Who Are We?

We are the Iron Bank of Bravos, Since we do business with all parties of Westeros, Its very Important for us to recognize who is a Lanister(Who Always pays his debts :P) and who is not.

Inspiration

So, we try to analyze and understand the various bodies involved in our present day Financial Market and predict their credibility. Investment market have been known to be vulnerable by ample of market makers, And today being the auspicious day of the return of GOT, we start by looking at one of the key players in the game, Investment Advisers/Brokers. As the brokers are closely related to both, the markets and the investors,and hold huge significance in keeping the markets clean and stable.

What it does

Here, we are trying to use the data (provided by FINRA) about the brokers of different firms and backgrounds to:

  1. Perform Unsupervised Learning and identify correlations & similarities in the high dimensional space
  2. Assess the credibility of the brokers and creating a metric for estimating their credibility.
  3. Also, We work on building a Delinquent loan classifier, and look at borrowers who have defaulted on their loans(Dataset Provided by RiskSpan).

How we built it

After pre-processing and understanding the highly categorical nature of the dataset, we started out by looking at unsupervised learning approaches to make sense of the data. We applied classical machine learning and deep learning algorithms on the dataset to cluster them into groups and the insights we gained from the exercise helped us build a metric to estimate their credibility. We then did the same analysis on the borrower end and built a classifier to identify potential delinquent loans.

Challenges we ran into

Pre-processing and wrangling the xml data was fairly difficult. The data processing was also a challenge, because of the size of the dataset. Building the metric for broker credibility also took a lot of thinking.

Accomplishments that we're proud of

Being able to make sense of the data and coming up with statistical measures that can be used in the future for prediction and classification.

What we learned

State of the art machine learning algorithms like Self-organizing maps and K-Mode Clustering, assessing the features to build a metric. We also tried deep learning methods like self organizing maps and cluster.

What's next for The Iron Bank of Bravos

We can move a step further and analyze the investors of markets to find out market manipulation patterns.

Share this project:
×

Updates