KUSHAGRA SHAH

Analysis of Semantic Bounds in SDO using the NSGA-II - Research Assistant

   March 2022 - Aug 2022

• Varied UBSS threshold to promote semantic diversity in MOGP. Performed over 40000 independent runs on 6 binary classification problems.
• Using hyper-volume as a performance metric revealed that UBSS between 0.25-0.75 promotes a better average area of evolvedPareto-Front and PO-Front.

 Data Analysis on Rainfall Erosivity

   April 2022 

• Cleaned and transformed data of rain-station and precipitation of US.
• On a dataset of 51 states, performed PostgreSQL operations to find the wettest and driest state.
• Created raster of transformed data using QGIS and derived analysis on it by implementing IDW and TIN Interpolation, Contours and prepared a report.

Influence of Green Spaces on Psychological Health

   Nov 2021  

• Handled and captivated GEOJSON data analyse demographics of green spaces with effects on crime and mental health.
• Found Havering as the greenest(59.3%) and the City of London least(4.8%) green coverage.
• OLS regression model revealed a 1% increase in green space lowered depression rates by 0.137 per 1000.
• Crime rate and mental health account for 78% of depressive episodes per 1000 people.

Prediction and Analysis of student performance using a hybrid model of Multi Layer Perceptron and Random Forest

       Dec 2019                                     IEEE

• Analyzed multiple ML algorithms on a dataset of 480 students to predict the impact on the students’ performance.
• Divided the data in a 66:34 ratio of train and test, a hybrid model of MLP and Random Forest provided tangible results.

Contemplation of various Machine Learning Algorithms Under distinct datasets

       Aug 2019                                     IEEE

• Made a survey of 7 different machine learning algorithm under 3 distinct datasets.
• Figured out which algorithm best suited on particular type of datasets.

Neural Network and Deep Learning - Coursera Certified