https://github.com/sanketshevkar/sm445_theminerleague

Smart Hackathon 2020 FIS Global PS number: 445
https://github.com/sanketshevkar/sm445_theminerleague

Last synced: 28 days ago
JSON representation

Smart Hackathon 2020 FIS Global PS number: 445

• Host: GitHub
• URL: https://github.com/sanketshevkar/sm445_theminerleague
• Owner: sanketshevkar
• Created: 2020-08-01T09:16:39.000Z (over 4 years ago)
• Default Branch: master
• Last Pushed: 2020-08-05T07:18:39.000Z (over 4 years ago)
• Last Synced: 2024-11-07T08:51:00.498Z (3 months ago)
• Language: CSS
• Size: 13.6 MB
• Stars: 0
• Watchers: 2
• Forks: 1
• Open Issues: 0
• Metadata Files:
  • Readme: README.md

Awesome Lists containing this project

README

        
# Findex

## Problem Statement

Companies have to inform stock exchanges about

the corprate actions of the securities that are listed

on the exchange.

The same data is available on other websites in an

unstructured form.

In this problem statement we have to provide a

solution so that we can gather corporate action data

from uppaid sources.

//image

## Our Solution

Corporate action Miner Bot based on AI-ML.

### Web Crawler

User can search for a company and the crawler crawls through Financial Express and Money Control websites to search for the desired company name and most recent corporate actions.



### Manual Upload

1. Segment Classifier

   - The uploaded document is first classified into

     three major segments.

     - Equity

     - Mutual Funds

     - Debt

   - Algorithm - XGboost

   - Accuracy - 86.7%

2) Equity Classifier

   - Documents are further subclassified into:

     - Dividend

     - Book Closures

     - Stock Splits

   - Algorithm - Adaptive 1-D Convolutional Neural Network

   - Accuracy - 91.2%

   

3) NLP Models

   - Specific Details are extracted from the document for each class.

   - Algorithm - Trained Named Entity Recognition model in spacy for extracting relevant details.

4) Cloud Database

   - Final Corporate Action Details are uploaded to a MongoDB Cloud Instance.

5) Client Side

   - NodeJS REST-API connects to the MongoDB Cloud Instance and fetches and serves the data to the Client application.

   

## Architecture



 

# Built With

## **Client Side Tech-Stack**

    - NodeJs

    - ReactJs

    - React Native

    - MongoDB

    - AWS

## **Pipeline Tech-Stack**

    - Tensorflow

    - Flask

    - Spacy

    - Selenium

    - Scikit Learn

## Getting Started

Follow the instructions present in the README files inside each folder of the project.

### Prerequisites

You need to have anaconda, nodejs, npm installed as a prerequisite. Also you need to create and activate the anaconda virtual environment first.

```

conda env create --file findex_env.yml

conda activate findex_env

As a sudo/admin user run:

python -m spacy download en

```

# Performance

- Segment Classifier Accuracy - 86.7%

- Equity Classifer Accuracy - 91.2%

# Our Website

    [FinDex](http://35.154.254.195:3000/)

# Future Scope

We aim to develop, not only a small solution

but a whole mobile-first, cloud-first and

completely automated product for small

capped investors.

For enterprises, we aim to provide a flexible

product with plug and play modules to develop

a robust and automated solution.

# Acknowledgements

    [Spacy]https://spacy.io/usage/training

    [Bse]https://www.bseindia.com/corporates/ann.html

    [Hkex]https://www1.hkexnews.hk/search/titlesearch.xhtml?lang=en

    [PyMuPdf]https://pymupdf.readthedocs.io/en/latest/

    [Selenium]https://selenium-python.readthedocs.io/