vmueller13 / project4_sanantoniorestaurants

Final project to test the hypothesis: All the good taco places in San Antonio are located within 410.

Restaurant Scores in San Antonio

Table of Contents

• Introduction
• Features
• Getting Started
• Usage
• Processes and Technologies
• Challenges

Introduction

The San Antonio Restaurants Dashboard is a web-based application that allows users to explore and visualize data about restaurants in San Antonio. It provides various visualizations, such as a map view, charts, and a restaurant list. Users can interact with the dashboard to view restaurant details, scores, and price-range. By using the layer feature of the map, users can diplay restaurants by cluster of scores, thereby making it easy to find the highest scored restaurants in their own neighborhood! Additionally, we have added different tabs to the site that allow us to answer some questions about the dataset.

Features

• Interactive map showing restaurant locations
• Filter layers on the map to show restaurants grouped by score
• Restaurant list with sorting and filtering options

Getting Started

To get started with the San Antonio Restaurants Dashboard, follow these steps:

1. Clone the repository: git clone
2. Navigate to the project directory: cd
3. Install the libraries: psycopg2, chart.js
4. Run the included code to set up your database and import the csv in pgadmin4s
5. Change the information in the config.py file to your username, password and port for pgadmin4

Usage

1. Start the application:
2. Open your web browser and go to http://localhost:5000 to access the dashboard.
3. Explore the various visualizations and interact with the features.

Processes and Technologies

HTML, JS, css, postgresSQL and Python API Flask App

The first part of this project neccessitated the following libraries and techniques: flask, psycopg2, d3.json. We created a database using postgresSQL to hold the restaurant data and pulled this database through a Python API Flask App in order to visualize the data on a Leaflet map. In coordination with our HTML, JavaScript and css files, we created the dashboard as detailed above. With this dashboard in place, we moved on to the second part of the project.

Machine Learning

To begin this part of the project, we imported the libraries and techniques: sklearn.metrics, sklearn.cluster, kMeans, StandardScaler, pandas, matplotlib, imblearn.over_sampling, RandomOverSampler, Seaborn and numpy. We first cleaned up the data in Jupyter Notebook; dropping NaN values and unecessary columns so that we had a clean dataset at 558 rows and six columns. With this cleaned dataset, we were able to use get_dummies to convert categorical columns to numerical values and conduct a logistical regression. An assessment of the model gave us a Balanced Accuracy Score: 0.5, Confusion Matrix: [[ 0 9] [ 0 131]] and a Classification Report as seen below:

In order to further calibrate and test our model, we decided to create a logistic regression model with resampled training data. Using RandomOverSampler, we were able to achieve a Balanced Accuracy Score: 0.5250212044105174, Confusion Matrix: [[ 1 8] [ 8 123]] and a Classification Report:

Since there wasn't any significant improvement to the balanced accuracy score or the classification report, we recommend using the original Logistic Regression model for further analyis. To continue, we used kMeans to cluster the data. In looking at the clustering we noticed two significant insights. First, that the model clustered the restaurants based on the score (i.e. 4-7-5.0 scores in the same cluster) and second, that the model also took into account the price of the restaurant as well (i.e. $ mostly in one cluster, $$ mostly in another cluster etc.)

Questions from the Dataset

Ultimately, we used these clusters to answer some questions about the cuisine available in San Antonio.

• Question 1: What zip code should you stay in San Antonio to be close to delectable food?
• Question 2: Are all the good Mexican food places actually inside 410?
• Question 3: What are the highest scored food categories in Cluster 0: Top Shelf (4.7-5.0)?

After clustering our data we created these visualizations to answer these questions:

Question 1

This bar chart shows the zip codes that are contained in the top shelf cluster (cluster 0). The zip code with the highest number of highly rated restaurants is 78258 or Stone Oak. This is a high end neighborhood in San Antonio where property is valued at half a million dollars or more. The second highest zipcode is 78209. Alamo Heights/Terrell Hills is right off of Broadway, near a lot of new development in the city, we now know that this includes good places to eat too! The third highest zip code is 78216. This zip code is near Ridgeview and Thousand Oaks. With the shopping center here as well as other quality establishments, this zip code area is also near the airport, making it a prime spot for visitors to eat, shop and stay without missing their flights!

Question 2

This question explores a popular saying in San Antonio that all the good places to each are inside the 410 loop. While this might have been true at one time, this is not longer the case. Out of the 104 Mexican food places in the Top Shelf cluster, only nine of the restaurants are located inside the 410 loop. Therefore, we conclude that the development of San Antonio as the city grows, not only brings good food places outside the loop 410, but all over the city.

-- San Antonio Zip Codes

Question 3

This question explores what kind of cuisine does well in San Antonio and could be used by future entrepreneurs or current foodies to find out what cuisines do well in San Antonio. Surprisingly, the highest rated category is 'Burgers, American, Sandwiches' which beats out our predictions of BBQ or Mexican cuisine.

Challenges

Initially, when we started the machine learning step of the project, we started with linear regression. After receiving a negative r value, we realized that we had discrete values instead of continuous and needed to pivot to logistic regression instead. The other major challenge we faced was incorporating each of our visual elements into the tabs on the live web server. After troubleshooting as a team and revising the file path (many times!) we were able to successfully include all of our visuals.

As always, teamwork, communication and collaboration have helped us face and conquer each of these challenges (and more!). --The Cool Cat Coders: Asfia, Matty, Saroja, Taniya, Veronica