Data Science Projects

Forecast Daily Department Sales of Walmart Stores
Sales Forecasting | Timeseries (ARIMAX, LSTM)
Walmart’s hierarchical sales data for three US states (CA, TX, WI) is used to forecast daily sales for the next 28 days. Forecasting accuracy is important for appropriate inventory or service levels. I used traditional statistical as well as ML algorithm. Finally combination of both statistical and ML model helped to improve forecasting accuracy.

Predict Accent of a Speaker
Speech Recognition | Deep Learning (CNN, BLSTM)
Accented speech poses a major obstacle for automatic speech recognition algorithms. In this project, we built an accent classifier on top 5 accents of common voice mozilla dataset using deep learning models.

Topic Modelling on Covid-19 reaserch articles
Text Mining | NLP (SciSpacy, Topic Modelling)
CORD-19 is a repository of over 158,000 scholarly articles about COVID-19, SARS-CoV-2 and coronaviruses. There is a growing urgency to organize this information for the medical research community to keep up. We deployed NLP techniques to derive meaningful topics.

WiDS Datathon Challenge 2020
Healthcare Analytics | Tree Ensemble(LightGBM)
MIT’s GOSSIS community initiative has provided a dataset of more than 130,000 hospital Intensive Care Unit (ICU) visits from patients, spanning a one-year timeframe. The objective was to create a model that uses data from the first 24 hours of intensive care to predict patient mortality.

Predict Earthquake Damage
Geospatial | Clustering (KMeans, LCA), Tree Ensemble (LightGBM, Random Forest, Gradient Boosting)
Based on aspects of building location and construction style, our goal was to predict the level of damage caused to buildings by an event similar to the 2015 Gorkha (Nepal) earthquake.

Effect of Lifestyle on Ageing
Healthcare Analytics | Random Forest, Gradient Boosting
Using patient lifestyle data (demographics, economic, health, education aspects), developed a model to predict likelihood of ‘Arthritis’, ‘Angina’ and ‘Chronic Lung’.

West Nile Virus Prediction
Healthcare Analytics | XGBoost, Random Forest, Stacked Model
To aid City of Chicago & CPHD in development of effective solutions against deadly virus transmissions, created a predictive model for West Nile Virus based on weather, location & mosquito repellant spraying data