Healthcare,REAL-LIFE PROJECT

Appointment No Show Prediction

One of the real-life data science projects I have worked on with one of the best healthcare startups in the Silicon Valley. Appointment No Shows are a significant problem in the US-healthcare industry, and costs the hospitals and clinics a lot of money. The aim of this project was to build a model to predict patient no-shows in advance to improve patient health and reduce the costs. The final result was a model with an AUC-ROC score of 0.85 and a potential saving of $40000 anually for one of the business clients.

Finance, SVM , Algorithmic Trading

Using SVM to predict Buy/Sell for Reliance Stock

Comparing different SVM Kernels, specifically the linear kernel, polynomial kernel, and the Radial Basis Function Kernel, in their effectiveness when applied to predict buy/sell of Reliance Stock, on top of technical indicators(which are used by traders before making decision) as features for the machine learning model. Framed as a binary classification problem, Buy/Sell - Buy - if price is predicted to rise, Sell - if price is predicted to drop.

View Project

Graph Theory, Social Network Analysis, Machine Learning

Social Network Graph Link Prediction : Follower Recommendation

In this project, given a directed social network graph, the objective was to predict missing links to recommend Friends/Connections/Followers, using just the information of existance of a link(a directed edge in the graph) between 2 users, and leveraging the same to predict missing links in the graph(new possible users to follow). This is an important functionality for any social network to improve user experience. The dataset used for this project was an open-sourced dataset from Facebook.

View Project

Finance, Exploratory Data Analysis

Lending Analysis in Financial Sector

Lending loans to ‘risky’ applicants is the largest source of financial loss(called credit loss) for any bank/lending company. If we are able to identify these risky loan applicants, then such loans can be reduced thereby cutting down the amount of credit loss. Identification of such applicants using Data Analysis is the aim of this case study. Lending Club (a peer-to-peer lending company) wants to understand the driving factors behind loan default. The company can utilise this knowledge for its portfolio and risk assessment.

View Project

Sentiment Analysis, NLP, Deployment(Flask)

Twitter Sentiment Analysis

Human preferences are practically unpredictable. That's where the science of Psychology and Sociology comes in. This project demonstrated how we can prototype an NLP based sentiment analyzer within minutes to analyze sentiments on a variety of tweets. It is relatively a small side-project but useful. Finally, I delpoyed it using Flask API to make it more usable by anyone across the web.

View Project