Default profile banner
NC

Nikita Chopra

@nikitachopra

Freelancer Data Scientist at Upwork

Bangalore, India

UpworkIndian Institute of Science Bangalore

Nikita Chopra is an experienced Data Scientist with expertise in Machine Learning, Deep Learning, and Program Analysis. Professional experience includes developing predictive models for stock market analysis, customer segmentation, and churn prediction using techniques like LSTM, K-Means, and XGBoost. The candidate also has a strong academic background, including research on data races and static analysis for interrupt-driven kernels.

Experience

Freelancer Data Scientist

Upwork

FreelanceMar 2001 - Present

Stock Market Prediction: Used price and volume data of various US-based stocks available on Kaggle, to predict future stock prices. Worked with LSTM and various traditional ML models for this purpose. Sentiment analysis on product reviews: Scraped product reviews of various products from Flipkart and Amazon, and performed sentiment analysis on the reviews, to suggest the customer which website to buy a product from. World Happiness analysis: Used World Happiness data available on Kaggle, to determine the factors influencing happiness in a country.

Young Technical Leader

Airtel X Labs

Sep 2001 - Aug 2001

Device Recommendation Engine: Used calls, data usage, demographics, ARPU, messages counts’, complaints counts’, bundle information, apps usage flags and prior handset information to generate features. Used Random Forest Classifier on these features to predict whether a customer would jump from 2G/3G to 4G. Model Accuracy - 90%, Model Recall - 60%. B2B Churn Prediction: Used base data, SR data and utilization data to generate various features. Used XGBoost model to predict whether a B2B customer would continue using Airtel services or not. Model Accuracy - 65%, Model Recall - 65%.

Associate Data Scientist

Airtel X Labs

Sep 2001 - Mar 2001

Network Site Planning: Trained a semantic segmentation model to identify different classes - buildings, vegetation, barren land etc. This was done for rural sites, we first obtained the satellite images using Google Maps API, then ran the model to obtain the masks. We were able to achieve ≈ 0.6 IoU score for buildings and vegetation. Used the obtained masks to get density values for different classes and used these to estimate the number of buildings, and hence population and revenue within a certain radius. Customer Segmentation: Used raw data to find out the pattern in a customer’s location data, migrant label, rural/urban label, primary/secondary SIM label. Created 50 features using calls, messages and device data. Used K-Means clustering on these features to find 17 clusters in the data.

Education

Indian Institute of Science Bangalore

M.Sc.(Engg)

Computer Science

Jan 2015 - Jan 2018Grade: 6.0/8.0

Thesis Title: Data Races and Static Analysis for Interrupt-Driven Kernels

University of Delhi

M.Sc.

Computer Science

Jan 2013 - Jan 2015Grade: 89.6%

Thesis Title: Evolution of Conferences’ Co-Author Communities

Hansraj College

B.Sc. (Hons.)

Computer Science

Jan 2010 - Jan 2013Grade: 91.025%

Skills

Python
R
pandas
numpy
sklearn
opencv
PIL
matplotlib
Keras
Tensorflow
Pytorch
Spark
PySpark
Hadoop
LSTM
K-Means clustering
XGBoost
Random Forest Classifier
Semantic Segmentation
Sentiment Analysis