Jivitesh Poojary

Our team 'Random Variables' won the Indiana Medicaid Data Challenge. Our solution focused on identifying regions in the State of Indiana where mental health care facilities could be improved.

We were awarded a certificate, an opportunity to present our analysis at the Midwest IT conference and a cash prize of $1000. The presentation was given to decision makers from - challenge organizers Regenstrief Institute, Inc., KSM Consulting, Indiana HIMSS Chapter, Indiana Family and Social… Show more

Our team 'Random Variables' won the Indiana Medicaid Data Challenge. Our solution focused on identifying regions in the State of Indiana where mental health care facilities could be improved.

We were awarded a certificate, an opportunity to present our analysis at the Midwest IT conference and a cash prize of $1000. The presentation was given to decision makers from - challenge organizers Regenstrief Institute, Inc., KSM Consulting, Indiana HIMSS Chapter, Indiana Family and Social Services Administration and Indiana MPH.

https://public.tableau.com/profile/jivitesh.poojary1464#!/vizhome/INMedicaidChallenge-MentalHealth/ProjectOverview?publish=yes

https://prezi.com/view/w8lmPrFwuUAa4oclYSyI/
Show less

We aim to build a repository of Shiny apps where the user can create quick visualizations by uploading data. Some of the visualization types include:
- Sankey flow diagram
- Stream graph

- The dataset is from a mobile blood donation vehicle in Taiwan. The Blood Transfusion Service Center drives to different universities and collects blood as part of a blood drive. We want to predict whether or not a donor will give blood the next time the vehicle comes to campus.
- Using exploratory data analysis some of the data unnecessary attributes were removed. This step was performed in R.
- The class probabilities were predicted using the predict_proba feature in decision tree and… Show more

- The dataset is from a mobile blood donation vehicle in Taiwan. The Blood Transfusion Service Center drives to different universities and collects blood as part of a blood drive. We want to predict whether or not a donor will give blood the next time the vehicle comes to campus.
- Using exploratory data analysis some of the data unnecessary attributes were removed. This step was performed in R.
- The class probabilities were predicted using the predict_proba feature in decision tree and random forrest classifiers of the scikit package.
- Data is courtesy of Yeh, I-Cheng via the UCI Machine Learning repository Show less

- The objective of the competition was to forecast outcomes of all possible match-ups in the 2017 tournament.
- Primarily used five out of the eight available files, the files were joined and a new variable was created to obtain the difference between the seed rank for a given year.
- Tried two techniques - a logistic regression model and Elo benchmark
- Was able to reach top 25% of the participants

The data primarily captures road accidents in UK during 2015 and has 70 features/columns and about 250K rows. Data has been fetched from Open Data Platform UK and is being shared under Open Government Licence. For more details refer to Open Data UK.
Link - https://www.kaggle.com/akshay4/road-accidents-incidence

- Our objective was to explore the data and predict the severity of the road accidents.
- This being a real data, a significant proportion of time was spent in cleaning the… Show more

The data primarily captures road accidents in UK during 2015 and has 70 features/columns and about 250K rows. Data has been fetched from Open Data Platform UK and is being shared under Open Government Licence. For more details refer to Open Data UK.
Link - https://www.kaggle.com/akshay4/road-accidents-incidence

- Our objective was to explore the data and predict the severity of the road accidents.
- This being a real data, a significant proportion of time was spent in cleaning the data
- Because of the nature of the response variable, logistic regression was the model of choice. Show less

The Places & Spaces: Mapping Science exhibit introduce science mapping techniques to the general public and to experts across disciplines for educational, scientific, and practical purposes. Website - http://scimaps.org/

- We analysed the Apache command logs provided by the client. data cleaning was performed in python to remove unwanted characters in the URLs and also to remove URLs which seems to be accessed by a crawler or a bot.
- Then data was aggregated in XML, JSON and CSV… Show more

The Places & Spaces: Mapping Science exhibit introduce science mapping techniques to the general public and to experts across disciplines for educational, scientific, and practical purposes. Website - http://scimaps.org/

- We analysed the Apache command logs provided by the client. data cleaning was performed in python to remove unwanted characters in the URLs and also to remove URLs which seems to be accessed by a crawler or a bot.
- Then data was aggregated in XML, JSON and CSV formats to integrate with the base visualizations.
- The sankey flow diagram was implemented in R Shiny using NetworkD3, the word cloud was implemented in Gephi, the bot activity tracking and tree map were implemented in Tableau.
- The visualizations helped the clients gain insights regarding improving their website to increase user traffic. Show less

