Research Project :
Copper Alloy Discovery using Machine Learning
Until recently, the process of discovering new alloys has relied largely on a combination
of trial and error, expert judgment, and intuition. As a result, the discovery process
is slow and expensive.
In this collaborative project, we propose a cheap, fast and reliable Smart Alloy
Generation System that utilizes machine learning to obtain multi-element
alloy compositions and processing conditions with user-specified property parameters.
Includes -
- Web Scraping
- Predictive Modelling
- Model Optimization
- Python
- Scikit-Learn
Backtesting Wisdom - The Performance Analysis Engine for Algorithmic Traders
Automating the backtest performance analysis using data science and web development
best practices.
Algorithmic trading uses a computer program that follows a defined set of instructions to
place a trade. In practice, traders backtest their trading bots on historical prices to
estimate their potential future returns.
Unfortunately, there are hundreds of different combinations to backtest. Proprietary
software, such as MetaTrader, can perform these backtests and return the performance
results for these combinations individually. However, it can take up to weeks to
reliably estimate the HOLISTIC picture of a strategy's performance.
This project aims to reduce the amount of time traders spend on testing strategies.
Includes -
- Cloud Application Deployment on Heroku
- Database Setup and Deployment on Postgresql Server
- File Storage on AWS S3
- Backend with Python Flask
- Frontend with HTML, CSS & Bootstrap
- User Handling (Authentication and Security)
Scraping Product Information from Grocery Retailer's website
Aenean ornare velit lacus, ac varius enim lorem ullamcorper dolore. Proin aliquam
facilisis ante interdum. Sed nulla amet lorem feugiat tempus aliquam.
Includes -
- HTML Parsing with BeautifulSoup
- Data Structuring with Pandas
- Market Simulations and Analysis with Python
- Visualizations with Seaborn and Matplotlib
Optimizing Deep Learning Model Architecture using Evolutionary Algorithms
In this project we will develop an optimal performing Neural Network by making an
Evolutionary Algorithm to modify and optimize the following -
1. Number of Hidden Layers
2. Number of Neurons in each Hidden Layer
3. The Activation Function for each layer
4. Learning Rate
Includes -
- HTML Parsing with BeautifulSoup
- Data Structuring with Pandas
- Market Simulations and Analysis with Python
- Visualizations with Seaborn and Matplotlib
EigenFaces - Building a Facial Recognition System from Scratch
This project was done as part of my final year Computer Vision course at the
Australian National University.
Eigenfaces refers to an appearance-based approach to face recognition
that seeks to capture the variation in a collection of face images and using the
captured information to encode and compare images of individual faces.
In this project, I have used Python and images from the YaleFace database to build and
test such a system!
If you are interested, feel free to refer to the code and the step by step report of the
project on GitHub and implement this on your own.
PS - It's really fun!
Includes -
- Python
- Singular Value Decomposition
- Eigenvectors and Eigenvalues using PCA
- Visualizations with Matplotlib
Sentiment Analysis of Product Reviews from E-Commerce Clothing Store
Aenean ornare velit lacus, ac varius enim lorem ullamcorper dolore. Proin aliquam
facilisis ante interdum. Sed nulla amet lorem feugiat tempus aliquam.
Includes -
- State of the Art Pre-trained Sentiment Classification Neural Network - Flair
- Text Preprocessing with NLTK
- Creative Feature Engineering
- KMeans and Heirarchical Clustering
- Data Manipulation and reshaping with Pandas
Leveraging Unspervised Clustering for Identifying Traffic Mortality Trends in USA
Aenean ornare velit lacus, ac varius enim lorem ullamcorper dolore. Proin aliquam
facilisis ante interdum. Sed nulla amet lorem feugiat tempus aliquam.
Includes -
- Exploratory Data Analaysis
- Kmeans Clustering
- Result Analysis with Pandas
- Data Visualization with Matplotlib
Extracting the Hottest Topics in Artificial Intelligence from 10,000 Research Papers
Aenean ornare velit lacus, ac varius enim lorem ullamcorper dolore. Proin aliquam
facilisis ante interdum. Sed nulla amet lorem feugiat tempus aliquam.
Includes -
- Topic Extraction using Latent Dirichlet Allocation
- Natural Language Processing using NLTK
Scraping Product Information from Coles' website
This is web scraping library designed to facilitate easy extraction of product details
from the Coles website. This project uses Selenium for browser automation, following the
Page Object Model pattern to structure the code.
Includes -
- Browser automation with Selenium
- HTML Parsing with BeautifulSoup
