Introduction

Restaurants and bars play a significant role in shaping neighborhoods within a city. In many cases these establishments can shape the environment and reputation that a neighborhood has. A new, popular restaurant and a unruly, noisy bar can each affect its surrounding area for better or for worse. This page analyzes data regarding the distribution of restaurants and bars around the city of Syracuse, aggregating key indicators to the census tract level.

Overview

The page is comprised of six sections. First, the introduction which contains an overview of the data and how to modify it. Second, the setup required to conduct the analysis. This includes loading the original yelp dataset and census tract shapefiles. Third, a couple of reusable functions are defined in order to minimize the amount of code and the number of changes required for future modifications. Fourth, the data is aggregated to the census tract level. The focus of the analysis is on key indicators (rating, reviews, concentration of establishments) but there are numerous other indicators in the original dataset that could be explored in the future. Fifth, a sample of the output is displayed before generating the output file. Sixth, the appendix which contains information regarding the source data.

Future Enhancements

In order to expand upon this analysis, additional columns can be added to the aggregated output data. This can be done by copying an existing aggregation section, prefixed with “Aggregation:” and modifying the two sections of code in the following way:

In the first code section, the original data stored in the “dat” data frame and creates a new column in teh output aggregated data frame enttitled “yelp.data”. To modify, follow these steps:

Change the name of the column in the yelp.data data frame from “RATING” to a new name.
In the aggregate.tract function change the column you want to aggregate in the source data frame entititled “dat”.
Conditions can be added to subset the original data frame as is done in the example below. However, the dat$FIPS condition must match.

yelp.data$RATING <- aggregate.tract(dat$RATING[dat$RATING > 0], dat$FIPS[dat$RATING > 0], "mean", yelp.data$TRACT, 0)

In this code section, the data is plotted in a map by census tract. To modify, follow these steps:

In the plot.census function call, change the first argument to use the new column name that was created in the above section.
Optionally, modify the fifth argument (integer value) to change the number of decimals in the legend.
Optionally, rename the title of the plot, last argument, to a more applicable name.

plot.census(yelp.data$RATING, yelp.data$TRACT, color.slice, color.ramp, 1, "Average Rating by Census Tract")

Setup

Load Packages

library(maptools)
library(sp)
library(geojsonio)
library(RCurl)

Set Parameters

#URL for census tract shapefiles for the city of Syracuse
url.shapefile <- "https://raw.githubusercontent.com/lecy/SyracuseLandBank/master/SHAPEFILES/SYRCensusTracts.geojson"

#Parameters for plotting aggregated data across census tracts

#Color range from low to high
color.range <- colorRampPalette( c("steel blue","light gray","firebrick4" ) )

#Number of buckets data will be sliced into
color.slice <- 10

#Categories of establishments considered bars or nightlife
bars <- c("Bars", "Beer Bars", "Gay Bars", "Dive Bars", "Beer, Wine & Spirits
", "Pubs", "Breweries", "Sports Bars", "Lounges", "Wine Bars", "Nightlife", "Music Venues", "Dance Clubs", "Pool Halls")

Reads in geojson shapefiles for syracuse census tracts.

syr <- geojson_read(url.shapefile, method="local", what="sp")
syr <- spTransform(syr, CRS("+proj=longlat +datum=WGS84"))

Load Yelp Dataset and prep data.

setwd("..")
setwd("..")
setwd("./DATA/PROCESSED_DATA")
dat <- read.csv( "yelp_processed.csv", stringsAsFactors=FALSE )

dat$RATING[is.na(dat$RATING)] <- 0
dat$PRICE_HIGH[dat$PRICE_HIGH == 999] <- 90

head(dat)

Functions

Function to aggregate data.

Arguments:

dat.agg - vector of data to aggregate
dat.groupby - vector of data to group dat.agg by (vector length must be the same)
agg.fun - function to be applied in aggregation (sum, mean, etc)
dat.tract - vector of FIPS codes for census tracts in Syracuse
dat.default - default value to apply to census tracts with no data available

aggregate.tract <- function(dat.agg, dat.groupby, agg.fun, dat.tract, dat.default){
  temp <- aggregate(dat.agg, list(dat.groupby), FUN=agg.fun)
  tract.order <- match(dat.tract, temp$Group.1)
  agg.result <- data.frame("TRACT" = dat.tract, "DAT" = rep.int(dat.default, 55))
  agg.result$DAT <- temp$x[tract.order]
  agg.result$DAT[is.na(agg.result$DAT)] <- dat.default
  return (agg.result$DAT)
}

Function to plot aggregated data across census tracts.

Arguments:

dat.agg - vector of aggregated data to plot
dat.tract - vector of FIPS codes for census tracts in Syracuse
slices - the number of buckets to group data for colors
rcolor - color ramp to apply to the buckets of data
places - number of decimal points to round legend to
title - title of the plot

plot.census <- function(dat.agg, dat.tract, slices, rcolor, places, title){
  color.vector <- cut(rank(dat.agg), breaks=slices, labels=rcolor)
  color.vector <- as.character(color.vector)
  color.order <- color.vector[dat.tract]

  plot(syr, border="white", col=color.order, main=title)

  legend.ranges <- get.legend(dat.agg, slices, places)
  legend("bottomright", bg="white", pch=19, pt.cex=1.5, cex=0.7, legend=legend.ranges, col=rcolor)
}

Function to generate labels for the legend on each of the plots. Note: this is only called from the plot.census function defined above.

Arguments:

agg.dat - vector of aggregated data
slices - number of buckets that the data is split into for plotting
places - number of decimals to round the legend to

get.legend <- function(agg.dat, slices, places){
  dat.max <- max(agg.dat)
  dat.min <- min(agg.dat)
  dat.range <- (dat.max-dat.min)/slices
  dat.last <- dat.min
  legend.ranges <- vector(mode="numeric", length = slices)
  
  for( i in 1:slices ){
    range.min <- dat.last
    range.max <- dat.last + dat.range
    legend.ranges[i] <- paste(round(range.min, places),"-",round(range.max, places))
    dat.last <- range.max
  }
  return(legend.ranges)
}

Data Aggregation

Convert to spatial points and perform spatial join to census tract shapefiles.

dat$coord <- dat[ ,c("LONGITUDE","LATITUDE")]
yelp <- SpatialPoints(dat$coord, proj4string=CRS("+proj=longlat +datum=WGS84"))

Plot restaurants and bars across Syracuse.

par(mar=c(0,0,0,0))
plot(syr, col="grey90", border="White")
points(yelp, pch=20, cex = 0.7, col="darkorange2")

Match yelp data to census tract. Store census FIPS code in yelp dataset.

tracts.yelp <- over(yelp, syr)
dat$FIPS <- tracts.yelp$GEOID10
dat$COUNT <- rep.int(1, 932)

Create a new data frame to store aggregate information.

yelp.data <- data.frame("TRACT" = syr$GEOID10, "YEAR" = rep.int(2017, 55))

Store color ramp to plot aggregated data by census tract. Ramp scales from low to high based on parameterized color range and number of slices.

color.ramp <- color.range(color.slice)