Stock Analyzer

Using sentiment analysis and algorithms to predict the directions of stocks and analyze global economy through news articles, as well as provide a dashboard which provides the user with lots of data.

Comment

Inspiration:

As the stock market grows, the inequality in financial resources has resulted in only the few profitting off individual stocks. The biggest reason for this is extremely expensive and private data and analysis which the average person could never afford, the Bloomberg terminal is $24,000 PER YEAR. We created stock analyzer in order to democratize computational financial analysis through sentimental analysis and algorithms. It can be

What it does:

Searches 10+ credible economic sources (CNBC, SeekingAlpha, MarketWatch, MotleyFool, FinancialTimes) and analyzes hundreds of articles about the stock through sentiment analysis.

How I built it:

Using the python modules BeautifulSoup and the Google Natural Language packages, an advanced webcrawler built from BeautifulSoup passed raw article text through the Google Natural Language engine to evaluate. Sentiment scores were rated on a scale from -0.9 to 0.9 with a magnitude, representing the strength of the sentiment. The sentiment scores from each and every article were aggregated to calculate the final average score for the stock.

Challenges:

We initially wanted to train our own model using the Fast.AI and the Stanford IMDB Movie Review Dataset. Given the time constraints we had to complete this project, it was unrealistic to complete due to a egregious training time. We also wanted to harvest data from Twitter using Tweepy (Twitter API -> Python), but the application process rendered it unrealistic as well. Finally, the Google Cloud Compute Engine was down for a prolonged time, delaying the progression of the project in its earliest and most crucial stages

Accomplishments that we're proud of:

We were able to compile a robust algorithm that had an uncanny extent of accuracy. The webcrawler we built is a dynamic web crawler that can search thousands of resources and detect unrelated articles and junk URLs, therefore optimized for large scale data collection and analysis. For example, AliBaba is expected to spike in revenue, therefore bolstering the stock in the coming week due to a nationwide sale. Our program was able to synthesize the positive sentiments from the detected article, labelling it with a average sentiment score of 0.60 on a -0.9 to 0.9 scale, -0.9 being a negative outlook to 0.9 being the most positive.

What I learned:

We were able to use our previous background in Python and web development to create something we didn't know we could do before. This was our first experience with Beautiful Soup and Google Natural Language API. We learned to successfully combine it to create something practical and effective

What's next:

We want to expand our sources and automate our process so users can get real time updates every hour or so instead of a manual system.

Built With

Share this project:

Updates