- Visualizations: Images we used to run streamlitapp.py
- Downloads/CSV_1: CSV's used in fundamental_charts.ipynb
- Downloads/CSV_2: CSV's used in machine_learning.ipynb and stock_sentiment.ipynb
- fundamental_charts.ipynb: Key performance indicators (ratios) used for our deployed investor dashboard.
- machine_learning.ipynb: Machine learning code.
- stock_sentiment.ipynb: sentiment analysis code.
- streamlit deployment: https://albakomito-g12-project-3-streamlitapp-3zw1yp.streamlitapp.com/
- Note: Code instructions are all contained within each notebook.
Our project's purpose is to build on our previous work and use Classification and Sentiment Analysis to demonstrate that our renewable energy portfolio is a good investment for environmentally conscious and socially responsible investors. This project utilizes a neural network via Classification to analyze the projection of the Renewable Energy portfolio prices and if they would provide a positive outcome. As a comparison, we use NLP sentiment analysis to further illustrate these renewable energy stocks are a great investment. Once we have gathered our data, we will create a user-friendly interactive dashboard where investors can view our findings. Our hope is that with the knowledge provided, investors will make a more informed decision to invest in these companies, and hopefully achieve market-beating returns.
Below, we briefly describe how we gathered the data, developed our code, and key findings after analysing the data. Finally, we use Streamlit (https://albakomito-g12-project-3-streamlitapp-3zw1yp.streamlitapp.com/) as a frontend development tool to display our results in a user friendly UI.
When preforming our classification analysis, we found that our results were not always accurate or conclusive. When gathering the data we included closing prices, earnings per share (eps) and revenue in our analysis. Our initial analysis (without closing prices, eps and revenue) yielded an exceptionally low accuracy score of 33%.
Upon adding eps, revenue and closing prices to our analysis, our classification model yielded an accuracy score of 59%, which (while higher than our model without the features) was not conclusive.
To address this, a future step in this analysis could be to explore different models or analyze different time frames to generate superior accuracy scores.
Our next step was to create a sentiment analysis using Google News API and observe results in hopes of positive outcomes. However, our results (illustrated below) were again inconclusive as some months were positive and others were more so on the negative side.
Another issue we encountered was insufficient data or news on some of the stocks we chose to create a more comprehensive sentiment analysis on our portfolio.
While we are very passionate about this idea and renewable energy investing, we conclude that our Classification Model and Sentiment Analysis do not provide enough information to investors to make an informed decision. However, we encourage investors to view our fundamental analysis charts (Cumulative returns, PE multiples, Cumulative Portfolio Profits vs S&P500 cumulative profit) to compare the superior historical performance of our renewable energy stocks against fossil fuel companies and the S&P500 as a starting point for their investment decisions, bearing in mind that past performance is not an accurate predictor of future performance.
We would have to pursue this idea in more depth and require more time to try new analysis and approaches to gather better data and results. That being said, Renewable Energy investments are in principal a great idea to explore further in the future.