
Model Experimentation and Optimization: Our continuous effort will include experimenting with various configurations of BERT models and other machine learning techniques, aiming to optimize the predictive analysis for ESG ratings.

Through these steps, we aim to build upon our initial efforts, enhancing the system's capability to predict ESG ratings from global news data with greater accuracy and comprehensiveness.