Ankur Patel

To track the health of New Yorkers to influence health program decisions, reports were made to see trends analyze the dimensions. The robust data included neighborhood, borough, and citywide estimates on a broad range of chronic diseases and behavioral risk factors.
Programs: Tableau, Excel, Python, Word

The bank has been seen unusual churn rates for their customers (churn is when people leave the company). I created a classification model to tell the bank which of the customers are at the highest risk of leaving.

The Polish Companies Bankruptcy Data Set (http://archive.ics.uci.edu/ml/datasets/Polish+companies+bankruptcy+data) will be used for building a classification model to predict companies that will bankrupt. Visit the link for the details of this dataset.

-First place winners
Research and innovation for sustainable improvements in welfare, which was the goal for the Estee Lauder Companies hackathon. We made a reward system for recycling the empty ELC containers using technologies listed below.
Software - CloudSQL, VisionAPI, Python
Hardware - AmazonEcho, Arduino, IRSensor, LEDDisplay

The Human Resource team who hires employees in the company analyze salary data before offering. As a data scientist helping the HR, I built a regression model to predict if the new candidate is speaking the truth or bluffing.

The churn dataset was used to predict the churn rate of the customers. It's the percentage of subscribers to a service who discontinue their subscriptions to the service within a given time period.

Using the Energy Efficiency Data Set (http://archive.ics.uci.edu/ml/datasets/Energy+efficiency), a regression model will be built to predict the y1 Heating Load and y2 Cooling Load. Firsly, an EDA will be performed to overview the dataset, and the data will be preprocessed. After that, the model will be built using multiple linear regression, ridge and lasso regressions.

The first routing app that provides the safest routes by using ArcGIS API and city wide traffic statistics. The app works by predicting the probability of vehicle crashes at a certain location given several inputs like precipitation chance and amount, temperature, time of day, past crashes, and sinuosity of the road.
Programs: Python, JSON

2nd place- Group developed a SaaS app called Seeker that uses Google Cloud Video Intelligence API for object detection and time intervals, and Yuuvis for storage and Search Service API to jump to time of searched object
Programs: Python, JSON

- Created a web app that predicts MLB game outcomes for user's fantasy team
- Analyzed MLB players' data from 2018 to show users the comparison and rankings
Programs: Python, HTML, JSON

An IoT device with of an arduino attached with sensors of temperature, humidity, and vibration/shocks that logs data back to the cloud. It solves lack of visibility of the cargos' condition in transit. It can also be used by insurance companies for filed dispute claims.

Created NLP modules to extract features of text using Apache Spark for big data of 4 million Amazon reviews
Program: PySpark

Competing against 6000 participants from 24 countries, group created an audio and image classification app to notify users of real-time and location-based crime incidences for evacuation and to get connected with lawyers during claim cases
Programs: Python, JSON, HTML

Multilayer perceptron (feed forward neural network) with hidden layers on iris dataset
CNN for image classification in MNIST, SVHN, and Cats & Dogs dataset
Program: Python

Women’s fashion online shop reviews analyzed to predict whether the review author would recommend the product using NLP.
Program: Python

Predicted forest cover types from classification of cartographic variables of 580,012 measurements. I used Logistic Regression, LinearSVC, Nearest Centroid, Decision Tree, and Random Forest classifiers with Bagging, VotingClassifier, and GridSearchCV ensemble methods.
Program: Python

The ann-thyroid dataset from UCI Machine Learning Repository was analyzed to predict the hypothyroid class using PCA, TNSE, and clustering with KMeans, Agglomerative, and DBSCAN. After viewing in a Decision Tree, API and NMI was used to evaluate the performance. EllipticEnvelope, OneClassSVM, and IsolationForest were used to detect outliers, and the AUC and precision was visualized.