- The data obtained from Federal Reserve of St. Louis’ FRED data repository, contains the unadjusted U.S. monthly unemployment rate from January 1948 to January 2017.
- We analysed the data by breaking this time series down into components (trend, seasonal and oscillation), and finding out what might predict the unemployment rate.
- After studying different possible causal parameters we concluded that percentage change in Real Growth Domestic Product (RGDP) was inversely related to… Show more

- The data obtained from Federal Reserve of St. Louis’ FRED data repository, contains the unadjusted U.S. monthly unemployment rate from January 1948 to January 2017.
- We analysed the data by breaking this time series down into components (trend, seasonal and oscillation), and finding out what might predict the unemployment rate.
- After studying different possible causal parameters we concluded that percentage change in Real Growth Domestic Product (RGDP) was inversely related to Unemployment rate, and on an average had a time series lag of up to 1.5 years in advance.
- The data can be obtained here - https://fred.stlouisfed.org/ Show less

- Used R (ggplot2) to visually explore the data of hurricanes in the US, to study if there was a meaningful difference between the distribution of damage caused by hurricanes with female names and the distribution of damage caused by hurricanes with male names
- We observed that the data had a bias in the naming of hurricanes, and concluded that any observed difference between the damage caused by feminine-named and masculine-named hurricanes, was due to a few outliers.
- A 2014 paper… Show more

- Used R (ggplot2) to visually explore the data of hurricanes in the US, to study if there was a meaningful difference between the distribution of damage caused by hurricanes with female names and the distribution of damage caused by hurricanes with male names
- We observed that the data had a bias in the naming of hurricanes, and concluded that any observed difference between the damage caused by feminine-named and masculine-named hurricanes, was due to a few outliers.
- A 2014 paper published in PNAS was titled “Female hurricanes are deadlier than male hurricanes.” The paper can be found here - http://www.pnas.org/content/111/24/8782.full Show less

- The datasets contains transactions made by credit cards in September 2013 by european cardholders. This dataset presents transactions that occurred in two days, where we have 492 frauds out of 284,807 transactions. The dataset is highly unbalanced, the positive class (frauds) account for 0.172% of all transactions.
- The primary objective of the project was to reduce the false negatives at a same time maintaining the accuracy.
- Our approach was bootstrapped random undersampling of the… Show more

- The datasets contains transactions made by credit cards in September 2013 by european cardholders. This dataset presents transactions that occurred in two days, where we have 492 frauds out of 284,807 transactions. The dataset is highly unbalanced, the positive class (frauds) account for 0.172% of all transactions.
- The primary objective of the project was to reduce the false negatives at a same time maintaining the accuracy.
- Our approach was bootstrapped random undersampling of the majority class followed by bagging of these indivisual subsamples.
- We applied Random Forest Classifier, Adaboost and XGBoost iteratively to obtain the best parameters for each of the models.
- The implementation was done uisng the scikit machine learning toolkit in Python.
- The best result we obtained was of zero False Negatives with 60% accuracy for Random Forrest Classifier
- The accuracy was compared with brute force application of Random Forrest algorithm using Weka Show less

- The objective of the project was to accurately predict the house prices in Ames, Iowa as part of the Kaggle competition “House Prices: Advanced Regression Techniques”. The data was provided by Dean De Cock from Truman State University.
- Feature selection was performed in R using Boruta technique.
- Different approaches Linear Regression, Logistic Regression, Lasso, Ridge, Adaboost and XGBoost were tried to obtain the best results. This implementation was performed in Python using the… Show more

- The objective of the project was to accurately predict the house prices in Ames, Iowa as part of the Kaggle competition “House Prices: Advanced Regression Techniques”. The data was provided by Dean De Cock from Truman State University.
- Feature selection was performed in R using Boruta technique.
- Different approaches Linear Regression, Logistic Regression, Lasso, Ridge, Adaboost and XGBoost were tried to obtain the best results. This implementation was performed in Python using the Scikit Machine Learning libraries.
- We obtained the best results using the XGBoost algorithm giving us a rank in the top 20% of the competition. Show less

- The objective of the project was to predict a space-delimited list of ads associated with a display block, ordered by decreasing likelihood of being clicked.
- Applied maximum likelihood estimation and probability adjustment using Adaptivesmoothing techniques, used Follow the Regression Leader (FTRL – Proximal) algorithm for predicting the probability of ads being clicked.
- Implementation was done in python on IU High Performance Computing servers.
- The evaluation metric for the… Show more

