Quantitative Analyst.
Raptorarima, is a python package which is an implementation of Auto Regressive Moving Average Model from scratch using python. This API Supports to model time series data using box-jenkins model. Please find API code here.
Raptorfinance, is a python package which retrives stock price via webscrapping YahooFinance. This API Supports to retrieve data especially works for NSE/BSE stocks. Please find API code here.
GPA: 3.89/4
GPA: 9.3
GPA: 3.6
Co-developed a C++ application to extract earnings and price data for Russell 1000 stocks and impact on stock price. Implemented a preprocessing module to retrieve and clean stock price from yahoo finance using LibCurl. Classified stocks into 3 groups based on surprise percentage of EPS. Performed random sampling from each group of 30 stocks and computed excess return with SPY as benchmark. Calculated AAR for each group and deviation to compare impact of EPS on stock price.
We believe that its very crucial to safegaurd a person's digital presence. Especially in today's world where tech is very much included in our lifestyle. We developed a web application which gives the user a unique Identity based on face snap, and other users can search someone else's profile to check if they are who they are claiming to be. Please visit here to checkout our application. Please watch this demo video of our working beta to understand the application better.
This repository contains various trading strategies which use zerodha kite endpoints to make trade decisions given user authorised credentials. Please check out the code repository here , which contains implementation of long short beta, mean reversion, open range break and volume weight average pricing strategy.
Assessed stock returns to identify optimal allocation of capital for a stock portfolio and trained a model using historical data of S&P 500 for selected technology stocks. Used Pandas and Seaborn for EDA in Python. Generated a Random assignment of weights to portfolio of stocks, allocated ideal weight by leveraging efficient frontier considers stocks with highest Sharpe ratio. Utilized the past data of portfolio with static optimal weight to predict future portfolio returns applying linear regression and LSTM with 82% and 89% accuracy respectively.here
Applied Boost, STL library and object-oriented principles to create an option pricing system. Applied Black Scholes pricing method for European options, developed functions to compute Greeks. Employed a numerical method pricing with Monte-Carlo and finite difference methods for pricing European Options.here
The project states whether buying a particular stock is a good or bad investment based on recent 10k tweet’s polarity based on the stock and historical performance of the stock. The Twitter sentiment analysis retrieves a list of the last 10000 tweets posted in english containing the symbol introduced and they are later stored in a list of Tweet class, defined in tweet.py with the tweet's text and polarity from TextBlob. Please check out the code repository here.
The ARCH or Autoregressive Conditional Heteroskedasticity method provides a way to model a change in variance in a time series that is time dependent, such as increasing or decreasing volatility. An extension of this approach named GARCH or Generalised Autoregressive Conditional Heteroskedasticity allows the method to support changes in the time dependent volatility, such as increasing and decreasing volatility in the same series. Implemented the project in python, please find the code repository here.
Face mask detection using local camera frame by frame and classifies the frame with mask or without mask with probability and boundary box. I used 7959 images to train the models. The dataset is composed of WIDER Face and MAFA. Please check out the code repository here .
We used ‘yfinance’ API in python to collect the data and using user input we collected the dataset between a given time interval. We normalized the Dataset using MinMaxScaler. Using ‘Open,High,Close,Low,Volume’ with batchsize=2 and Time_Steps = 2 as hyperparameters to LSTM, we predicted the stock closing price of google with 87% Accuracy. Please check out our code here
As part of our major project we used Convolutional Neural Network to train our model on the Ravdess voice dataset, our model classifies each voice clip into one of 8 different emotions. As part of data preprocessing we first translated voice data into voice graph images using Librosa python library and then trained the CNN model on the image dataset. We classified the image graph data using our own CNN network architecture ( 2 conv layers, 2 relu, softmax layer and 2 dropout layers), we used gradient descent to optimise the cost function of classification error. Please find our code repository here , this link redirects to our project documentation.
Project for recommendation system on Amazon Product Purchase Data set. A popularity based recommendation system. Created an instance of User Similarity based Collaborative Filtering model. Please check out our code here
Used Headline dataset to classify sarcasm level, instead of tweet dataset as Headlines are more professional and syntactic in nature. Created features from headline database and generated weight matrix using GloVe embeddings. Used a Bidirectional LSTM model to train the data based on the features computed from the dataset. Please check out the code repository here .
We have used German Credit Dataset and Classifiying the users who are defaulting credit.Using RFC feature importance, we are perform feature engineering to identify the important features. Using Ensemble techniques (Random Forest/Adaboost/ Gradient Boost) we are able to achieve an accuracy of 90% on Train and 80% on Test data. here
Solving programming challenges in Leet code and Hacker rank.
Apart from being a Quantitative Analyst, I enjoy meeting new ppl and reading books.
PS:If you like coffee, we have much to discuss. ;)