Flights delay

Analyzing flight details and predicting arrival delay.

The goal of this project is to predict flight delays by observing the predictive power of a number of variables on the 2015 domestic flights dataset.

Broadly, the following steps were done to accomplish the project objectives:

• Investigate missing values
• Perform Exploratory Data Analysis: Delays based on airlines, delays based on airports, delays based on distance of the trip, delays based on day of the week.
• Apply linear regression model.

• R-squared is 91.68%.
• Final equation of model is : Balance_Hat = 2.70757128343479 + 0.019111162332402735*LATE_AIRCRAFT_DELAY +0.019317385511321186 * AIRLINE_DELAY+ 0.020272263585622848*AIR_SYSTEM_DELAY+ 0.019478108752049868*WEATHER_DELAY +0.0008544833163888557*DEPARTURE_DELAY

• Github link to the project

Credit Card Fraud

Fit and compare 5 classification models (Logistic Regression, Decision Tree, Random Forest, Bagged Trees, Gradient Boosting) to predict credit card fraud or not.

It is important that credit card companies are able to recognize fraudulent credit card transactions so that customers are not charged for items that they did not purchase..

The original data set comes from Kaggle: https://www.kaggle.com/mlg-ulb/creditcardfraud. The datasets contains transactions made by credit cards in September 2013 by european cardholders. This dataset presents transactions that occurred in two days The original dataset contains roughly 280K observations. I have simplified down to roughly 3000 observations. The aim of this project is to compare a few ML models and compare.

The dataset contains only numerical input variables which are the result of a PCA transformation. The original features and background information about the data is not available due to confidentialy issues. Features V1, V2, … V28 are the principal components obtained with PCA, the only features which have not been transformed with PCA are 'Time' and 'Amount'. Feature 'Time' contains the seconds elapsed between each transaction and the first transaction in the dataset. The feature 'Amount' is the transaction Amount, this feature can be used for example-dependant cost-senstive learning. Feature 'Class' is the response variable and it takes value 1 in case of fraud and 0 otherwise.

In this project, we will fit and compare the following 5 classification models to predict credit card fraud or not::

• Logistic Regression
• Decision Tree.
• Bagged Trees.
• Random Forest.
• Gradient Boosting.

• In terms of F1 score, Logistic regression has the best performance with a score of 0.893, followed by random forest (score =0.891) then boosted trees (score=0.889) then bagged trees (score=0.885). As expected, a single decision tree (score=0.87) doesn't perform too well.
• In terms of AUC score, Boosted tree has the highest AUC score (0.973), followed by Logistic (0.972) and decision tree (0.927) is the worst.

• Github link to the project

Heart Disease

To perform exploratory data analysis and implement a model to predict the presence of heart disease in a patient.

Heart disease describes a range of conditions that affect your heart. Diseases under the heart disease umbrella include blood vessel diseases, such as coronary artery disease, heart rhythm problems (arrhythmias) and heart defects you’re born with (congenital heart defects), among others. In this project report, we will be applying data analysis techniques and statistics approaches and comparing the attributes for classifying whether a person is suffering from heart disease or not, using one of the most used dataset — Cleveland Heart Disease dataset from the UCI Repository. Dataset can be found from Heart Disease UCI: (https://www.kaggle.com/ronitf/heart-disease-uci).

The data has 13 attributes which are independent factors. The ‘goal’ or dependent factor is the presence of heart disease in patients. We need to predict the presence of heart disease by analyzing the correlation between the attributes and the dependent factor.

Techniques involved:

• Data Cleaning
• Exploratory Data Analysis
• Train/Test Split
• Logistic Regression
• Building ROC curce and finding AUC

• Final AUC model : 0.8524

• Tools : Python, Numpy, Seaborn, Matplotlib, Statsmodel, Sklearn

• Github link to the project

Grok Investments

Research, development and analytics of Grok Investment website.

• The objective of the project was to redesign Fidelity's website homepage to make it more attractive and engaging to millennials.
• Performed search engine analytics, social media analytics, competitor website analytics, and Fidelity website audit to identify top trends, products, features, and search queries. The data gathered from the search engine platform (Google) and social media platforms (YouTube, Twitter) was obtained through web scraping tools/algorithms.
• Developed a website, “grokinvestments.com", based on the data collected in the research phase. The prominent features on the website are Chatbot, quiz, curated product list, and learning portal. •Product development tools- React and Google Firebase.
• Collected data on user engagement and analyzed information gathered from google analytics, A/B/n test and survey results.
• Analyzed the data through topic modeling, identifying correlations, and top trends.
• Analytics/Testing tools: Google optimize, Google analytics, Tableau.
• Increased SEO score from 82 to 100, increased accessibility score from 79 to 84, and received an overall rating of 9/10.
• Identified top advising sources, the relationship between income and investment risks, and the relationship between age and investment risk tolerance.

• Website link
• Report link

Airbnb

Database design, development and customer behavior analysis.

The project focus was to understand how a company as an online marketplace offering homestays utilizes a digital platform with their Database and Relational Database Management System model to manage its property listings and stakeholders (Guest and Host only in this case). Through this data repository, we explore and analyse how AirBnB enables both Hosts and Guests to make informed, data-backed decisions for interaction choices in both online and offline world and also track their organization’s periodic and overall performance. Our focus is to make scalable data by designing RDBMS for a city (San Francisco, California, USA) that can be expanded further in a similar fashion into states and countries to cover their area of presence.

• Github link to the project

Stores data project

Store level prediction based on MAPE.

Predicted the store level ( Global, Division , Department) at which the Mean Absolute Percent error is the least for the lift obtained through promotional and baseline sales. Concepts/techniques used : Cross validation, Non-linear weighted regression, data manipulation.

• Github link to the project

Project Name

Lorem ipsum dolor sit amet consectetur.

Use this area to describe your project. Lorem ipsum dolor sit amet, consectetur adipisicing elit. Est blanditiis dolorem culpa incidunt minus dignissimos deserunt repellat aperiam quasi sunt officia expedita beatae cupiditate, maiores repudiandae, nostrum, reiciendis facere nemo!

• Date: January 2020
• Client: Threads
• Category: Illustration

Close Project

Project Name

Lorem ipsum dolor sit amet consectetur.

Use this area to describe your project. Lorem ipsum dolor sit amet, consectetur adipisicing elit. Est blanditiis dolorem culpa incidunt minus dignissimos deserunt repellat aperiam quasi sunt officia expedita beatae cupiditate, maiores repudiandae, nostrum, reiciendis facere nemo!

• Date: January 2020
• Client: Explore
• Category: Graphic Design

Close Project