High Level Comparison of COVID Cases in Canada
Real time cases fetched from COVID-19 Canada Open Data Working Group API CCODWG. Data fed into SSIS live from web source into an ETL pipeline, then reported off of Tableau following Kimball Methodology.
Predicting COVID Hospitalization & Deaths in Haryana
Multiple models were compared using IBM Modeler, and Neural Networks was selected after comparing summary statistics. The timeseries GIFs were generated using PIL and geopandas in python
Image Classification Using Deep Learning (CNNs)
Collaborated on an algorithm that uses Gradient Boosting Classifier and Multidimensional convolution in scipy to classify satellite image tiles to color coded features.
JavaScript & MapBox API based Web Map
A Mapbox API based web map showing route between two points designed to be fetched from a GeoJSON. This was an attempt to solve a business problem for an Amazon DSP's self-owned courier website to show a route based on the start and end points given by customer on a quotation form.
Supervised Classification of Annapolis Valley, Nova Scotia
Satellite image captured from Landsat 8 OLI dated Sept 2019 was classified using Supervised Classification algorithm in PCI Geomatica using Maximum Likelihood Method on a total of 15 training areas later merged into 5 classes as shown.
Traffic Collisions - ETL and Prediciton Pipeline
A fully automated ETL pipeline in SSIS that loads data in MS SQL Server after staging, transforming and creating dimension & fact tables. The resultant data mart is used to compare the traffic collisions in HRM vs Canada and visualized in Power BI.
Displaying GPS Tracking points in QGIS after designing a spatial database in PostgreSQL
The GPS device log data along with other attributes was modelled in PostgreSQL as per shown ERD. Spatial queries were executed using PostGIS. The data was then exported to QGIS using a query layer and displayed in a map as shown.
Spatially Constrained Multivariate Clustering in Montreal using K Means Clustering
Optimum clusters were defined with Montreal as a DA (dissemination area) using Sturge's rule , C = 1 + log(n). Clustering was done in IBM Modeler using KMeans after comparing with Kohonen's and TwoStep algorithm. The insights were used in ArcGIS Pro to build the final solution.
Image Enhancement : Distinguishing Urban Sprawl from Green Patchess
Satellite imagery from USGS Earth Explorer captured from Landsat 8 OLI was used to enhance an image of New Delhi airport area (25x26 sq kms). Standard Linear Enhancement was used in PCI Geomatica after carefully examining the LUTs.
Creating Trees Geodatabase & Mapping Dominant Species
Created geodatabase from feature classes gathered from various sources including Nova Scotia Department of Natural Resources and Forestry GIS Data to map dominant tree species, tree type, stand age, & crown closure in an area.