Citrics

You can find the project Here. The API Docs for the DS portion can be found here Here

High level overview presentation.

Deep dive into cleaning the data.

Datasets used

Contributors:

Scott Maxwell	Matthew Sessions	Luke Townsend	jimmy 'Zeb' Smith

Steven Reiss	Stephanie Miller	Amy NLe	Robert Tom

Project Overview

Citrics provides statistics on 28,925 different locations in the United States that are available for viewing. This was created with a team of web developers and data engineers. These statistics include information about housing prices, employment, industry, lifestyle and much more, sources are listed below.

Links to team documents:

Trello Board

Product Canvas

Live Front End/URL

Tech Stack

• Python
• Flask
• Docker
• Jupyter Notebooks
• Mongo DB
• AWS Elastic Beanstalk/Amplify/S3/Route 53
• AWS PostgreSQL

Predictions

The following models are using a K-Nearest Neighbors KD-Tree algorithm from the Scikit-Learn Python Library

Clicking on the links for each model will take you to the .py file for that model within this repo

Housing Model:

Features & Metrics Used:

• Median Rent
• Occupants per room
• Housing by bedrooms
• Vacancy Rate
• Rent Pricing
• Historical Property Value
• Historical Property Value Growth by %

Industry Model:

Features & Metrics Used:

• Industry Types
• Health insurance
• Salary
• Commute & travel time
• Retirement
• Unemployment

Culture Model:

Features & Metrics used:

• Education
• Language
• Ethnicity
• Birth Rate
• Population

Reverse User:

A recommendation questionnaire that supplies the user with a recommended city Features & Metrics used:

• Population
• Income
• Monthly Housing
• Temperature Preference
• Industry*

*Note that Industry is optional on the website, due to lack of adequate data for all cities within the database.

Time series-housing:

Features & Metrics used:

• Education
• Language
• Ethnicity
• Birth Rate
• Population

Time series-Industry:

Features & Metrics used:

• Education
• Language
• Ethnicity
• Birth Rate
• Population

Note for Library Conflicts:

• AWS' Elastic Beanstalk has a hard time runing Numpy and Scipy. These libraries power Sklearn. Also, the joblib library had a hard time running models that were trained on different operating systems. Once we found models that worked, we exported the code to a python script and ran it on a Linux based machines running Python 3.6. We then used Docker/Dockerhub to contain and ship our flask app, and then connected it to AWS to test in Elastic Beanstalk. These steps allowed us to seamlessly deploy predictive models.

Data Sources

• Census Bureau
• Zillow Housing
• Longitude & Latitude
• Population Growth
• Weather Data
• Bureau of Labor and Statistics

Notes: How we pushed our data to MongoDB

Python Notebooks

Documation Notebook

Fixing City Names:

• Fixing City Names_1
• Fixing City Names_2

Different types of data and sources

• Longitude & Latitude
• Population Growth
• Weather Data

Housing Data

• Cleaning Zillow Data
• Price Checking
• Mapping Lat & Long to Zillow

Models (for suggesting similar cities):

• Housing KNN Model
• Industry KNN Model
• Culture KNN Model

Models (Time series):

• Housing Forecasting
• Industry Forecasting

Model (Questionnaire):

-KNN Unclassified

How to connect to the data API

You can find documentation for the API here