Finding Donors (CharityML)

Date:

14 Sep 2018

Site link:

Jupyter Notebook URL

Algorithms:

Gaussian Naive Bayes (GaussianNB), Decision Tree Classifier, C-Support Vector Classification

Technology:

Python, Jupyter-Notebook

Problem Statement: Build an algorithm to best identify potential donors.

• My goal was to evaluate and optimize several different supervised learners to determine which algorithm will provide the highest donation yield while also reducing the total number of letters being sent.
• Implemented a pipeline in Python that will train and predict on the supervised learning algorithm given. I used the Grid Search method to tune the parameters of all algorithms and Gradient Boosting Classifier to extract features importance.

Indian Paper Currency Prediction

Date:

June 2020

Code:

GitHub Repo

Demo:

https://indian-currency-prediction.herokuapp.com/

Algorithms:

Convolutional Neural Network (CNN)

Technology:

Python, Jupyter-Notebook, Flask

Problem Statement: Build an app to prediction indian paper currency.

• Collected the images from search engines and trained a Convolutional Neural Network (CNN) model to predict the 7 types of Indian paper currency i.e. 10, 20, 50, 100, 200, 500, 2000.
• Deployed the machine learning model on Heroku with Flask back-end and secured the app with CSRF protection.

Image Classification with PyTorch (Deep Learning)

Date:

13 Nov 2018

Site link:

Github Repo URL

Technology:

PyTorch Framework, Python, Jupyter-Notebook

Problem Statement: Building a command-line application to predict flower class along with the probability.

• AI image classification and machine learning utilizing the PyTorch framework.
• Used transfer learning on pre-trained architectures including vgg11, vgg13, vgg16, vgg19, densenet121, densenet169, densenet161, and densenet201.
• Trained dynamic neural networks in Python with GPU acceleration with 85% accuracy.

Customer Churn Prediction (PySpark)

Date:

09 June 2019

Site link:

GitHub Repo URL

Blog:

https://medium.com/@rowhitswami/customer-churn-prediction-of-a-music-app-using-pyspark-d65b8f5be047

Algorithms:

PySpark, TF-IDF, SDGClassifier

Technology:

Python, Jupyter-Notebook

Problem Statement: Customer Churn Prediction from a Music App Spark

• Used PySpark to analyze the data of a fictional music app Sparkify to identify the factor affecting the customers who are most likely to churn.
• Trained machine learning model on IBM Cloud with the accuracy of 83.87%

StackOverflow Survey '17

Date:

20 Feb 2019

Site link:

GitHub Repo URL

Blog:

https://medium.com/@rowhitswami/4-crunching-facts-of-stack-overflow-developer-survey-17-441eb082331f

Algorithms:

Exploratory Data Analysis

Technology:

Python, Jupyter-Notebook

Problem Statement: Analyzing the public data of StackOverflow Survey 2017

• Analyzed the StackOverflow public data of over 64,000 developers around the globe for the year 2017.
• Answered questions like, how education may influence the salary, gender ratio of developers across the world, the rate of increase in salary with years of experience and does more language implies more salary hike in IT sector.

Identify Customer Segments with Arvato

Date:

24 Dec 2018

Site link:

Jupyter Notebook URL

Algorithms:

KMeans

Technology:

Python, Jupyter-Notebook

Problem Statement: Identify Customer Segments with Arvato Dataset

• The data and design for this project were provided by Arvato Financial Services. I applied unsupervised learning techniques on demographic and spending data for a sample of German households.
• Preprocessed the data, apply dimensionality reduction techniques, and implement clustering algorithms to segment customers with the goal of optimizing customer outreach for a mail order company.
• The objective was to find relationships between demographics features, organize the population into clusters, and see how prevalent customers are in each of the segments obtained.

Disaster Response Pipeline

Date:

13 Apr 2019

Site link:

Github Repo URL

Technology:

Python, Jupyter-Notebook

Problem Statement: Building a machine learning pipeline

• Analyzed disaster data from Figure Eight to build a model for an API that classifies disaster messages.
• Created a machine learning pipeline to categorize the events so that it can send the messages to an appropriate disaster relief agency and deployed as a web app.

Twitter Sentiment Analysis using Hadoop

Date:

Oct 2017 - Mar 2018

Client:

Personal

Technology:

Hadoop, Flume, Hive, Twitter Streaming API, Python

Twitter, one of the largest social media site receives tweets in millions every day. This huge amount of raw data can be used for industrial or business purpose by organizing according to our requirement and processing. This project provides a way of sentiment analysis using hadoop which will process the huge amount of data on a hadoop cluster faster in real time

Data Analysis (Freelancing)

Date:

24 Apr 2018

Site link:

GitHub Repo

Client:

Lcodi56

Technology:

Python

Sales data was given in CSV format and the task of this project was to derive valuable insights from the raw data, like:

• Which product was most sold?
• What payment modes were used for purchasing products?
• What is the most common payment method for the United States?
• What was the earliest time of the day a transaction occurred?
• Were there repeat customers? Discuss possible issues.

Data Analysis (Freelancing)

Date:

04 May 2018

Site link:

GitHub Repo

Client:

Lcodi56

Technology:

Python

The dataset contains information of people die from "Diabetes Mellitus" between 1999-2015. The task of this project was to derive valuable insights from the raw data, like:

• In what state has the most deaths occurred?
• Over what period (start and end) was the data collected?
• What were the total number of deaths for 2006?
• What state had the least deaths in 2001?
• How many people die from "Diabetes mellitus" over the entire reporting period?