OKCupid Project

Introduction

In this project, we take almost 60,000 OK Cupid dating profiles, and we look to:

• run various statistical analysis to give us insight into the profile
• create data visualizations to easier see our analysis, and see patterns and trends in the data
• create various machine learning models to accurately predict variables using a users basic inputs

Dataset

Scraped from OKCupid, uploaded here.

Model

Three machine learning models were used:

• Linear Regression
• K Neighbors Classification
• Decision Tree Classification

Summary

This project was intensive in data cleaning as a lot of the data was either missing, or non-sensical. Automating the data cleaning process and the analysis on the newly cleaned data was easily the longest/hardest part here. In terms of modeling, our most accurate model was a decision tree with almost 80% accuracy (and a macro 80% F1 score). Overall - the decision tree process was the most efficient in our selection of our output variable.