- The objective of the project was to predict a space-delimited list of ads associated with a display block, ordered by decreasing likelihood of being clicked.
- Applied maximum likelihood estimation and probability adjustment using Adaptivesmoothing techniques, used Follow the Regression Leader (FTRL – Proximal) algorithm for predicting the probability of ads being clicked.
- Implementation was done in python on IU High Performance Computing servers.
- The evaluation metric for the project was Mean Average Precision (MAP) with 12 as the number of predicted nodes
- We obtained the best results using FTRL-Proximal giving us top 15% rank in the competition. Show less

- Real Estate dataset: Analyze the dataset in understanding the impact of certain attributes in prediction of property prices.
- General Social survey dataset: Performing multivariate regression analysis, getting correlation and variance inflation factorsfor understanding the health index
- Baseball dataset: Executing bivariate regression analysis for understanding the attendance at a game in relation to win rate,salary, the type of league, etc.
- Economic survey dataset: Performing… Show more

- Real Estate dataset: Analyze the dataset in understanding the impact of certain attributes in prediction of property prices.
- General Social survey dataset: Performing multivariate regression analysis, getting correlation and variance inflation factorsfor understanding the health index
- Baseball dataset: Executing bivariate regression analysis for understanding the attendance at a game in relation to win rate,salary, the type of league, etc.
- Economic survey dataset: Performing ANOVA on few attributes and understanding the variation in the data Show less

The project was for a leading organization in electronic signature technology and Digital Transaction Management services for facilitating electronic exchanges of contracts and signed documents, where we helped them improve their quote to contract processes using a Apttus CPQ-CLM systems.

The solution enabled clients the sales, revenue and legal teams to improve their productivity significantly by seamlessly connecting the sales and contract processes into a single, efficient… Show more

The project was for a leading organization in electronic signature technology and Digital Transaction Management services for facilitating electronic exchanges of contracts and signed documents, where we helped them improve their quote to contract processes using a Apttus CPQ-CLM systems.

The solution enabled clients the sales, revenue and legal teams to improve their productivity significantly by seamlessly connecting the sales and contract processes into a single, efficient revenue-creation system. The implementation was done on the Force.com platform by leveraging tools like apex, visualforce, test class, Batch processes, data conversion and reports & dashboards. Show less

Was part of a specialist group for data conversation for one of the biggest semiconductor vendors in the world, worked closely with different technology teams in delivering a robust sales and pricing system

The project was for a leading American cloud computing company, where we helped them improve their sales processes using a CPQ system.

Created a CPQ system for the client using Apttus CPQ packages on Force.com platform resulting in significant improvements in the quoting and pricing processes. The system allowed the client to speed up deal cycles, close more deals, automate sales across channels, analyse real-time performance of sales using interactive dashboards and gain visibility… Show more

The project was for a leading American cloud computing company, where we helped them improve their sales processes using a CPQ system.

Created a CPQ system for the client using Apttus CPQ packages on Force.com platform resulting in significant improvements in the quoting and pricing processes. The system allowed the client to speed up deal cycles, close more deals, automate sales across channels, analyse real-time performance of sales using interactive dashboards and gain visibility into changes that could impact revenue targets. Show less

- The project was about extracting and recognizing handwritten characters or machine texts from scanned images which could then be used in a wide variety of applications
- Character recognition was done using a Neural Networks algorithm using two techniques: identity matrices and the other being ASCII codes
- We made use of different types of neural networks namely: feed-forward back propagation network, Elman back-propagation network and the fitting network. The simulation was done using… Show more

- The project was about extracting and recognizing handwritten characters or machine texts from scanned images which could then be used in a wide variety of applications
- Character recognition was done using a Neural Networks algorithm using two techniques: identity matrices and the other being ASCII codes
- We made use of different types of neural networks namely: feed-forward back propagation network, Elman back-propagation network and the fitting network. The simulation was done using MATLAB Show less

- We developed a website for a hospital to leverage hospital services where patient need not perform any paper-work while getting admitted and treated, by providing a seamless application that would handle the information of thousands of patients and provide efficient healthcare services
- We were primarily engaged in Server side scripting using JSP (Java server pages). The interface was built using IBM Tools – WASCE for application server, DB2 as database server