Predicted California house prices using a 1990 Census dataset of 20,640 observations. I used regression methods of Linear Regression (OLS), Ridge, Lasso, ElasticNet, and GridSearchCV, and Machine Learning techniques to cross validate using GridSearchCV.
Program: Python

Employed machine learning, ensemble, and statistical analysis for feature selection and tuned models for predicting heart disease with most accuracy
Program: R, IRKernel

Phone app that assists visually impaired people in providing mobility and independence through Machine Learning and Artificial Intelligence
Programs: Python (written by Divyendra)

Automobile engine factors on vehicle's mpg: Investigated the impact of a number of automobile engine factors on the vehicle's mpg. Made a linear model, used it to make predictions, and compared with actual values using error differences. Created a residual plot and histogram for the errors.
Program: R, IRKernel

Length of stay predictions to avoid readmission using 89 attributes for liver disease patients. After EDA, created models using sklearn linear models and ensemble.
Programs: Python, R, Excel

Developed several GUI using frameworks like Kivy, Tkinter, and Flask in Python.
Programs: Python, KV, HTML

On Google Cloud Platform console, created a project in which BigQuery service was used to import csv files to query.
Program: SQL

Using Python, created table and loaded a data set into a local running Postgres server. Also developed Python database applications with the PostgreSQL database server.
Program: Python, PostgreSQL

Finance Analytics modeling with time series data for stock returns, data visualization, volatility, OLS, and back testing.
Program: Python, Quantopian

Using Google Drive API and Google Sheets API, downloaded a JSON file with credentials to connect that were used in a Python script to process.
Program: Python, Google Sheet API, Google Drive API

As a Graduate Assistant, assisted the team to create a model for optimal asset allocation for US retirees using Bloomberg data from 1983 to current. The model was a machine learning classification algorithm, which was selected after coming multiple, and based on recession and pullbacks during 60 or 120 month rolling horizon.
Program: Python, Bloomberg Terminal

Using a dataset of revenue earned by sales reps in time interval, analyzed by using functions in Excel. Created a report using pivot table with progress chart and dashboard. Also used Python to create and modify the Excel sheets.
Programs: Excel, Python

Assists new artists in getting recognized by connecting them with influencers to promote their music to users. Incentives are set in place by the artists to attract influencers. Underlying the platform, all transactions are tracked by use of Blockchain.

Web scraping in Python using BeautifulSoup and requests libraries for articles on Quartz in America. The data was then written to an Excel.
Similarly, the FedEx Newsroom's default limited articles were scraped, and the data was saved to an Excel.
Program: Python

Association rule mining is a technique to identify underlying relations between different items. An association rule has two parts: an Antecedent (if) and a Consequent (then). There are three major components of Apriori algorithm: Support, Confidence, and Lift. In this data set, Churn will be only consequent while others can be either.

The Churn data set will be used to predict Churn depending on VMail Plan, Intl Plan, and CustServ CallsSet (which will be set ordinal).

Using Starbucks location data from Kaggle, Transportation Ridership data, and Census Bureau API for Census tracts and demographic data, developed a logistic regression model to predict probability of Starbucks in a census tract.
Program: Python, Tableau

BIRCH, which stands for Balanced Iterative Reducing and Clustering, is an algorithm appropriate for large data sets. Here, we will classify loan applications as approved or not (the flag response variable), using the following predictors: debt-to-income ratio, FICO score, and request amount.

Analysis of 2020 Democratic Presidential Candidates:
Cory Booker, Kamala Harris, Amy Klobuchar, Bernie Sanders, Elizabeth Warren
Analyzed tweets about the candidates from Apr 4-11 uniformly by state, and tweets from the candidates' personal Twitter accounts from Nov 1 - Mar 31.
Program: R

Tested the metabolic phenotype of mice in different conditions – glucose, insulin, and pyruvate – by collecting and analyzing blood glucose levels.
Program: Glucometer, Excel

Analyzed bodyweight of mice in different conditions over time.
Program: Python, Excel

Identified the genotype of mice by examining their DNA sequence by polymerase chain reaction.
Program: Spectrophotometer, Excel

Tested for potentially harmful bacterial contamination through protein immunoblot, an analytical technique to detect specific proteins.
Program: Spectrophotometer, Excel

Performed patent research; established a business model design; and developed a para-percussion assembly to promote physical activity and entertainment for wheelchair users

Assisted in column chromatography and performed mass spectrometry, which is automatic metadata collection, to measure protein's purity; created graphs of the data in Graphical Analysis software.
Program: Spectrophotometer, Graphical Analysis