https://github.com/rishisankineni/quovo-data-challenge

Data Challenge - Data Scientist position - NYC startup
https://github.com/rishisankineni/quovo-data-challenge

data-science machine-learning natural-language-processing

Last synced: about 1 month ago
JSON representation

Data Challenge - Data Scientist position - NYC startup

• Host: GitHub
• URL: https://github.com/rishisankineni/quovo-data-challenge
• Owner: RishiSankineni
• Created: 2016-10-22T23:13:58.000Z (about 8 years ago)
• Default Branch: master
• Last Pushed: 2017-05-02T17:16:28.000Z (over 7 years ago)
• Last Synced: 2024-06-24T23:28:36.947Z (6 months ago)
• Topics: data-science, machine-learning, natural-language-processing
• Language: Jupyter Notebook
• Homepage:
• Size: 595 KB
• Stars: 1
• Watchers: 2
• Forks: 3
• Open Issues: 0
• Metadata Files:
  • Readme: README.md

Awesome Lists containing this project

README

        
# Quovo - Data Science startup based in NY.

# Data Challenge - Data Scientist Position.

Quovo - Quovo is a provider of account aggregation and data analytics technology for finance.

## Total Equity Funding

$15.2M in 3 Rounds from 8 Investors

# Challenge details:

Thank you for taking the time to apply to Quovo. We like to send potential candidates a SHORT coding test/exercise so 

we could get a sense of how they approach problems. This also gives you the a good opportunity to see if Quovo-style 

challenges are a good fit for you. Don't go crazy on time, we'd just like to see enough progress on it where we can 

all have a conversation looking at your code together and talk about how you attacked the problem.

--

In each row of the included datasets, products X and Y are considered to refer to the same security if 

they have the same ticker, even if the descriptions don't exactly match. 

Your challenge is to use these descriptions to predict whether each pair in the test set also refers to the 

same security. The difficulty of predicting each row will vary significantly, so please do not aim for 100% accuracy. 

There are several good ways to approach this, and we have no preference between them. 

The only requirement is that you do all of the work in this file, and return it to us.

Hint: Don't be afraid if you have no experience with text processing. You are in the majority of applicants. Check out this algorithm, 

and see how far you can go with it:

https://en.wikipedia.org/wiki/Tf–idf

http://scikit-learn.org/stable/modules/generated/sklearn.feature_extraction.text.TfidfVectorizer.html

Good luck!

# My Approach

Calculated Unigram, Bigram,and Trigram. Used XGboost and log-loss to predict the probability instead of 0 or 1 prediction. You can take a look at the Quovo-submiss.ipynb for the code and issimilar_predicted for the output values. Also, you can use this link http://nbviewer.jupyter.org/github/RishiSankineni/Quovo-Data-Challenge/blob/master/Quovo-Submiss.ipynb to share the file. Thanks - Rishi

log loss- 0.56 (not bad). Nevertheless, will try to improve it by approaching the problem in a different way.