Graduate student with professional experience in data analytics, programming, and management with a desire to help businesses make data-driven decisions.
-
Experience
Simple Energy Pvt. Ltd,
Bengaluru- India FEB ’19 – AUG ‘19
Junior Data scientist
• Performed Univariate, Multivariate and multivariate LSTM(long short term memory networks) to analyze the pattern and predict future performance of vehicles. This was further communicated to the vehicle and other vehicles to avoid points of failure.
• Pre-processing and feature engineering of data was done before data was analyzed and visualized.
• Increased the performance of the vehicles by 60% by identifying points of failure and breakdown
Techie Aide, Bengaluru- India
SEP ’17 – FEB’18
Data Science Intern
• Performed analytics using Time series forecasting for BMRCL ie Bengaluru Metro trains
• Used Business forecasting using Time series for prediction of consumer traffic, month-wise
• Performed AR, MA, ARIMA, data cleaning, feature extraction, model selection, data analysis and data visualization of the time series data
• Forecasted the traffic income expectancy which helped in the increase of annual sales by 30%
-
Projects
• Built a classification model using Text analytics that distinguishes a fake video from a new informative video on YouTube by creating a corpus of text which is created by using TF-IDF to calculate the most important words for Breast cancer.
Manipal Hospital Harvard business review Case study:
Performed data cleaning by removing NA values, imputing NA values using KNN
Checked specificity and sensitivity of 3 class problem, found that the class was imbalanced and used techniques of Under-sampling, Over-sampling and SMOTE techniques to balance data
Identified problem of Quasi-complete separation and dealt with it by collapsing categories of predictor variable and running logistic regression by excluding the variable
Performed prediction using ensemble methods such as Random forest and Ada boost for multiclass and binary classification. For better results decided to make NPS score variable binary from multiclass