Projects

Machine Learning (4) | Data Analysis(4) | Operations Research & Economics (1)

My projects are based on experimenting various supervised and unsupervised machine learning algorithms. The goal was to first understand how algorithms work and then running the code. You can find more theory in my blogs, where I have tried to explain the working of machine learning algorithms from optimization point of view.

Languages Used: Python, R, Apache Spark

• ML for e-Commerce – Predicting customer's purchasing intention using a Google analytics dataset
  An important dataset for e-Commerce space is data from Google Analytics. One of the major issue faced in the real world is imbalanced datasets where we have data from one class dominating the other. In this project, I try classifying if the customer will generate revenue or not from the click data, i.e., previous website visits and duration of visit etc, while also providing solutions to deal with an imbalanced dataset.
  
  Code


• ML for IoT - Classifying muscular movements into hand gestures
  Coming from a Mechanical Engineering background and with previous experience with the robotics team, working with sensors always fascinates me. The dataset consists of multiple data measurements from sensors attached to arms for recording the hand movement. Using Logistic Regression and KNN classifiers I try to identify the hand gestures.
  
  Code


• ML for IoT - Classification of room occupancy using sensor data
  Another commonly used IoT dataset, I have tried identifying if the room is occupied or vacant. The dataset consists of recordings from multiple sensors placed within the room to record the environmental conditions. In the project, I have used two of the most basic classification methods, Naive Bayes and Logistic Regression.
  
  Code
  


• ML for Service - EDA and gender classification from movies ratings using Spark
  Knowing your customer is an important factor for business growth. In this project, I analyze the dataset to gain insights on ratings, rankings and customers. Additionaly, a common problem faced in the service industry is customer knowledge. At times people do not fill up the optional fields in the forms, which result in loss of customer knowledge. To address this, I also predict the gender of the customer from the ratings and the genre of the movies.
  
  Code

An important part of Data Science is data cleaning and Visualization. Even though the machine learning algorithms can give us the prediction, not all projects require machine learning. In this section, I have done few data cleaning and Visualization projects with various tools.
Tools Used: Python, R, Apache Spark, Tableau

• Exploring New York Times Best Sellers book data with Spark SQL
  This is from my course, Mining of Large Datasets using Spark. I have used SQL in Spark to compute various tasks like joining and computing various aggregations to explore sales and ratings.
  
  Code


• Data mining and visualizing the New York taxi traffic for mobility and customer patterns using spark and matplotlib
  Another commonly used dataset is the New York taxi dataset. Understanding the mobility of riders like peak periods, traffic areas and fare policies, I have done analysis on this dataset using Spark and matplotlib for visualization.
  
  Code|Report


• Dealing with missing data – A case study using Bridge Classification dataset
  Not all the datasets are cleaned and ready for predictive analytics. Missing values are one of the most common problem in datasets. In this experiment, I have discussed and tried various methods to deal with missing values.
  
  Code


• Implementing FP-Growth algorithm to identify commonly used ingredients in different cuisines
  While supervised learning is helpful in knowing the future, I believe unsupervised learning is equally helpful in understanding the current data. One of the methods within unsupervised learning is Association Rules. In this project, I have experimented FP-Growth in Spark to understand if there are any patterns between the cuisine and ingredients.
  
  Code

Industrial Engineering is all about designing and improving processes. Supply chain is one of the most complex and capital intensive processes in the real world due to which it also becomes one of the most sought after research topics in Industrial Engineering. After taking advanced courses in Applied Operations Research and Logistics and Supply Chain Management, I thought of sharing few of my coursework in the field.

• Dynamic programming approach for optimization electricity purchasing and inventory management
  We applied one of the optimization methods, dynamic programming approach, using python to find optimal purchasing quantity of electric power from the grid. This approach can further be extended to various supply chain problems to identify optimal order quantities to be placed.
  
  Code